lOMoARcPSD

Appunti, tutte le lezioni - Operations management - a.a.

2015/2016

Operations management (Politecnico di Milano)

 lOMoARcPSD

7/3/16
EXAM

Written exam in July:

- 1 ex: LEAN
- 2 ex: QUEUEING
- 2,3,4 days to correct the exam
- 3,4 days after the written exam, the oral exam with theoretical questions

FINAL MARK: 50% written exam, 50% oral exam. (30 written, 18 oral: 24)

Mid-term written exam: 7 MAY in the morning

- 1 ex (queueing) + written questions to answer
Final mid-term written exam: first week of July
- we have to do only the second part of the written exam (only lean exercise)
- oral exam about the second part of the program

OPERATIONS MANAGEMENT: introduction

Why operation management is important? (Yahoo answer)

Operations and processes

All processes are

INPUT-TRANSFORMATION-OUTPUT
systems that use transforming
resources to work on transformed
resources in order to produce products
and services.
 lOMoARcPSD

Operations and functions involved

Functions involved in Operations are product design, engineering, purchasing, production, distribution, planning,
after sale support and quality.

Operations management evolution

Operations management has expanded from treating only the core production processes in manufacturing
organisations to include service organisations (a), non-core operations processes (b) and processes in other
functions such as marketing, finance and HRM.

Operations management organisational depth

Operations and process

management requires
analysis at three levels:
the supply network, the
operation and the
process.

In fact, products are

assembled in many
number of plants, like the
airplane which is
assembled in more than
100 plants.
 lOMoARcPSD

WHY THIS COURSE?

1. Operations are central to the company’s business and mankind’s sustainability: in term of economic
sustainability, because each company needs to make money, environmental sustainability, because each
company tries to produce green products and social sustainability, because companies take care of their
commuting employees, trying to merge their needs and those of the company in order to reduce the
amount of time spent commuting.
Relevance in a company:
1. Spending on material purchase can be 50-80 % of turnover/total cost
2. Spending on personnel can be 30-50 % of turnover/total cost (50-80 % in services)
3. The annual cost of the facilities (plants, information systems, etc.) can be 30-40 % of turnover/total cost
4. Investments are high and are binding for a long time, so making the right choices is crucial:
o a site can cost tens of millions of € and take 2 years to be realized
o develop distinctive competences even more
2. Skills mismatch: the skills in the Operations area are critical and strongly impact the competitive advantage.
For example, there are manufacturing positions that need particular skills and sometimes people looking for
jobs do not have them. Skills are important both in industrial companies and in the service ones. People are
even more important in service companies, where the service is delivered by workers, who need to train a
lot to guarantee a certain quality of the service.
3. Operations contribute to the company strategy: operations are used to implement companies’ strategies. If
operations cannot implement a strategy, this means that the strategy is not feasible.
Strategy has two aspects:
o Where to go
o How to get there
4. Services: services are a special world and a rapidly growing ones. Operations in services are at dawn:
companies are struggling to find service-specific skills.
EX) Carpigiani makes machines for ice cream: they are very valuable products and they last a lot. But how
can the company grow? Adding features or services to the machine, like adding a sensor that inform the
customer when to clean the machine.
EX) GDP: Gross Domestic Product. GDP percentage in services has increased during the years in all the
countries. The same happened to people employed in services (percentage)
Integrate Manufacturing and Services:
A good part of services is due to outsourcing policies
Over 70 % of services is targeted at industrial companies or created by industrial companies
But...
Services are intrinsically different from the production of goods
Success in manufacturing products does not necessarily mean to be good in delivering
services
More and more companies develop a product-service integrated offer
1. LEAN manufacturing: it represents a strong innovation which is taking place in the management of
companies and organizations. it started in Japan in Toyota, which had to enter the European market being
cheaper than the incumbents. To do this, it had to produce a very reliable product with the lowest costs,
being very efficient and so competitive at global level.
1950: Toyota produces in 1 year as many cars as GM produces in 1 day
1970-1980: Toyota grows rapidly in the USA and it’s one of the most appreciated trademarks: great worry
for the 3 sisters (in Europe imports are subject to quota)
Causes interpretations (1970-1980)
• Dumping
• Strong dollar
2007: Toyota is the world's largest car manufacturer

Operations management purpose

To provide the knowledge to manage and improve operational processes (operations), at strategic, tactical and
operational levels, in both manufacturing and service companies.
 lOMoARcPSD

TRENDS IN THE OPERATIONS: TOWARDS SUSTAINABILITY

8/3/16
BACKGROUND: THE CHANGING ENVIRONMENTAL AND SOCIAL MEGATRENDS

Megatrends
Europe’s manufacturing technology platform Manufuture has identified 8 megatrends:
1. Changing demographics
2. Globalisation & future markets
3. Scarcity of resources
4. Climate change
5. Dynamic technology & innovation
6. Global knowledge society
7. Mass customisation
8. Sharing global responsibility
These have a considerable impact & drive structural trends in nearly all manufacturing sectors.

CHANGING DEMOGRAPHICS
o Future generations: future generations in 2050 will be characterized by an increase in number of people,
especially in Africa and an increase in the age of people. In the European Union the number of elder people
will increase significantly, thanks to better life conditions that make people live longer. One of the main
consequences is the pension issue: to give pensions to a higher number of people we have only two
solutions:
- increase the retirement age
- increase the productivity
o People move to cities: a wold map describes cities with more than one million inhabitants between 1975 and
2025. It shows that people are moving from rural areas towards cities to find jobs. Cities become bigger year
over year and for this reason, people spend too much time to commute.

Is this demographic evolution a sustainable trend? People need water, food, many extra resources and to get them
they also need economical and political stability.

SCARCITY OF RESOURCES
o Demand for most resources has
grown strongly since 2000 and this
trend is likely to continue to 2030.
Water consumption will be higher
than food, because water is used
also in industries.

o Commodity prices have increased

sharply since 2000, erasing all the
declines of the 20th century. Every
time there was a global event, prices
increased/decreased according to the
event (World Wars, Oil Shock 1970s
when companies introduced oil with
too low prices and there was no oil
enough in the market with respect to
the demand, so price increased. Oil
price is down now).

NOTA BENE: Systems are interconnected in a

global arena (everything has a global
dimension).

9/3/16
 lOMoARcPSD

o Atlas Copco is a multinational company which is developing technologies to extract raw materials from the
mines. In order to support the higher demand of materials, also technology needs to be improved.

o RARE EARTH ELEMENTS: where the raw materials are? What kind of raw materials are important?
Some countries are luckier than others, for example Chile and Canada are very lucky for copper. Despite of
the quantity, some countries (Australia, Brasil, China, etc) are lucky for the presence of strategic elements,
like Germanium and Titanium which are rare elements used for transistors and batteries for smartphones.
Rare earth elements are a collection of 17 chemical elements, including Neodymium. China is the luckiest
country with the highest number of rare earth elements, together with India, Brazil and Malaysia and for this
reason it has a very strong commercial power. The real problem is the the ability to extract those materials.
(Access to natural resources has always been a reason to fight, like World Wars).

EX) Cell phones and other goods are composed of many components, so how can we reduce the risk of accessibility
of materials/components? We use materials easily accessible, already present in the production countries.

o URBAN MINING: in order to reduce the risk of low accessibility to raw materials, the most popular approach
is to extract materials directly from the existing products that are at the end of their lifecycle. Today
recycling is becoming more and more popular and necessary to make production sustainable and to
overcome the problem related to the extraction of raw materials.

o CIRCULAR ECONOMY: when a company uses as raw materials to produce new products the same materials
present in the final products which are at the end of their lifecycle and are recycled. Companies adopt circula
economies to reduce the supply risk, to become independent from competitor’s decisions, from the market
and from suppliers.

We could run out of many elements within the next 20 to 100 years due to the demographic growth and growing
demand of goods, so companies need to be more productive.

WATER CONSUMPTION
o World’s water supply: about 97.5% of all water on Earth is salt water. Only 2.5% of all the water on Earth is
fresh water. This is still enough, even if it is a very small percentage and it is composed of:
70% of fresh water is frozen in Antarctica and Greenland icecaps
Most of the remaining fresh water (30%) isn’t iced, but it lies too deep underground to be accessible
or exists as soil moisture
Only 1% out of 2.5% earth’s fresh water is available for withdrawal and human use
o Fresh water availability: the population is spread all over the world and water should be where humans are.
MAP 1: Freshwater availability in term of cubic meter per person and per year. This is not the right figure to
look at, because we should consider the improved water.
 lOMoARcPSD

MAP 2: Access to an improved water source in % of total population. In this case we consider the litres per
person of improved water, water that people can drink, which is more significant in term of availability of
scarce resources.
EX) Africa is growing a lot in term of population and the presence of improved water is very critical, due to
deserted areas. Moreover, African countries are political and economical instable, so Africa is not a good
investment opportunity.

o Water consumption:
The grey band represents the difference between
the amount of water extracted and that actually
consumed. Water may be extracted, used, recycled
(or returned to rivers or aquifers) and reused
several times over. Consumption is final use of
water, after which it can no longer be reused. The
fact that extractions have increased at a much
faster rate is an indication of how much more
intensively we can now exploit water. Only a
fraction of water extracted is lost through
evaporation.

EX) How many litres of water you need to grow and

produce one kg of rice?
How many litres of water you need to grow a cow
and produce a kg of meat?
The water drunk by the cow and the water needed for the land where the cow eats is a huge amount. There will be
not enough water in the future, so it is not a sustainable practice.

o Numbers:
50% is the number of people who don’t have access to the quality of water available to the citizens
of Rome 2000 years ago
40 billion is the number of hours spent each year in Africa due to the need to collect and haul water

o Coca Cola case: there was a plant of Coca Cola in Kala Dera, India, which was very close to water areas in
order to extract water from the ground. This practice led to a decline of the ground level and a critical
reduction of the amount of water available for the population. Indian people had serious problems in
extracting water from the ground, because they had go deeper than before. They went to court, claiming for
the terrifying impact of Coca Cola plant practices on local communities. Coca cola lost the cause.

This is an example of social sustainability, how companies can impact human lives.
 lOMoARcPSD

Sustainability in Manufacturing
Manufacturing is responsible for energy
consumption and CO2 emission, but it has also
the opportunity to reduce its impact developing
and applying new technologies and best
practices. In fact, new technologies and best
practices could save between 18% and 26% of
current primary energy use in global industry.
Sustainability can be used as a strategy by
companies in order to increase revenues (today
people are more and more interested in green
products) and to reduce costs and environmental
impact.

Productivity opportunities
How to be more efficient? There are many opportunities in term
of energy consumption, water consumption, land and steel
usage. Developing countries account for 70 to 85 percent of
productivity opportunities, so they have huge opportunities,
while developed one less. This is due to the fact that developing
countries have lower prices of raw materials, land, labour, so they are more productive, while developed countries
use these elements less because of the impact of their cost.

McKinsey outlines how sustainability can create value. There are several resource-related value-creation levers for
businesses.
RETURN ON CAPITAL:
- Sustainable operations: reduce operating costs through improved internal resource management
- Sustainable value chains: improve resource management and reduce environmental impact across value
chain to reduce costs and improve product’s value propositions.
- Green sales and marketing: improve revenue through increased share and/or price premiums by marketing
resource-efficiency attributes. For example, companies provide to marketing people numbers and figures to
inform customers that the product has been produced in A class plant, for example for fridges we have A,
A+, A++, but sometimes they also do Green Washing: it is a tactic used by companies to persuade customers
about green operations adopted by the company in order to be more sustainable, but in reality the company
is not sustainable at all.
 lOMoARcPSD

BACKGROUND: THE CHANGING OPERATIONAL MEGATRENDS

GDP by region (graph): we need to be careful, because countries with highest GDP are growing stronger, like
China, USA and OECD Europe.

Asia reclaims its natural position: from

an historical perspective, the East (Asia)
is regaining the powerful position of the
past. It is always important to have an
historical understanding of economical
events, looking at trends, which are
affected by culture, religion and human
perceptions.

Manufacturing value added by country:

in the 80s USA was the biggest one and from then till 2010 China grew a lot, becoming the second one.
In Europe, Italy is the second important one after Germany. The biggest European manufacturing economies
(areas) are: Stuttgart, Bevier and Lombardy. STUTTGART has the biggest automotive industries, like
Mercedes, Bosch and Siemens (suppliers of automotive industry, so they grew thanks to the growth of
automotive industry. The same for Magneti Marelli with Fiat in Italy, which created also a lot of spin-offs).
Also Indonesia grew a lot thanks to lower costs.

GDP growth gravitates to the East and South. Developing countries’ weight is increasing a lot and the middle
class is growing. In fact, middle class consumption is the most important one for manufacturing and the best
markets are those placed in countries where the middle class is growing, like India and China.

GDP per capita: the highest value will be in USA, Japan and Europe and in countries that are growing a lot
Computing power increases rapidly: from 1970 till 2012 the number of transistors in Intel microprocessors
reached 1400 millions.
Connections: data networks connect the continents, like undersea cables and data transmission capacity is
growing fast.

PRODUCTION: WHAT KIND OF?

How the products of the future will look like? How will we manufacture the products of the future?

1. PRODUCTS and PLANTS BECOME GREENER

Despite the economic downturn, customers still care about green products because the forces driving the green
movement haven’t changed:
Energy prices are still volatile
Concerns about product and food safety are still high
 lOMoARcPSD

Governments remain interested in environmental issues

There’s a more compelling business case than ever for reducing costs
The environment is not getting any cleaner on its own
Consumers’ demand for green products is still rising, both for those who systematically look for and often purchase
green products and for those who a higher price for green products is acceptable. Are people willing to pay more
money for green products? Obviously, but sometimes it depends on the previous price.
Also plants are becoming greener because the production process affects the environmental impact of a product.

EX) Which car is the most sustainable? There is a trade-off. Seat is the most polluting while WV is the most
expensive. One is gasoline and the other is diesel. 65% of the environmental impact is due to the usage of the
product, while 35% is due to the production. It is not enough to consider only the usage of the product, but we
always need to take into account the production process. We will choose the type of car according to adding
information related to production process and the whole lifecycle of the product.

2. REBOUND/BACKFIRE EFFECT
Thanks to the technological innovation we can produce more energy efficient products/assets, so less energy is
needed to produce or use the same product. But in the real world we have to face two different effects:

With a better technology and less energy used, it means that the production process is cheaper and also the usage
of the product is cheaper. In this way the product is used more by people, so humans have a wrong perception of
their overall environmental impact, because the usage, and not the production, increases the negative impact.

REBOUND effect: the real benefit expected with the design of the product is not achieved, because reducing the cost
of the production and the cost of the product, the usage of the product is higher and the environmental impact
higher. More energy consumed than expected, but less than before the technological innovation.

BACKFIRE effect: companies kill themselves, because they produce such cheap products that people overuse them
and more energy is consumed than before the technological innovation, creating a PARADOX.

3. PRODUCTS BECOME MASS CUSTOMIZED

EX) M&M and mass customization:

4. EVOLUTION OF THE MANUFACTURING STRATEGY

Product variety-volumes graph: high personalization (variety), high time needed to produce it and low volumes
produced.

Manufacturing production started many years ago, in 1850, with CRAFT PRODUCTION.
In 1908 started the MASS PRODUCTION. We consider the example of FORD mass production: his company’s
strategy in operations was to minimize costs of production focusing on standardization (no customization at
all, all cars were black, so he produced only one product with the same characteristics replicated). The aim
was to make the cheapest possible the product to sell more. The only competitor of FORD were HORSES. The
total cost of ownership of Ford was less than horses, because horses needed costs of buying, cost of
maintenance and costs related to their death.
From 1950 until 1980 there was the spread of MASS CUSTOMIZATION. After World War II, people had more
money so companies started to create variance in products to answer different customers’ needs, increasing
also the volumes produced.
 lOMoARcPSD

Around 2000 GLOBALIZATION changed everything. The curve changes because the same globalized
company has to provide a lot of variety in all countries, but at the same time it has to merge all different
strategies for all different markets, so operations became more and more complex.

5. SERVITIZATION
Products are becoming services:

Product: many years ago manufacturing companies started the production of products
Service to support product: with the increase of competitiveness companies added services to products, in
order to deliver customers experiences.
Service to differentiate product: companies started delivering more complex services to differentiate
products, but that was not enough, because after a while differentiation was completely eliminated.
Service is the product: companies start with the design of the services, then the product is conceived as the
instrument to deliver the service. For example, today cars are not anymore a product bought by customers,
because customers are more focused on the car service rather than the product itself.

EX) Car sharing (B2C): you do not “purchase” and “possess” a car anymore, but you purchase “usage” of a
car.
EX) Selling “heat” instead of heating systems. (B2B) We don’t buy anymore the heater, but we pay a fee to
the company who sell heaters according to the temperature desired in the house. The machine has a remote
control and the maintenance is done by the company who owns it and produce it. If the machine has a
damage, the company is in charge of adjusting it, as a service provided by the company to customers.
EX) Selling flying hours of engines instead of aeronautical engines. In this case (business model) the engine is
produced by RollsRoyce and it is not sold to an airbus as a component, but it is still owned by RollsRoyce
even if it is build on the airplane. The owner of the airbus pays a fee only for the hours in which the engine is
working. RollsRoyce guarantees the full maintenance and operation of the engine.

6. TRENDS IN PRODUCTS
These examples show how the design of the product is changed over time. The product is not anymore sold to
customers who should be in charge of its maintenance, because to design products as cheap as possible, probably
the maintenance is going to be very expensive and customers’ wouldn’t buy them. Today to be more competitive in
the market companies decide to own the product and maintain it, in order to reach a better quality (paying more at
the beginning), but reducing the cost of the maintenance. In this way the TCO (total cost of ownership) for the
company now is lower than it was in the past for customers. In this way companies can minimize the overall cost of
products.

Products are designed in order to be energy aware in order to reduce green house gas emissions. There are several
alternatives like biofuels, electricity, hydrogen, etc. The choice of driving power depends on price, availability and
operating range.
- Biofuels benefit from the use of existing distribution network
- Electricity is affordable, but operating range is short (at least for the moment)
- Hydrogen is the “cleanest”, but its distribution network is largely under planning
The future is a question mark, but all options should be kept open.
 lOMoARcPSD

Products are becoming lighter and stronger to improve raw material and energy efficiency. Companies are producing
using different materials in order to manage in the best way raw materials and they want to optimize energy
consumption, for example lighter the car lower the energy needed to accelerate (like stop and go cars). The same for
airplanes.

Products are becoming smarter, due to mechatronics and IT.

Information technology helps to improve energy efficiency
Digital steering and control systems enable improved ergonomics
Machines can be remotely controlled – humans are freed from working in uncomfortable and dangerous
surroundings
Machines can measure processes and output to optimise performance
Machines can send data on output, lapsed time, and their own condition over mobile internet
Improved conditions for fleet management
Need for repair and maintenance can be anticipated to minimise downtime

HOW TO REACT: THE VISION FOR A COMPETITIVE MANUFACTURING

Global CEO survey: executives of different countries were asked to rank which are the most important drivers that
impact the competitiveness in manufacturing. The ranking says that the talent driven innovation is the most
important driver in competitiveness. The most relevant ones are:
- About energy cost and policies, Italy has higher costs than other European countries.
- Cost of labour and material
- Healthcare system is important because for example Vodafone had to move one of its plants from far away
in England closer to London, because they couldn’t found people to move there (in the middle of nowhere).
Everything is interconnected. If companies want to attract and recruit people, they need to provide an
adequate offering not only to customers but also to employees.

China’s transformation: in this table the same ten drivers are used by the same CEOs in order to define how
countries are attractive (1 very negative, 10 very positive). There is a very high correlation between labour and
countries. Germany, USA and Japan are very good performing, for example it is easier to find top engineers and this
gives a chance to companies to be more competitive than others. On the other side, China and India have the lowest
cost in term of labour and materials. If a company need a lot of human power, it’s better to move to China. This is
how we should use this table, to take decisions according to companies’ strategies.

INNOVATION STRATEGIES
There are new strategies that are becoming more and more relevant today and the strategies of the future are:
 lOMoARcPSD

1. FROM DELOCALIZATION TO “GLOBALIZATION 2.0” (RE-SHORING)

Off-shoring: companies move plants (business operations) far away from their country of origin, in order to find a
dynamic equilibrium. In the past, Italy was the manufacturing country of Europe, but then it lost competitiveness,
because other countries became cheaper and so companies started moving to East Europe and Asia etc. The same
happened in USA, where companies moved to Mexico.

Re-shoring: companies transfer their business operations back to their country of origin. They expect net labour
costs for manufacturing in China and USA to converge by around 2015 so re-shoring is the only solution to minimize
total costs. Moreover, re-shoring defines the end of the practice of awarding business on the basis of price tag.
Today business is awarded according to the minimization of total costs.

Graph: salary of factory-workers fully loaded across time. In 2000 the ratio of the average wage rate of China over
the one of USA is equal to 3% (with the salary of workers in USA companies can pay 300 employees in China) while in
2015 it is equal to 17% (this means that companies can have 5/6 workers with the same salary in the USA).
Productivity, which is man power (output/resources), in China is 3 times lower than in USA, so it is not convenient
anymore to move away from local market.
With the off-shoring strategy companies decided to produce far away from their country of origin, but:
Transportation was a very big issue in term of delay (because the time to market and lead time were
becoming shorter and shorter according to the “amazon effect”) and a very important competitiveness
driver (customers’ expectation increased a lot across time).
Production costs increased
Customization made the off-shoring strategy less relevant

2. URBAN FACTORY
In the past factories were moved outside the cities, creating dedicated areas, because companies needed space and
because they pollute a lot. One of the main drawback was the commuting of people, which changed the way people
lived. In the past areas were very small and distances smaller, so commuting was not a real issue, but now cities are
very large and commuting is a very big issue for workers. For this reason, companies started changing the way
factories are designed and where they are located. Today factories are:
Cleaner and greener
Transparent factory: the factory has a showroom, a way to sell the product itself. For example, there is a
plant of WV that is glass made, completely transparent where people can see workers producing cars, to
show customers how good they are in producing cars.
Local and small
Living and working
Vertical and dense
3. INDUSTRIAL SYMBIOSIS
With industrial symbiosis companies share mutual benefits through:
 lOMoARcPSD

Co-evolution of products-process-production systems

Physical exchanges of materials, energy, water and/or by-products: two companies with nothing to share,
like one company working in market A producing a and the other working in market B producing b can share
energy and electricity building cogeneration plants (which pay back only for two companies and not for only
one), for example, one is using steam and the other electricity.
Optimised interaction of manufacturing with transport and critical infrastructures: 2/3 companies can share
the same infrastructure to reduce transportation costs.

4. FAB LAB
5. FROM PRODUCT/SERVICES SYSTEMS (Product centred approach) TO SERVICES THROUGH PRODUCT
(Solution-oriented approach)
6. FROM USER-CENTRED DESIGN TO USER-WELL BEING DESIGNED
7. INNOVATION AS A BUSINESS MODEL ITSELF

MANUFACTURING TECHNOLOGIES
All these trends must be supported by manufacturing technologies. There are many new manufacturing
technologies becoming more and more important.

Advanced manufacturing processes that will be focused on by the ‘Factories of the Future’ partnership are:
Additive manufacturing (i.e. 3D Printing)
Photonics-based materials processing technologies
Shaping technology
High productivity & ‘self-assembly’ technologies (see video)
Methods for the handling parts, metrology & inspection
Flexible sheet-to-sheet & roll-to-roll
Innovative physical, chemical & physiochemical processes
Replication equipment for flexible, scalable production
Integration of non-convention technologies

ADDITIVE MANUFACTURING (3D PRINTING)

3D printing creates new opportunities:
Capable of producing forms that cannot be made by traditional methods: it allows to produce very complex
products (with different shapes)
Light but strong structures one possible application
Enables distributed production of spare parts: Siemens has 3D-printed spare parts for gas turbines since
January
Additive manufacturing is not likely to replace traditional methods in serial production: instead of melting
metal we add layers over layers of material, using laser technology. GE is starting to utilise additive
manufacturing methods in production. The main applications can be see in industries like:
Aerospace components: less material, less weight and stronger mechanical characteristics than the
previous products
Automotive components
Medical sector: using resins companies produce new dental/bones plants and prothesys
Military sector: equipment, drones, weapons
Others: consumer products, bikes, mini-mes, artistic objects, jewels, toys.
Barbie: traditionally a container contains 13000 boxes, transporting a lot of air. If we transport the
polymer powder, we can deliver 250 000 boxes of Barbie in a container. In this way we change
completely the business model, positioning local 3D machines which produce the final product,
increasing companies’ flexibility and customization of products. The polymer is always the same raw
material as in the past.

The only draw back is the low volume production, in fact 3D works well only with low volume production,
because with low volume production goods this doesn’t work well.

SELF ASSEMBLING TECHNOLOGY

 lOMoARcPSD

Objects/components are designed in order to be assembled without the intervention of workers. Workers program
physical materials to build themselves, so they are not automated machines, but auto-assembling materials. Smartly
design 4D systems. 3D autonomous self-assembly materials generate non random shapes.

VIDEO: https://www.youtube.com/watch?v=0gMCZFHv9v8

THE NEW GENERATION OF ROBOT

The first industrial robot Unimat was installed in 1961 to GM’s Ternstedt factory in New Jersey
• 2 470 000 industrial robots had been sold by the end of 2012
• 1 235 000 – 1 500 000 industrial robots were in commission
• Robot density (number of industrial robots per 10 000 employees) is highest in Korea, Japan and Germany
• Highest density in automotive and electrical/electronics industries
• Use is growing in small and medium- sized companies
• Robots become more affordable, simpler to use and capable of collaborating with humans
Collaborative robots are robots which are co-working/collaborating with humans. They have sensors, cameras and
proximity sensors to detect where objects are.

ICT4MANUFACTURING

INDUSTRIAL REVOLUTIONS:
- First phase: the first industrial
revolution happened when
water and steam power were
used as a source of energy
- Secondo phase: it started
when companies used
electricity as a source of energy
together with new
manufacturing concepts
- Third phase: computers in
manufacturing, like PLC
- Fourth phase: smart factories,
industries 4.0

In the last 5/10 years, different IT are

becoming mature to provide systems in manufacturing:

ICT MEGATRENDS
- Collaboration: companies design
products 24h a day, sharing
projects and files trough the
world between countries with
different time configurations to
exploit better time and reduce
time to market.
- Mobility: companies rely on
mobile devices in order to be able
to be connected 24h a day and
wherever employees are.
- Connectivity: IT, internet of
things, with sensor that detect
information
- Intelligence: means
understanding and using data
detected by sensors in order to
understand/predict trends and take decisions.
 lOMoARcPSD

INTRENET OF THINGS (IoT)

Internet connects also object and not only computers.
We do not make reference to mobile phones, but cars and any
object that can detect data and can transfer it anywhere in the
world. In order to do this, we need to create the digital
representation of an object (Avatar). Starting from a physical
object, like a car, we need to detach the real object from its
physical representation, like in the case of Google Car, which
guidance is based on the digital representation. With the Avatar
we create a digital world, separated from the real one.

IoT is not enough thinking in a broad way, because

the business is made by people who produce and
make decisions. We should connect also:
People: Internet of Brains
Processes: Internet of Business Process
Services: Internet of services

virtualizing them to get a virtual representation of

the real world and manage everything in a better
way, thanks to the new INTERCONNECTED
MANUFACTURING.

MOS – MANUFACTURING OPERATING SYSTEM

MOS is an evolving CPS based Manufacturing
Management System to support a rapidly changing
business environment:
 lOMoARcPSD

BIG DATA
Sensors are cheap and can collect data easily, producing a lot of data and creating the big data issue. The problem is
not related to storing, which is easy to be solved, but with the decision making process, because it is difficult to
elaborate and use data to take decisions.

CYBER-PHYSICAL SYSTEM (CPS)

All real parts are duplicated in a virtual “cyber” world, with cloud systems, taking decisions in the real world and then
moving to the virtual one.
Cyber is the virtual/digital part, physical is the real part: they are merged together: each real object has a
digital shadow, which reflects the characteristics of the real object.
Communication everywhere and every time: future infrastructure will support the access to information
everywhere and every time without any specific installation/parameterization needs
Production and products will be intelligent: production resources will be automatic and will connect to each
other (M2M). Products know their own production system.
CPS creates new business potential in industrial services: analysis of data provides better machines and equipment,
improved processes and more efficient production.

COGNITIVE AUTOMATION
Important trend related to the importance of people in manufacturing. Cognitive automation means that instead of
having robots we have workers who are enable to perform better their activities due to the fact that we provide
them technologies to perform better tasks.
EX) Volvo plant. The worker reads real time information to better perform the task.

MODELLING, SIMULATION AND FORECASTING

All previously mentioned enablers are reinforced by the capability of simulating manufacturing processes or
forecasting the behaviour of manufacturing systems & processes. Advances in ICT are moreover enabling the further
development of simulation & forecasting tools.
Research in this area will focus on:
• Modelling & simulation for the (co-)design & management of integrated product- process-production systems
• Virtual modelling spanning all levels of the factory life-cycle

THE PEOPLE
Factories of the future are expected to create a large amount of employment opportunities for citizens. Factory
workers are key to competitiveness, but challenges such as changing demographics & news skills must be addressed
(we should design manufacturing processes and businesses according to how people will interact with technology).

Policies should address the following items:

New approaches to accommodate different demographics (aging of the population will bring old workers in
factories with different skills and abilities applied to technology)
New technical, educational & organisational ways to increase attractiveness of factory work
New approaches to development of skills & competences
New ways to organise factories: Human-centred work environments
Ways to integrate future factory work into social patterns
 lOMoARcPSD

14/3/16
OPERATIONS STRATEGY
All companies, to survive in time, have to build and keep a specific advantage that differentiate them from
competitors. Strategy of operations should be always in line with the strategy of the company, because operations
create vale for the company. We have already seen how companies compete, focusing on something that
differentiate them against competitors and managing this differentiating advantage is the essence of strategy.

STRATEGIC LEVELS
We need to define what level of the company’s organization we are considering:
Corporate level: resources allocation between different markets and products. In which business we want to
be? Each country in which a company operates has its own CEO and all the CEO around the world have the
same responsibilities.
Business unit level: what are the markets’ needs and how to satisfy them. How do we compete? We need to
adjust the offering according to market and consumers’ needs. Different strategies across the world in
different markets should be adopted, because BUs are like independent companies within a company.
Functional level: support the company in satisfying market needs. How we deliver the product/service? How
we support the company in the implementation of the strategy? Operations, Marketing, Accounting, R&D
and many other activities.

TRADITIONAL APPOROACH
TOP DOWN APPROACH: BUs implement locally the strategy received by the top and
then they execute it. Decisions are taken at corporate level.

Drawback:
Decisions are taken at corporate level and operations are supposed to do whatever is
decided at centre level (price, volume, marketing activities, etc…).

Wrong believes about operations are:

Operations are mostly technical: operative aspects are details
For the operation system there’s a one best way of improvement: managerial choices are not necessary

This approach doesn’t work in highly competitive environments, not stable and with lot of variance where the speed
of variation and change is very high. Companies fail in implementing this type of approach.

FORCES OF CHANGE
Relevant factors outside the company:
- Offer > Demand: for example, in the automotive industry the current production capacity of all the plants
which produce cars all over the world is 2.5 bigger than the demand, so plants are designed in order to
create more value than the requested ones. Companies invest a lot because they are pretty sure that the
demand will increase, as their market share. The problem is that every company is doing the same and they
steal market share between each other, so the main problem is the increase of costs.
- Customization
- Globalization
- Speed of technological development: today technology changes are faster than market changes and
customers’ need. In fact, today we have more a technology pull approach rather than a market pull
approach, because we should not wait for the market and do what customers wants, but we should do the
contrary.

Factors influencing company’s resources:

- Economical (wealth increase)
- Cultural (education level)
- Social (authority acknowledgment)
- Technological innovation
- ICT
These are all constraints for companies when they develop a strategy, they influence the way decisions are taken
 lOMoARcPSD

EX) If in USA companies rely on BOSS decisions which must be accepted and implemented by workers, in JAPAN
BOSS decisions rely more on workers’ opinions, supported by very effective communication.
If strategies are imposed by the corporate level only, they will never work efficiently and effectively, but Japanese
showed that managers should train workers and communicate with them to be successful.

EX) Fiat has implemented over time a new approach for the just in time production, which has already been
implemented by other companies in the automotive industry.

INNOVATIVE APPROACH
Functional strategies cannot be independent one from another, nor they
can be the sequential outcome of the business strategy; rather VP of
main functions have to communicate and interact to define the strategy
at BU and corporate level.
R&D, Operations, Marketing, etc must not be independent, but they
should communicate (this is a drawback of the traditional approach).
Companies should create strategies going horizontally leveraging on
interactions.
This approach, also called INTEGRATION & BI-DIRECTION APPROACH is
based on:
- Interaction
- Mutual approval

Business model: how the company plans the strategy Business strategy
Operating model: how the company realizes the strategy Functional strategies

Any time we plan a strategy and we decide what should be done in the plan, there is always a GAP between what we
have decided and what we realise at the end (what at the end is implemented).
The difference can come from unplanned events: during the years something new could came, something
unplanned at the beginning, so companies need to be flexible. Company’s flexibility can be seen from two
perspectives:
on one side, companies need to be flexible in the decision making process, like deciding not to do something
already planned (not realized) or to do something new not planned before (emerging)
on the other side they should detect what could be possible unplanned events in order to be able to react
quickly and efficiently (especially in fast changing environments). For these reasons sometimes companies
need to change the plan defined at the beginning.
-
WHY STRATEGY IS IMPORTANT?
To create a sustainable advantage, it’s necessary a significant amount of time. This advantage is rarely the outcome
of just one excellent critical choice, but it’s more the outcome of many little good choices.
 lOMoARcPSD

EX) If you want to go to Venice, you need to consider many small choices to take (if you go the wrong day, the visit is
useless, for example, due to weather, traffic).

To take many little good choices a clear line is needed: DELIBERATE STRATEGY
In highly turbulent environment it’s not possible to plan everything in advance, but it’s possible to be ready:
EMERGENCY STRATEGY
There are two ways to deal with emergency strategies:
- Deliver a resilient strategy: if something happens, the main strategy/direction is not affected by the event
- Being reactive: being able to look in advance to the future, trying to anticipate the unforeseen events, not to
avoid them

Positive events are threats:

- When we are not able to exploit them
- When they reinforce competitors

EX) If you want to go to Venice, you need to consider many small choices to take (if you go the wrong day, the visit is
useless, for example, due to weather, traffic).

STRATEGY
We should keep sensors
according to top down and
bottom up approaches: delivery
and emergency strategies merge
together.

THE VOICE OF THE CUSTOMER

What is the dimension in which customers are affected by operations?
Performances affected by operations, which must be considered in defining a strategy, are:
- Time: how fast a company is
Delivery speed: how short/long is our lead time (LT)
Delivery reliability: in term of day, time window, etc (variance or way we respect the delivery time)
Price (cost): customers want to pay products the less possible
Quality
Quality of design (specifications): for example, a low cost car has low quality of design. It represents
how companies position the product according to customers’ segments: low quality in the lower
segments, high quality for premium products in the higher segments
Quality of conformance: it is the difference between the designed performance and the real
performance, so it monitors if a company is compliant with what customers require.
Flexibility
Product
Customization
Variety
 lOMoARcPSD

Plan
Service

This is the check list to design a strategy, because customers look at these performances before buying a product.

TIME
Time to formulate the offer
Time to confirm the order
Time to deliver (delivery speed)
Delivery reliability (timeliness)
The main difference stands in:
- Make to stock (MTS): the product is in the stock and when companies receive orders, they use inventories
to deliver products to customers.
- Make to order/Purchase to order/Assemble to order/Etc: the product is not in the stock and when
companies receive orders they need time to deliver products. There is a high risk to lose customers due to
the high OCT required between the order receiving and the order delivering.

PRICE
Customers are becoming more and more smart, so it is even more difficult to cheat them. Price is composed of:
- Cost of purchase
- Cost of usage
- Cost of maintenance
- Cost of update/upgrade/expansion
- Cost of disposal

SERVICE
It depends what do you mean with service.
- Delivery (MTS: make to stock): goods availability at the WH
- Other areas:
Training: when you need someone to train you to use a certain complex product
Technological improvement
After sale support

OPREATIONS’ OBJECTIVES
The performances previously mentioned are operations objectives very difficult to be achieved. There is not a best
system on all the dimensions. The main problems are related to:
1. Finding the right compromise among all these features
2. Markets and customers are different
It’s necessary to align Operations and Market.

MARKET-OPERATIONS ALIGNMENT
Companies need to decide where they want to be in the market, so their positioning, which depends on customers’
needs and where competitors are. The positioning of a company depends on what are the desired performances, so
how to design them?
First of all, we should understand the market, then we need to define the possible levels of each performance. Each
company has to define the desired level for each performance and it should understand what are the achievable
performances of operations. We need to design the levers, so 3 areas used to achieve the achievable performances.
This is a way to design and share the strategy with stakeholders.
 lOMoARcPSD

DESIRED PERFORMANCES ACHIEVABLE PERFORMANCES

RECONCILIATION
We should find the right compromise

STRATEGIC LEVERS
Structural design
Infrastructural design
Delivery management

STRUCTURAL DESIGN CHOICES

This is only a checking list, because we need to go through to understand if we have considered everything in order
to implement the best feasible strategy.
Overall production capacity, its division and localization (where and how much):
These drivers must be considered to design the capacity:
- Entry barriers: with the protection of entry barriers, we can invest more in the long term because
we are protected
- Face demand variability: with a high variability we should not design a very high capacity. We
should be flexible.
- Economies of scale: increasing the production capacity, the cost per unit decreases, because the
fixed costs are spread on higher volumes. It is convenient to have larger plants with higher
production capacity than small plants with small production capacity
- Volume flexibility: how good we are and how cheap it is to increase/decrease the volume capacity
- Minimum unit of increase/decrease: normally flexible plants decide a fixed amount 10/100 usually
to increase/decrease the volume capacity step by step
- Work cost
- Competences availability, services, …
- Response time
- Closeness to market

Strategic make or buy (outsourcing)

Technologic process and equipment: shall we use people or robot to assemble products? It depends on the
cost, flexibility, quality.
Mechanization/automation grade: we need to choose if using the same mechanization/automation grade in
all plants or only in some of them.
 lOMoARcPSD

Type of plant design of the system: graph

Supply chain configuration

INFRASTRUCTURAL DESIGN CHOICHES

Competences needed and their management
Responsibility allocation
Team vs individuals
Managing by objectives or procedures
Functions integration (design and manufacturing, marketing and planning, …)
Incentive systems (driving people behaviour, so it is a crucial decision because wrong decisions lead to wrong
behaviours)
Information flows between:
- Different functions
- Different hierarchical levels

DELIVERY MANAGEMENT
How we manage deliveries. We need to design:
Operations planning and control system
Choice of how to meet demand (eg. MTS, ATO, MTO)
Choice of how to realize the product
Supply chain coordination system
Maintenance managing and realization systems
 lOMoARcPSD

16/3/16
STRATEGIC CAPACITY MANAGEMENT
If a company does not have enough capacity cannot produce: capacity is important!

Characteristics of capacity strategies

- Timing of change (when)
- Magnitude of change (how much)
- Attention to transient (be careful)

Key issues
Critical parts, decisions, attentions that we need to pay when when talk about capacity increase (not complete list):
1. Lead time to complete the change
2. Flexibility to change: are we able to change our mind? Is the current technology that we have able to
increase the capacity in the future? How easy is to change the capacity that we have? It is easier in the
manufacturing industry (like oil refinery), with less risk, because it is easier to follow the demand, but in the
process industry it is a huge issue.
3. Economies of scale: they have an impact on the size of the increase. EX) If we have one oil refinery and we
want two of them, we can have problems in separating equally the demand.
4. Forecast of demand trend: nobody forecasted the financial crisis, which impacted a lot the production of
many industries
5. Forecast’s uncertainty (fulfilment/demand): the effectiveness of our forecast
6. Competitors’ behaviour: if we increase the capacity to increase the market share, we need to take into
account the reaction of competitors, so if they allow us to increase our market share or not.
Companies should expect some reactions, so they should forecast not only events but also reactions of other
companies in the scenario. Company’s expectations are affected by behaviours of other companies, so we
need to take care of the uncertainty of the system and predict possible scenarios to be ready to react
without failing.
We especially need to take care about this issue when we do strategic capacity planning, which is a long
term decision. Today, due to globalization, the speed of changes in the market is increasing and long term
decisions are becoming shorter: the risk of decisions becoming rigid in the long term is higher, so companies
try to be as much as possible flexible in the long term. The only way to reduce risk is to share risk and
reduce the impact.

7. Requested service level/customers’ behaviour

Capacity is not an operational issue only, but it is also a

market issue. Some of the success of the increase of the
capacity depends on marketing, so marketing and
operations are becoming closer. Better is the quality of the
information that we take from the market, lower is the risk
of the decision that we are going to take. Marketing is not
just making right products, but also understanding and
anticipating the future, providing the company right
information to take right decisions and reduce risks.

CONSIDER ECONOMIES OF SCALE

Volume and unit cost curves: the unit cost goes
down if we increase the capacity of the plant, not
if we produce more!
NB) We are comparing different plants with
different production capacities. (spring question)
For example, what would be the unit cost if I use a
plant with 1000 units’ capacity? With 2000 unit?

Is the economy of scale always going down?

 lOMoARcPSD

It can be too big/large, that it can become a DISECONOMY OF SCALE, where the cost increases, due to high
complexity and high traffic between a large number of WH.

CONSIDER RISK
All values are estimates: all the values that we are considering for our estimation may not be present in the
future.
To get more precise data there is a cost. Sometimes it is even impossible: to have more detailed information
means increasing costs and companies not always decide to get them or they cannot get them.
Sensitive analysis helps in identifying key elements: sensitive analysis is used to test all different possible
scenarios in order to find what are the factors that impact our decisions (if changing a factor there is no
impact, this factor doesn’t impact our decision. If there is an impact, we need to evaluate that factor
carefully)
A certain degree of uncertainty is unavoidable, it is therefore necessary to be prepared for it
Data are not only estimate, can be influenced: we do not have to take everything for granted, because the
system is not always a closed loop system. We should keep our data updated. (e.g. sales forecast depends
also on marketing actions)
Passions is a key element for success

CONSIDER TRANSIENT
Transition phase: to change something and adopt/train people it takes time, so companies need to anticipate
possible scenarios. Transition is an important aspect which companies need to take into account.
It is impossible to get all benefits from the outset
Performances during the transient period can be permanently disrupted: if you stop your plant or if you
reduce your production rate or if you change the layout, you can lose customers permanently
It may require a long time to make the conditions to change: prepare the change can take more time than
the change itself. For example, to move the equipment from Linate to Malpensa took only 5 hours while the
planning phase took more time.
In case there are not the right conditions it may be better to postpone change: postpone the change to
prepare it in the right way
Anticipate (avoid) problems: a work breakdown structure and a project planning allows to anticipate and
avoid problems

TIMING AND MAGNITUDE OF CHANGE (when and how much)

LEADING AND LAGGING STRATEGIES
The curve represents the demand in term of quantity/time
(units/day, unit/week) which increases across time.

LAGGING (Purple line): capacity lags (comes after) demand

We are always late We increase the capacity only when the increase
matches the demand. (Capacity remains fixed unless we do not
change it). We usually produce less than the demand.
Demand > capacity

LEADING (Red line): capacity leads (comes before) demand

When the demand reaches the capacity we are ready to do another
step, to increase capacity.
- Drawback: we are always unsaturated, because we have more capacity than the demand, so we do not work
at the best of our capacity (that allows us to have a faster response to an increase of the demand). It costs
more because to have high capacity we need to pay more, so we have higher investments and because of
the unsaturated capacity.
- Advantage: we are able to respond to an increase of the demand.
Demand < capacity

These are the two extreme situations.

 lOMoARcPSD

LEADING:
PRO CON
Always spare capacity for opportunities Higher production unit cost
Faster response time Outbound cash flow
Better delivery reliability Higher impact from overestimating demand
Lower impact of uncertainty and unforeseen events
Lower impact from underestimating demand

LAGGING:
PRO CON
High plant utilisation Longer response time
Low production cost Lower delivery reliability
Lower impact from overestimating demand Higher impact from underestimating demand

Other approaches are:

SMOOTHING

Smoothing with inventory means using the excess capacity of one

period to produce inventory which can be used to supply the under
capacity period (when the capacity is lower than the demand). It is a
kind of intermediate solution used to anticipate the demand
increasing the production capacity.

FILLING PRODUCTS

In some cases, it’s possible to use the plant for other productions.
When companies have extra capacity they can decide to produce
products, which do not match their core competences, to avoid the
waste of capacity. It is only a possible solution, because it is not
always possible to do this.
Using filling products is like modifying the demand. It means that:
It’s possible to influence demand
Not all the products are the same
It’s useful to identify which product sacrifice in case capacity is
not enough

EX) CREMERIA: they produce seasonal products, which are produced only seasonally like Panettone and Colomba (4
months for Panettone and 3 months for Colomba). During the rest of the year, to saturate their production, they
produce snacks and other sweets, like frozen croissants, moving to different distribution channels and designing
them (like using dedicated trucks which can reach bars). These products are called filling products.
We are considering capacity not only about production, but also about transportation, in term of capacity of trucks
which deliver products to different points of sale (bars). The distribution capacity is affected by bars’ requirements,
so many times it was unsaturated, making costs increase.
In the meanwhile, the main ice-creams distributors, which are Unilever and Nestlé, became interested in the new
product, frozen croissants. They forced bars to buy frozen croissants in order to be able to buy ice-creams, so
Cremeria lost a lot of market share due to its higher costs of transportation. It decided to buy an ice-cream producer
company (Samson) to compete on the market (in term of prices) and to have the same portfolio of product as
Unilever and Nestlé.

With croissants they saturated their production capacity, but not the distribution channel capacity, so they
needed to buy Samson which saturated the transportation capacity.
 lOMoARcPSD

17/3/16
OUTSOURCING

In some cases, it’s possible to use external supplier to balance

capacity and demand. It is the opposite of the filling product.
The under capacity is covered by outsourcing to
suppliers/third parties the missing production.
It is convenient in term of price and cost when the full
variable cost of suppliers is lower than the variable cost to
produce products internally.
EX) If the price is 10€/unit, to produce it cost 6€/unit and
outsourcing 5€/unit: it is more convenient to outsource.
We do not need to take into account the full fixed cost.

Comments:

The lifecycle effect on capacity strategies

Is the lifecycle of the product impacting the choices and strategies of the company? YES.

1. At the beginning of the life of the product we are forecasting

the future growth of the product. But if it doesn’t grow, the
curve will go up and immediately fall down without reaching
the maturity, going out of the market.
Going for a lagging capacity could be risky, because you are
pushing the product creating expectations without selling it.
It’s better, at the begging of the lifecycle of the product, to have
more capacity, using a leading strategy, to be able to deliver the
product to the market.
Sometimes, it’s better or it could be used to have no product
available in the market as a market strategy.

EX) When Porsche launched the Cayenne model, they advertised it, but customers had to wait at least 6 months to
have it from dealers. This was a decided strategy to create and raise expectations in the market, even if the company
had enough capacity to produce it: the result was that everyone wanted it.

2. When the product is growing we should reduce the impact of not delivering the product, so it’s better to
have inventories, using the smoothing strategy.
3. When we reach the maturity, we could switch to a lagging strategy, because the forecast is more reliable
and the situation is more stable.

NB) This is a general approach, not a rule, because companies need to adjust/adopt a certain strategy according to
the information, the cost, the market expectation, customers and competitors’ decisions.

Capacity planning
Capacity planning should not always be planned with a fixed/constant
interval and the size of the capacity increment should be decided
according to the forecast.
There is a slow growth of the demand at the begging, where small
increments take more time to be implemented, then the forecasted
demand increases with a higher speed, so the capacity increments are
higher and with less time needed to be implemented.
 lOMoARcPSD

Forecast issue – Demand and lead-time uncertainties

Looking into the future is very risky and therefore it could happen to have many types of uncertainty.
Uncertainty means that, when we do the forecast, we should consider that it could be wrong (for example, at the
end the demand could be higher or lower). For a better planning of the situation we should create a range of the
forecast. The range is bigger when the time goes by.

Graph
Uncertainty can be on 2 dimensions:
Y axis: demand uncertainty (> or < than the forecasted
demand)
X axis: capacity increment uncertainty (uncertainty in the
delivery of the capacity increment: it can be late or faster
than expected)

Planned lead time (Y-X) is the lead time before the next
increment. X is the starting point, while Y is the next step
where we are after the increment of the capacity. We can
have a late increment, when we are late in increasing the
capacity (latest finish time) or we could have an early
increment, when we ask our suppliers to be faster so when
we are early in increasing the capacity (earliest finish time).

The possible scenarios that we are going to face are:

A) Capacity on stream early, but demand on lower boundary of forecast

Actual demand goes to the lower forecast range, which means that it is late
and lower possible. If the capacity increment is as earlier as possible,
companies face the worst situation in term of capacity: the capacity
increment happens early and the demand is late, so companies have to deal
with a very high extra capacity.

B) Capacity on stream late and demand on upper boundary of forecast

Actual demand goes to the upper forecast range, which means that it is
early and highest possible. If the capacity increment is as late as possible,
companies face problems in term of capacity: the capacity increment
happens later than the actual demand, which is early, so companies have to
deal with a lack of production capacity.

Scenario analysis
This analysis is done to understand what are going to be the costs if these scenarios happen, to face the possible
scenarios and be ready to react properly.
To anticipate possible constraints/opportunity: opportunities like over capacity which can be used to test
new products. Companies can also transform constraints in opportunities.
To prepare countermeasure: if we do not fulfil the demand, we need to explore potential suppliers to
activate them if it is needed (outsourcing).
To limit the impact of unforeseen events
 lOMoARcPSD

To prepare a faster, more efficient and more effective response

Being ready has a cost, but what would be the effect if you are not ready? No risk no gain.

Now we deeply analyse the Y axis, the SIZE OF THE INCREMENT

We should decide the size of the increment, because it influences our decisions. The two options/strategies are:
Small capacity increment (400 units): it is easier to follow the demand, so less risky.
Large capacity increment (800 units): it is more difficult to follow the demand, so higher risk. It could not
payback the investment, due to the very high over capacity.
The best option depends on the cost of the increment.

Elements to consider to define the size of the increase:

- Economies of scale: big capacity increments provide costs reduction, while small increments provide small
benefits in term of costs It is better to do big increments
- Demand uncertainty (stage during the life cycle): At the beginning of the lifecycle the uncertainty is high,
while in the maturity phase the uncertainty is lower It is better to do small increments at the beginning
and large increments at the maturity phase
- Financing availability: without money companies cannot do changes. Big capacity increments need a lot of
money, while small increment only a small amount of money
- Over/underutilisation costs and Outsourcing possibility: both are linked to the possibility of using filling
product, smoothing and outsourcing approaches
 lOMoARcPSD

17/3/16
MODELS FOR OPERATIONS STARTEGY
Performances
They refer to characteristic of the products:
ORDER QUALIFIER (Q): it’s a feature of the product that
customers and the market expect the product to have. EX)
Shampoo for children: it should not affect children’s eyes.
It is a basic expectation.
Graph: if your product has a very good performance in
term of the qualifier characteristic, it doesn’t increase
sales, because it is not perceived as a differentiation
feature.
Q is necessary to be considered by customers

ORDER LOSER (QQ): it is a feature which, if not owned by

the product, can have a very negative impact.
EX) If we want to buy a Gucci bag we expect it to be very
expensive. If someone is selling it with a very cheap price,
nobody will buy it. Increasing too much the performance
by reducing the cost could be counterproductive. This is an
order loser.

ORDER WINNER (OW): an order winner characteristic

differentiates the product from competitors’ products. The
company should advertise it. The more OW characteristics
you have the higher is the number of orders you will get.
Graph: there are 2 OW: 1 star and 3 star. 3 star has a
higher impact on the number of orders.

Characteristics:

1. Non linearity: we use as example OW characteristics:

GREEN: diverging behaviour (at the end OW has a
parabolic trend)
RED: saturation behaviour (when you put too much
effort in increasing the performance).

QQ and OW depend on the market perception.

- If the price is too high (higher than a reasonable
value): you don’t sell it
- If the price is too small: QQ, counterproductive

How to define them:

- Set of representative customers
- Set of significant orders (behaviour): when you start to quantify, there are always surprises
- Interview: to get information
- Ranking: rank orders Evolution in time tool
- Distribute 100 points

MARKET SEGMENTATION
Which market segments? Market should be segmented and divide Q, QQ and OW according to market segments.
Search and underline differences
Group of homogeneous needs (segments)
A customer can be part of different groups (promotional Vs standard packing)
 lOMoARcPSD

Operations Point of View

EVOLUTION IN TIME
It is a tool with all typical performances (we can also add others) evaluated:
OW: using numbers (from 1 to 100)
Q
QQ

Comments to the evolution in time: an order qualifier is not forever, it can

evolve.
It could be that an OW becomes a Q in time: at the beginning of e-
shop the speed of delivering products was an order winner, while today it is not anymore a winner, just a
qualifier.
It can also happen that a Q becomes an OW: market expectations have changed. Expectations change with
the market evolution.

MAP OF PERFORMANCES
This tool is used to visualize and understand how a company is performing against its competitors, also during time.

INTERVENTIONS PRIORITY
Importance
Distance
Competitors
Difficulties

PROJECT SELECTION MATRIX

Let’s assume that we want to go from the orange to the red line in time. We should analyse different projects
(actions in price, packaging, speed, …) to implement our strategy. All these actions require efforts (time, people,
competences, money) and the impact of each project is not necessary the same (for example they depend on the
money availability). The previous tools described where simple and powerful tools, but not used in most of the
times, while this one is very used.
 lOMoARcPSD

LIFECYCLE S CURVE
We should always keep in mind in what stage we are.

OPS STRATEGY EVOLUTION

How is the role of operations strategy changing in the company with time?
The role of operations can be seen through 4 different stages:
1. Correct the worst problems
2. Adopt best practice
3. Links strategy with operations
4. Give and operations advantage

At the beginning the role of

operations is crucial while the
role of the market is less
important due to the fact that
whatever the company is going
to produce there will always be
innovators and early adopters.
Then the role of the market
becomes more and more
important because companies
need to increase their sales
and revenues to become
profitable so operations cover
only a supporting role.
 lOMoARcPSD

A TYPOLOGY OF OPERATIONS

How a company is positioned

according to drivers of
performances of operations:

Volume of the market

Variety
Demand
Visibility (how our
performances are visible to
customers’ perception)

EX) 4 v analysis for some retail banking processes

BREAK EVEN CURVE

The curve is linked with the growth of the demand.
Revenues is a linear curve.
If the price is fixed: more output means more
revenues, but more output needs an increment in
capacity: costs do not have a linear trend.
Variable costs increase linearly with output.
They grow less than revenues (on the
contrary we will never make money)
Fixed costs increase according to the increase
of capacity

Fixed cost break: capacity increment where we

introduce more fixed cost.
 lOMoARcPSD

There is an area of volume where we suffer, so it is very dangerous and we should not stay there. Even though we
have the break, sometimes the slope is not so high and we could end up in this risk area.

Break even point: when the company starts making money.

R=P*V

SUMMARY
1. Understanding what is for your product Q, QQ, OW
2. What are those performances used to position yourself against competitors and customers?
3. To do projects which impact the performances, be careful
4. Be careful, according to the stage of the product lifecycle, the weight of Q, QQ, OW changes (according to
expectations)
5. Be careful, there are different segments which need to be addressed in the right way
6. Within the same company you could have different mix of operation levers: what is the proper mix to be in
the market?
6/4/16
TRADE OFF
TRADE OFF (no change curve)
Strategy can be based on many performances and we saw that, for example, Amazon influenced many of these
performances in its sector. Companies should try to improve as many as possible performances.
Trade off is a theoretical concept which can be represented
by a curve.

The trade off curve is used to compare any performance with

cost performance. The better we are, moving A right, the
higher the costs.
A - right: worse variety performance (low variety) but better
cost performance (low costs)
A - left: better variety performance (high variety) but higher
cost performance (high costs)

How to use this curve? How to use this concept to support

the designing of our strategy?
The first reasoning starts asking ourselves were we are today
and were we are going to be in the future.

TARGETING (to compare ourselves with the market)

Targeting is a way to use the trade off no change curve.
If we move along the curve (A A’) we are changing our
position (from one segment to another one in the same
market). In both cases it could be that Market expectations
are not along the curve we are able to go through. The curve
represents the options we have according to the structure of
our operations management. The no change curve represents
ourselves: all the possible options we have today.

We can have a GAP if we benchmark ourselves with the

Market. If we have a gap, we have a problem.

Comparing Market with A’: the Market has a better cost performance and variety performance, so competitors are
offering in the market better performances. The two stars of the Market are better than any position of our
company along the curve, so in both segments we are worse performing than the Market.
If the star of the Market would be down: the cost performance would be better, but the variety performance worse.

This is a tool used to understand where we are and were we want to be.
 lOMoARcPSD

THE EFFICIENT FRONTIER (to compare ourselves with competitors)

Another way to use the no change curve is to benchmark ourselves with competitors. The dots represent different
companies. If we are on the curve, we are efficient: we are in the best frontier we could have.
Be careful: the shape of the curve is not necessarily symmetric as here. The curve is not part of a circle or something
like that. It depends on KPIs used to measure the performances and the performances considered.
The curve called “efficient frontier” represents the best mix of the performances present in the market.
If we do some kind of improvements, like operations system improvement or technological innovation, we are
moving our curve somehow, like up on the right (B B’). Later we can also move along the new curve, which
becomes the new efficient frontier. This is a way to be more competitive than competitors.
How to move to the new curve? Is it always thanks to a new strategy or more different strategies? More strategies.

OVERCOMING TRADE OFFS

In this picture, the no change curve is the curve where there is A.
Assume that we are in A and we want to do some improvements.
What should we improve? We start changing only one
performance:
- We keep the variety the same and we try to reduce costs
(EX: improving lean manufacturing, reducing set up time,
less stocks, less space, better layout, less people. All these
have impact on costs): A D
- We increase flexibility and we keep costs the same: A C

EX) If we reduce set-up time, what happens to a machine which

produces 10 items? The volume increases and we can spread variable costs on a higher volume of products.
1. Is the set-up cost a variable costs or not? No
2. How the set-up cost is made? If it is made of man power (workers to prepare the machine), it is VC or FC?
If the worker who prepares the machine is the same who works with the machine and if he does not do any
other additional activity: it is a FC. The set-up cost doesn’t cost us more, because we are already paying the
worker. But this has an impact on the saturation of the machine:
If the machine is unsaturated and if we increase the set-up, there are no additional set-up costs (in
term of opportunity cost)
if the machine is saturated and we decide to increase set-up, we increase set-up costs. This costs us
the difference given by the unitary cost to buy the product outside from the supplier and the unitary
cost to produce it.

SET UP COSTS DENPEND ON THE SITUATION WE HAVE

If we reduce the set up time:

we can create more capacity, increasing the variety performance: thanks to higher capacity, it is easier to
change from product A, B, C and to reduce the batch size. It is easier to fulfil customers’ demand for
different products at the same time.
we can reduce costs, increasing cost performance.
It depends on our choice.
 lOMoARcPSD

SECTOR CURVE TRADE OFF (to compare ourselves with the sector
and other companies of the sector)
A and B are two companies of the same sector and they have
different no change curves (we can see only a part of them).
To design the sector trade off curve (no change curve of the sector)
we need to interpolate the performances of all the companies of the
sector.
Assume we are B. With this mix, B is on the sector curve, so it has a
chance to be successful in the market. If B would not be on the sector
curve, it wouldn’t have no solution to be successful in the market.
The only solution is to move back on the sector curve.

CHIEDI APPUNTI
But remember, this is not necessarily the best option for B in terms of MARGIN.
This is linear? The increment cost and service level are linear? NO. The better we want to be, the more we have to
pay. If we want to provide less SL, this costs us less money.
How much money we do? If we provide low SL, we get few money, nobody wants to buy the product. Better we are,
better chances we have to get more money. Getting better we get more money.
How the market is increasing? With a second derivative. If we still increment our price? At the end the nr of people
WTP high prices become less. The second curve.
- If we are too low in the SL, we loose money
- There is a minimum level we need to have to start gaining profit
- If we are too high in SL we cannot fin anyone who would buy the product

There is an optimum position to be: where the margin is the highest possible. This optimal position is B, at the
intersection. Moving along the B curve, means losing margins.

POTENTIAL CURVE
How to move frontiers? Moving frontiers is not
always happening upwards and with the right size.
Companies can have a higher/stronger increment in
one dimension (one driver of competitiveness) and
not on the other one.

COMPANY F
The meaning behind this movement is:
for low variety (standardized products) they
are very good in reducing costs. If they
more to high variety (larger portfolio of
products) they cannot improve their cost
performance as before. F can increase the
potential curve staying in the low variety
performance.
COMPANY G: they are good in variety, but
not in costs. The contrary.

Don’t fall into the trap: the shape of the curve

doesn’t always remain the same if we move. It is
not negative, but it can be a positive issue;
sometimes we need to adjust the company position to the sector curve to be more efficient.
 lOMoARcPSD

DRAWBACKS of trade off analysis:

1. We are considering a performance at a time (performance vs cost performance). This is a limit.
2. The problem of designing the curve and simulate (for targeting)
3. Understand where competitors are and build their curve (for efficient frontier). Normally companies hire
people from competitors to know them.
 lOMoARcPSD

21/3/16
MANAGING WAITING LINES - Queueing theory

QUEUEING SYSTEMS – Introduction

General questions:
1. What is queuing system and why does it form?
2. Why it is important to study and to manage queues as best as possible?
3. What are the queuing system applications?

QUEUEING SYSTEMS
A queuing system is formed by one or more “customers” waiting for a service by one or more “servers”.
Customers’ examples
People waiting at the cashier’s desk for a bank operation (financial service)
Pieces waiting to be processed by a lathe in a job-shop
Container waiting to be loaded on chassis
Corresponding servers
Cashiers
Job-shop’s lathes
Overhead traveling cranes’ array for the loading

Both pieces and people are treated in the same way and also many different situations are treated in the same way.

Example 1:
A bank historical data show that on Monday there are on average 12 customers/hour at the only counter.
The counter operator can manage to serve on average 10 customers/hour. Is a queue going to form?

Average arrival rate: 12 customers/hour on average at the only counter

Process service rate: the counter operator manages 10 customers/hours
The queue is going to form.

Example 2:
A bank historical data show that on Monday between 10 and
11 am there are on average of 10 customers/hour (1 customer
each 6 minutes) at the only counter. The counter operator can
manage to serve 12 customers/hour on average (1 customer
each 5 minutes). Is a queue going to form?

Average arrival rate: 10 customers/hour on average at the only counter 1 customer each 6 minutes
Process service rate: the counter operator manages 12 customers/hours 1 customer each 5 minutes
The queue is going to form, but it depends.

NB) The queue itself is not a problem, but there are other related elements that create issues.

NB) Average service rate should be higher than the average arrival rate to have structural balance and to be able to
avoid the queue.

Variability and uncertainty:

Variability: is the same concept of variance, so it is related to statistics. We have an average, which is
associated to a certain kind of variability, standard deviation and other statistical parameters (those of
casual variables: normal distribution, negative exponential distribution)

WHY DO QUEUES FORM?

Queues form due to the not perfect counterbalance between the demand rate and the service rate (demand and
offer).
Structural imbalance: servers are not able to satisfy the arrival rate (Service rate < Arrival rate)
Incidental imbalance: it depends on the situation and all events have different probabilities to occur:
 lOMoARcPSD

Variability: for example, if we consider 6 activities (A, B, C, D, E, F) which need to be accomplished in

1 hour by only one server. There is a kind of sequence we should follow and we should be able to
avoid any kind of theoretical overlapping between these activities. If one server is doing an activity,
like serving a customer, cannot serve other customers at the same time. The following activity will
not come naturally after the first activity ends (because we are considering casual variables). For this
reason, in reality we do not have structural balance, because the arrival time is a stochastically
variable, so we have to deal with queues.
Uncertainty

A FEW MESSAGES:
Usually the tendency is to consider average values only, which are a good indicator. But there’s always a
variability that needs to be considered and managed (standard deviation or variance, according to the
chosen one)
Forecast and historical data help, but they are not the truth. They are extrapolated with specific techniques
to be used as inputs for our considerations.
Structural condition of queue systems is that the average pace of customers’ arrival in the system is lower
than the one of the server’s distribution of the service: we will never operate with structural unbalance, but
with a simplified reality
Queues are a result, a symptom: queues are consequences of some choices that were made on the
production/service system. They are not just the result of mistakes, but also wrong considerations that led
to choices. EX) In the service systems, there are not stocks, but people, but we use the same considerations,
so this can lead to wrong choices. “Stock cover problem”

QUEUE INEVITABILTY

WAITING LINES vs STOCKS

There is a clear relationship between them, so we look at the similarities between production and service systems:
 lOMoARcPSD

CHARACTERISTICS STOCKS IN A PRODUCTIVE SYSTEM WAITING LINES IN SERVICES

Costs Capital opportunity cost: margins are Time opportunity cost: if the customer in line is not
put in stocks and not in other served, the margin are not profitable, so time is
resources crucial

Space Warehouse Waiting area

Quality of the Inefficiencies index: for example, it Negative impression: there is no inefficiency, but
system measures the impact of stocks in only negative impressions which inspire clients
lower quality production systems
Decouplement Allow to decouple the productive Allow job division and specialization: separating
process phases different tasks of servers to different customers

Utilization level WIPs allow to keep machines always Waiting customers keep the servers busy
busy (saturated)
Coordination There is no need of a detailed Allow to not precisely balance demand and offer:
scheduling (coordination) when there are a lot of customers waiting, I cannot
organize their arrival, so I try to put as much as
servers as we need according to specific situations.

WHY queues are particularly critical in the service companies?

Because pieces waiting do not create problems, while customers , which are human being, create more problems
related to psychological behaviours.

QUEUEING SYSTEM MODELING

CUSTOMERS’ BEHAVIOURS
Rejecting: when a customer or a product is rejected by the system because he/she/it doesn’t respect
acceptance requirements (constraints to enter the system). For example, if a piece arrives to a machine but
it is not expected to be worked with that machine, it is rejected. It refers both the production and the service
systems.
Balking: when a customer decides not to enter a queue because it’s already too long. It refers only to the
service system (customers).
Reneging: when a customer already in queue gives up the service (maybe because he/she is waiting too
much) and goes away without being served.
Jockeing: when a customer tries to trick the rules to get advantage, e.g. shifting from one queue to another
to shorten the waiting.

QUEUEING SYSTEM CHARACTERISTICS

Service process: with some features

Then there are two options:
1. No future need for service
2. Some defects which need to be corrected or a customer is not satisfied, so he renter the system to be re-served.

CALLING POPULATION
Customers’ population: it’s the input source. We can have a finite or infinite population:
 lOMoARcPSD

Finite: if the potential number of new customers for the system is significantly affected by the number of
customers already in the system. (If already one customer is in the system, he/she can influence of other
new customers who could come)
Ex: Arriving to a copier shared between three secretaries

Infinite: if the number of clients in the system does not affect the demand pace for the service made by new
customers. It is a potentially unlimited number, not in any case but it is probable.
Ex: people arriving at the ER

ARRIVAL PROCESS
The arrival process describes how customers show up. It’s described by
the distribution of inter-arrival times, that is the time interval occurring
between two consecutive arrivals (for example the first customer arrives
in 2 minutes and the second one in 4 minutes and so on). Generally, data
are collected by collecting the actual arrivals time.
Different empirical studies show that very often the distribution of inter-
arrival is well described by a negative exponential distribution.

Frequency graph: all bars in the histogram are the different time arrival of
customers. X: inter-arrival time.

ARRIVAL PROCESS - Negative exponential distribution

EXAMPLE: The average inter-arrival time between 2 patients to the doctor is 2.4 minutes. Considering that a patient
is just arrived, what is the probability that another will arrive in the next 5 minutes?

t in this case is in minute, so everything is reported in minutes. First of all, we calculate lambda. Then we use lambda
in the formula of F (5), where t=5. We obtain that F (5) = 87.6% which is the probability that a patient will come in
the next 5 minutes. This result reports both uncertainty and variability: we have the variance and uncertainty related
to the stochastically variable used.

ARRIVAL PROCESS - Poisson distribution

The Poisson distribution is in a one-to-one
relationship with the exponential
distribution. When the inter-arrivals time
are exponential, the number of events N(t)
that take place in a given time t is a Poisson
process.
The Poisson distribution has a discrete
probability function as:
 lOMoARcPSD

Poisson process is quite variable, not continuous. There is always the stochastically variable.

EXAMPLE: At a telephone exchange phone, calls arrive at a pace of 12 calls per hour. Knowing that the number of
calls can be shaped with a Poisson distribution, answer the following questions.
1. What is the probability that 10 calls will arrive in the next hour?

First of all, we consider the number of events, n = 10. Then, we have to consider the unit of measure of time, t = 1
hour (reference unit is hour). Lambda is an average, 12 calls/h.
We apply the formula to get P (10 in 1 h) = 14.8 %

2. What is the probability that there won’t be any call in the next 5 minutes?

We have n = 0 and the reference time is always hours, so we translate 5 minutes in hours and we get P ( 0 in 0.083h)
= 36.79%.

3. What is the probability to have more than 2 calls in the next 10 minutes?

If we have P (N >2), to get it we need to calculate P (N= 0), P(N=1) and P(N=2), so P(N>2) = 32.42%

SERVICE PROCESS

The service process describes how each server delivers the service, in
particular it defines its duration. It’s defined in terms of service times
distributions. The service times distribution is often well represented
by a negative exponential distribution, thanks to historical data
collected and demonstrations.
Usually data are collected by collecting the actual service times (we
have the average service/processing time).

NUMERICAL EXAMPLES
Pierre Cahon, barman of one of the most popular discotheque of St. Catherine Street, Montreal, serves cocktails at a
pace of 1 every 50 seconds. Recently, during an evening when the club was particularly full, every 55 seconds
someone was at the bar asking for a drink. How to shape this
system?

We have no structural unbalance and we need to model the

situation.
 lOMoARcPSD

NB) The queue are customers which are waiting in line.

The system is the queue of customers plus the server.

23/3/16

WHICH PARAMETERS CAN BE RELEVANT?

n=total number of customers.

q stands for the line (queue) and s for the system

This ratio is mediated by the number of servers in that given moment and that given system.
Po and Pn are probabilities referred to a given period of time (one hour or a time bracket).

NUMERICAL EXAMPLES
In the Pierre Cahon’s example case:
a) Assuming that Pierre is able to serve customers following a FCFS (First come fist served) method, how much would
you expect to wait before you’ll have your drink in hand?
b) How many people would you expect to find in line on average?
c) What is the probability that three or more people will be in the system?
d) What is the probability that you can be served immediately without none in line?
e) What is Pierre’s saturation level? (how busy is he?)

We are considering formulas for the M/M/1 system.

There are different modelling systems with different formulas.
Lambda and mu as a preliminary data must be calculated. Then
we apply formulas.
e) Ro is lambda over mu. Saturation rate is equal to ro.
90.8% of the time (hour of the day) the server is busy, so 90.8%
we will have queue. Even if in theory the service rate is higher
than arrival rate, we can have queue due to variability and
uncertainty of service and arrival processes.

NB) 90% of exercises are based on natural language and not numbers. Parameters must be known, definition and
formulas are given in a table. It’s a matter of interpretation
 lOMoARcPSD

NUMERICAL EXAMPLE
A consultant is searching for a way to understand the exact number of counters to leave open in a new branch bank,
in order to be able to offer a good service to the customer. How many counters should be left open in the rush hour
if you want to keep the waiting in line time below 15 minutes? Each counter will have its own queue of customers in
front.
The consultant, from previous surveys, knows that during the rush hour there are on average 100 customers per
hour and that each server at the counter takes on average 5 minutes to be able to serve each customer (arrival and
service distribution: negative exponential).

We know lambda and mu, so we can calculate ro = 8.33 which gives us information about how many servers at least
are needed in order to have structural balance.
Which preliminary indications can we give to the consultant based on these data?
Minimum number of servers has to be greater than 8,33 to avoid structural imbalance. At least 9 servers are needed
for this system.

Then we can calculate Wq and Ws (we use min as unit of measure). Ws is the serving time added to Wq. We need to
calculate Wq. We can see that, if we want to keep the waiting in line time below 15 minutes, we need to have 12
servers in the system.

The saturation varies with the number of servers of the

system, so we start from 9 to see how the system changes
varying the variable of number of servers.
The saturation level is almost linear, without gaps
between the values of different situations.
The Wq curve presents a big gap between 9 and 10. This is
the general behaviour of the system, which have been
demonstrated.
Increasing by 1 the number of servers, from 9 to 10
(+11%), the throughput time decreases of 60%.
 lOMoARcPSD

Just adding one resource to the minimal needed (9), the saturation decreases, but the important thing is that we
have a very huge decrease of Wq. The relationship between saturation and Wq is not one-to-one related (not
linear), but only a consequence.

WAITING TIME vs SATURATION

100% of saturation means that every second someone
arrives, while 0 means that there are no arrivals.

The waiting time doesn’t grow in a linear way with the

increase of the system saturation time.
Even with a system utilization rate rather low (around 60%)
there will be delays.
In the last part of the graph (increase), as far as we go from
the limit of 9, the slope of Wq increases a lot.

Even with a system utilization rate rather low (around 60%) there will be delays, so we have queues, a time that on
average is between 0 and 17.5

WAITING LINES MANAGAMENT IS USED FOR:

System sizing and designing (ex. consultant who needs to understand how many counters at least have to be
left open): to provide advice about how to size and design the system
Operative management, to evaluate costs and gains to improve the service (ex. barman who tries to
understands how to make his customers more satisfied): to evaluate the feasible solutions and choose
between different solutions.
Queue length (Long lines suggest a poor customer service and cause congestion and dissatisfaction)
Average waiting time in line (Long waits are linked to a poor service)
Average waiting time in the system (it can suggest problems with customers, servers’ efficiency or
capacity)
System utilization rate (Check costs without unacceptable reductions in services)

Application to size the systems, achievable goals:

SERVICE QUALITY PRINCIPLES

We define an objective and we try to model the system according to it. In this case the objective is about the quality
of the service level.
Average waiting time of the customer: banks, restaurants
EX) Sizing with the goal that the expected value of the time that the customer has to wait
before sitting at a table is less than 10 minutes.
The expected value of the waiting line should be less than a parameter.
Maximum waiting time: public services (ambulance, fire brigade)
EX) Sizing with the goal that the probability that the client has to wait more than 5
minutes before receiving the service is less than 5%.
The probability of having the average waiting time higher than a given parameter (w)
is lower than another given parameter which is a probability (%, alfa).

COST PRINCIPLES
It consists in the minimization of the sum of the wait and the service
costs (trade off).
For services we need to consider both the service cost (the cost
needed to serve the customer) and the wait cost (it is mainly an
opportunity cost: the cost of having a large waiting room can be
another component added to the opportunity cost, but it does not
represent the majority. In the opportunity cost considered as the
wait cost, there is also the opportunity that one customer is leaving
the queue). Each customer in line is a loss for companies, but at the
same time it is impossible to avoid queues, so they can only diminish
them as much as possible.
 lOMoARcPSD

EXAMPLE
G/G/1 and G/G/m systems formulae
The assumption of exponential arrival and processing times is convenient to mathematical computations.
However, in practice, process times in particular are rarely truly exponential. This is why it is important to have some
idea of how a G/G/1 and G/G/m queue behave. However, exact mathematical relationships are not possible with
such distributions. Therefore, some kind of approximation is needed. The one here is in common use and it is for
practical purposes. For G/G/1 systems the formula for waiting time in the queue is as follows:

In this case, when we do not know the type of distribution (G stands for generic distribution) and we can not make
some assumptions, we have to use this corrected formula (Kinman formula).
The last ratio is corrected with two coefficients of variability, which give information about the variability and the
dispersion of variables: Car, for arrival and Cse, for services. They tell us how much standard deviation there is
compared to parameters. There is a demonstration behind this formula, it is not just empirical.
We can have Wq for systems with generic arrival and service processes.

NUMERICAL EXAMPLES
“I can't understand it: We have worked out our capacity figures and I am sure that one member of staff should be
able to cope with the demand. We know that customers arrive at a rate of around 6 per hour and we also know that
any trained member of staff can process them at a rate of 8 per hour. So why is the queue so large and the wait so
long? Have at look at what is going on there please.”
Sarah knew that it was probably the variation that was causing the problem. Over a two-day period when she was
told that demand was more or less normal, she timed the exact arrival times and processing times of every
customer. Her results were as follows:

We start collecting the initial

parameters. Especially for services the
coefficient of variability is very high (3.5
means 350 %).

“Ok, so I see that it's the very high

variation in the processing time that is
causing the queue to build up.
How about investing in a new
computer system that would
standardise processing time to a
greater degree? I have been taking with our technical people and they reckon that, if we invested in a new system,
we could cut the coefficient of variation of processing time down to 1,5. What kind of a different would this make?”

Reducing the variation of the process time has reduced average queuing time from 2 .48 hours down to 0.61 hours
and has reduced the expected number of people in the queue from 14.68 down to 3.66.
 lOMoARcPSD

Using specific information system, in order to give a regular service processing time, they can also influence the
variability keeping the average the same. It happens that if before the distribution is like a standardized normal
distribution, then it becomes thinner (piu alta e piu stretta). This is the impact of changing only variability, reducing
the standard deviation.

SATURATION – QUEUE LENGTH AND VARIABILITY

For a given saturation level, according to variability, we will have
different queue lengths (or with a given queue length we can have
different saturation levels according to variability variation).
The higher the variability, with the same level of saturation, the higher
the length of the queue.

Graphical representation of the impact of variability on the length

of the queue.

(X) is the AS IS system performance. The queue is too long (then also the waiting time is too long). It is set a new
queue length target. Starting from (X) where there is high saturation level and long waiting time, we want to
improve the situation moving to:
(Y) If we just increase the number of servers/resources, we reduce the number of customers in line. We spend a
lot of money investing in machines and resources.
(Z) We can keep constant the resources, but we can reduce variability. We can also try to improve the variability
of the arrival rate, not only the service variability (service rate).

NB) We can work on two different types of variability separately (one or the other) or together.

QUEUE CONFIGURATION
The configuration is how the specific queue is organized.
1. Single queue 2. Multiple queue 3. Take a number 4. Infinite Servers

SINGLE QUEUE
Pros:
Assure a FCFS type of service
Avoid concern that another queue could be faster
 lOMoARcPSD

Minimize the average waiting time (MPORTANT/HARD thing): this type of queue minimizes the average
waiting time. If we have some problems with servers, like 4 servers and one of them is not available, all
customers are still served by others. We start from the assumption that there is no balking phenomenon, so
customers can change the line.
Reneging actions are less frequent: because usually the line is fast
Cons:
It could “scare the customer”

MULTIPLE QUEUE
Pros:
Allow to diversify the service: each queue is dedicated to a specific service. In general, for example in banks
they are used for practical reasons.
Allow to diversify the work
Customer can choose a server of his/her liking
Balking actions are less frequent
Cons:
With the same number of servers, the average waiting time is greater

TAKE A NUMBER
It is a variation of a single queue system where customers can take numbers.
Pros:
Assure a FCFS type of service
Avoid anxiety that another queue could be faster
Minimize the average waiting time
Customers have the possibility to relax and dedicate themselves to other things during the wait: it is a
matter of the psychology of the client. He/she can organize himself to do something. It is a soft aspect for
services.
Cons:
An absent-minded customer could risk to loose its turn
It could "scare the customer" (if he/she doesn't have anything else to do): I know the number and maybe it
is very high so I could be scared by it. It is a matter of psychology, because sometimes without the number
people decide to wait anyway.

INFINITE SERVERS
It is like having infinite servers

Examples
 lOMoARcPSD

ELEMENTS OF CHOICHE
1. Service customization degree
customization: MULTIPLE QUEUE
2. Service times variability
variability and unpredictability (like the failure of a machine): SINGLE QUEUE

QUEUE CAPACITY
Limited capacity: the number of customers in line plus the ones that are being served is limited. Customers
who arrive after this capacity is overfilled are rejected. Like in airports when airplanes take off, they have
limitations in term of day and hours.
Unlimited capacity: the number of customers in line plus the ones who are being served is potentially
unlimited. Like when we have a lot of space for customers (the space is not really unlimited, but the space is
not a constraint considering the average arrival rate).

NB) In exercises we consider unlimited queue!

TO SUMMARIZE QUEUE CONFIGURATION:

Queue is formed by customers waiting to be served
It’s possible to recognize between:
Queue configuration:
o Single queue
o Multiple queue
o Take a number
o Endless servers (self-service)

o Limited queue
o Unlimited queue

QUEUE DISCIPLINE
Queue discipline (or “ranking rule”) is the rule or set of rules with whom it’s established in what order customers are
served, it’s often strictly related to the queue configuration: according to changes of the system, the service
changes.
Static: the pertaining order between the customers that are present doesn’t change in time and/or at the
changing of the system conditions (FCFS, SPT,...)
Dynamic: the pertaining order between the customers can change in time and/or at the changing of the
system conditions

Priorities: some customers could be more important for some reasons, so we can assign some priorities to them
and not to other customers in the same line. They are the most important thing in queue discipline.

How many actions and what kind of actions we can do in order to influence and manage our queue?
3 things:
Levers and counter measures demand side: we can act on the demand side
Levers and counter measures offer side: we can act on the offer side
Levers and counter measures for soft aspects/related to psychology of waiting: we can act on the psychology
of customers (queue psychology) using inputs used for services

LEVERS AT DISPOSAL/ IMPROVEMENT AREAS

SERVICE – OFFER SIDE
Configuration changes
Add/Remove/Move resources (saturation of resources)
Decrease service time
Technology utilization
Training (of operators)
Increase resources flexibility: for example, in a multiple queue with different and specialized servers, we can
train them, changing the configuration using a single queue, to improve the efficiency of the system.
 lOMoARcPSD

Decrease service times variability: more related to organizational problems and skilled workers or problems
with machines (technical problems)

ARRIVALS - DEMAND SIDE

To overcome the main problem, the solution is booking: scheduling the customer arrival to be sure about the arrival
process of customers. This is used to optimize the profit (to match the demand and the offer in the best way
possible) and reduce the variability, acting on the efficiency.

Decrease uncertainty level

Booking
Improve forecasts
Decrease variance
Incentives for non rush hour
Booking

PSYCHOLOGY OF WAITING
Service laws:
1. Satisfaction = Perception - Expectation
Perception is more important than reality. If expectations are higher than perceptions, there is low
satisfaction, the contrary the satisfaction will be higher.
2. It’s hard to play catch-up ball
First impression is the most important, usually related to the queue of shops or buildings.
A psychological way to attract customers is to work on queues.

Example:
Measure the customer satisfaction is important to
understand where the critical points of the system
are. This is a customer experience map of a visit for
an X-ray investigation.

Each step can be analysed according to queue

theory and psychology queue theory to avoid
having bad results according to customer
satisfaction.

PROSPECT THEORY (TVERKSY and KAHNEMAN,

contemporary psychology)

My utility increase as far as my “perception of gaining

something” increases. Gain or loss perception depends on
the benchmark.

Loss sensitivity is greater than gain sensitivity

The slope of the utility in the gain perception part is lower
than the slope in the loss perception part.
When I start to be disappointed about the service, the
situation worsens very fast. It is important to avoid the
negative perception of clients.
WAITING PRINCIPLES
The unoccupied time seems longer than occupied time: distract and entertain with small related or
unrelated activities.
EX) Small show at Disneyland, mirrors near elevators, recorded music, magazines at the doctor.
Pre-process wait seems longer than in-process wait: taking care of the customer as soon as possible to make
him/her “in-process”
 lOMoARcPSD

Anxiety makes the wait to look longer: communicate and update frequently the waiting time to the
customer
An unfair wait seems longer than a fair one
The more the service is of value or the more the customer senses its utility, the more he/she is willing to
wait
Waiting alone seems longer than waiting in a group
A customer exposed to an exaggerated wait will be a more difficult one to serve or an ex-customer

Therefore, be careful to all these factors and make sure to organize some actions to avoid that the customer senses
the supplied service in a negative way!
The very important thing to avoid is to have a customer who is not going to be anymore your customer. We need to
provide the right experience.

QUEUE SYSTEMS ANALYSIS TOOLS

Generally, there are 3 possible approaches for queueing systems:

Deterministic analysis:
Pros:
It is very simple and intuitive to apply
Cons:
Doesn’t take transitory into account
Doesn’t take arrivals and service times variability into account
Queue as an ON-OFF aspect, with any statistical concept behind and any variability

Queueing theory:
Pros:
Takes inter-arrivals and service times variability into account
Allows to calculate a set of significant variables (Wq, Lq, Pn, etc..)
Cons:
Doesn't take transitory into account: the system does not change over time, it is a stationary system with no
chronological considerations, but only logical considerations. It provides a picture of the situation without
time dimension. Brief and effective way, by hand, to model the system.
Mathematic analysis can become very complex or even intractable in case of complex waiting lines
o Non Poissonian arrivals
o Non exponential service times
o Customers complex behaviours (balking, reneging, jockeying)

it is based on statistics, with averages, standard deviation, distributions and so on. It is a rough analysis, but a fair
approximation, with an average mistake between 5 and 20% according to real results, so it is acceptable.

Simulation:
Pros:
Takes transitory into account
Very flexible
o For the type of systems that can be shaped
o For the type of data that can be obtained (ex. times (%) that a specific customer waits more than 1
minute)
o Economic and user friendly software are more and more widespread
Cons:
Time consuming
Skills needed to design, production and analysis

It can overcome the problem of the transitory, so it is much more precise. There is some available software, like
Arena which cost a lot, that can be used in the design and modelling phase of the system, which can be very
complex: hours of simulation according to the complexity of the situation. Unfortunately, in some cases, it is the only
approach that can be used.
 lOMoARcPSD

QUEUES THEORY ASSUMPTIONS

1. Stationary process: this assumption could be very restrictive: usually the arrival process isn’t a static process,
it changes during the day. For example, in a bank, there are very different values of arrival process in a whole
day. a possible solution is to make more analysis considering different time slot.
2. Peculiar customers’ behaviours are not expected: no balking, reneging, jockeying
3. True only for certain inter-arrivals and service times distributions: negative exponential distribution are
assumed for service and arrival rate

KENDALL’S CODIFICATION
Kendall's codification turns out to be A/B/c/K/m/Z
A identifies customers' arrival distribution
B identifies service times distribution
c identifies number of servers (if we have c it means that there can be any number of servers)
K identifies queue capacity (buffer by default: endless)
m identifies population dimension (by default: endless)
Z identifies service discipline (by default: FIFO)
Usually stop at the first 3 components (ex.M/G/1), in detail:
o M suggests a negative exponential distribution (Markovian)
o G suggests a general (i.e. arbitrary) distribution

EXAMPLE:

Most common models:

Standard M/M/1
Finite-queue M/M/1 (M/M/1/K)
Standard M/M/c
Finite queue M/M/c (M/M/c/K)
General Self Service M/G/∞: It's a system with an endless number of servers or a system where customers
serve themselves. A possible example is the supermarket where, at least during the acquisition phase,
customers do not have to get in line. The total number of customers in the process varies due to arrival and
service times variability. This model is also useful to shape, rounding situations where you rarely have to
wait (ex. Emergency services).
We use always infinite queues with M, both single and multiple queues: M/M/1 and M/M/c

EXAMPLE
Population: endless or very big
Arrival process: inter-arrival times are described by a negative exponential
Queue configuration: single queue without capacity restrictions and balking or reneging effects
Queue discipline: FCFS
Service process: 1 server with a negative exponential service times distribution

This kind of queue formally is the same as M/M/1, but in this case the queue has a finite dimension, so we use
M/M/K.
Pn = probability to not enter in the system
Pn = expected number of lost customers
 lOMoARcPSD

This model is particularly useful in estimating lost sales due to an inadequate waiting area or to an excessive long
queue.

EXAMPLES: How to shape the following systems?

1. At a post office 5 postal clerks serve customers following this logic: the first free clerk serves the next
customer in line (There's space 40 customers at top at the branch).

We have no information about distribution, so we can suppose M/M and there is a FIFO logic of a single
queue situation, so: M/M/5/40.

2. At the BuonaSpesa supermarket there are three cash counters. The firsts two serve customers indistinctly,
customers chose autonomously the line. The third cash counter is dedicated to customers with 5 products in
the bash at top. The supermarket capacity is not a restriction to customers' possibility of getting in line.

We can assume M/M, but we have to understand if there is single or multiple queue. Single lines are
independent, so we have 3 M/M/1.

3. From Milan - Linate airport, airplanes take off from the airport only strip.

There is only one strip so only one server. There are constraints and no information, we can only say
M/M/1/K. The service rate is the time for taking off. The arrival rate is a negative exponential. Limited
number of airplane in the waiting lines, K.
Another interpretation can be: G/G/1/K. If we change the assumption.
 lOMoARcPSD

30/03/2016
MANAGING WAITING LINES - Exercises
EXERCISE 1
In the first morning hours, customers arrive at the post office where there's a single "take a number" queue, at an
average pace of 54 customers/hour (Poisson) while each server can manage to complete a service in an average time
of 4 minutes (negative exponential).
a) If there are 6 servers at the counter: what is the average number of customers in the system, the average waiting
time in line and in the system?
b) Determine the smallest number of counters that need to be opened to keep the average time in the system lower
than 10 minutes
c) If an employee costs 30€/h and customer's waiting time stands for a cost of 20€/h for customer, what would be
the optimal number of servers according to purely economic considerations?

NB:
Single queue, take a number queue M/M/c system
- Arrival rate Poisson distribution (discrete distribution) is the same to say that inter-arrival time are distributed as
a negative exponential, so it’s a traditional case.
- Service rate negative exponential

a) Ls? Wq? Ws= Wq + average service time (4 minutes)

b) Ws < 10 minutes, smallest number of counters?

c) we have an objective function that should be minimized economic considerations!

SOLUTION:
The situation can be shaped as a M/M/6 system, where:
= 54 customers/hour
= 60/4= 15 customers/hour

THEORY FORMULAS PDF: we have data about and c (number of servers) and Lq values, which depend on both
and c. If we consider fixed, when c increases Lq decreases.

OSS: usually service rate is given, so we have to calculate Lq, Ls, Wq, Ws.

a) We know = / = 54/15 = 3.6 and c = 6, so we have from the table that Lq = 0.295 (average number of
customers in line).
Ls = Lq + = 0.295 + 3.6 = 3.895 customers
Wq = Lq / λ =
0.295/ 54 = 0.00546 hours around 20 seconds (19.67 seconds)
Ws = Lq / λ + 1 / μ = 0.295/ 54 + 1/15 = 0.072 hours around 260 seconds (259,67 seconds)

NB) Remember that every result is obtained in hours, because the service and arrival rate are given in hours, but we
need to convert them in seconds, so we need to multiply the results by 3600.

b) Goal: Ws < 10 minutes and having the smallest number of open counters.
The arrival rate and service rate are given, so they cannot be modified, the only thing that can be modified in Ws
(our changing parameter) is Lq which is a function of (which is given) and c. So, our changing parameter in Lq is c.
We consider the following equation:
10 (min)/60 (min/h) > Lq/54 + 1/15
 lOMoARcPSD

and we obtain that Lq < 5.4. is given so from the table, we know that c should be ≥ 5.

c) Goal: calculate the optimal number of servers that minimizes

service and waiting cost.
Waiting cost = 20 €/hour * Lq (it varies according to number of
servers: it’s not linear but decreases with the number of servers c. If
c increases, Lq decreases but not in a linear way)
Server cost = 30 €/h * c (linear with c)

GOAL: min 30*c + 20*Lq c=5

EXERCISE 2
At a photocopier shop, customers usually enter to make copies or send a fax. Once they have finished, they go to the
cash counter to pay. The customers' arrival rate is 30 cu/h. Of those, 2/3 enter to make copies while 1/3 to send fax.
The arrival rates are distributed like a Poissonian. On average a customer takes 2 minutes at the photocopier; 4
minutes at the fax; 1 minute to pay.
Supposing that service rates are distributed like a negative exponential:
1. What is the probability that the shop is empty?
2. What is the system average throughput time?
3. How many people on average are in line at the cash counter?
4. How many customers on average are at the shop?
5. What is the probability to have less than 6 people in line at the photocopier?
6. What is the probability to have 1 customer at the shop?

SYSTEM MODELLING:
Customers enter the system and they can do 2 different things: photocopies and fax. In this complex system of the
photocopier shop we have 2 flows, each of them with 2 stages (1 stage in common). We have 3 M/M/1 stages: one
for fax, one for photocopy, one for the cash counter.
1) some customers (2/3 = 20 customers/hour = ) go to the photocopier then they go to the cash counter
2) some customers (1/3 = 10 customers/hour = ) go the fax then they go to the cash counter
3) stage in common cash counter: = 20 + 10 = 30 customers/hour

SOLUTION:
1) P0?
 lOMoARcPSD

We have 3 M/M/1 stages and in each of the 3 stages P0 should be = 0.

This is a joint probability, which is a multiplication of the 3 P0 (to obtain the P0 of the whole system).
P0 of the whole system= 5.55%

2) Ws of the whole system?

We apply the formula of Ws for each of the 3 stages:
1. Photocopier throughput time: Ws = 1/ (μ - λ) = 1/10 hours = 6 minutes
2. Fax throughput time: Ws = 1/ (μ - λ) = 1/5 hours = 12 minutes
3. Cash counter throughput time: Ws = 1/ (μ - λ) = 1/30 hours = 2 minutes
The expected throughput time of the whole system is the WEIGHTED AVERAGE of all the throughput time of the
different flows (different families of customers of the system = family of customers that do exactly the same
process/flows inside the system) of the system.

Family 1 (2/3 of customers): photocopy cash counter: Ws = 6 + 2 = 8 minutes

Family 2 (1/3 of customers): fax cash counter: Ws = 12 + 2 = 14 minutes
WEIGHTED AVERAGE: Ws = 20/30*8 + 10/30*14 = 10 minutes

3) Lq at the cash counter?

Lq = ρ*λ / (μ - λ) = 0,5 customers

4) Ls of the whole system? (is the sum of the Ls of each sub-system)

Ls shop = Ls photo + Ls fax + Ls cash counter
Ls= λ / (μ - λ) = 2+2+1 = 5

5) Probability to have less than 6 customers in line at the photocopier

The probability to have less than k customers in the system (in line + in service) is equal to:

𝑃(𝑛 < 𝑘) = 1 − 𝑃(𝑛 ≥ 𝑘) = 1 − 𝜌𝑘

Probability to have less than 6 customers in line = probability to have less than 7 customers in the system (1
customer is supposed to be served while the other 6 are in line).

𝑃(𝑛 < 7) = 1 − 𝑃(𝑛 ≥ 7) = 1 − 𝜌7

20 7
𝑃(𝑛 < 7) = 1 − 𝜌7 = 1 − ( ) = 94.15%
30

6) What is the probability to have 1 customer

at the shop?

4 situations:
1) fax and cash counter empty and 1 in the
photocopy
2) fax and photocopy empty and 1 in the cash
counter
3) cash counter and photocopy empty and 1 in
the fax
4) fax, photocopy and cash counter are all
empty (it is not considered)
 lOMoARcPSD

CONSIDEREATIONS:
A complex system is the composition of more elemental sub-systems that interact among themselves
(mapping the system and identifying every step of the process).
Very often in a complex system transit different type of customers (knowing in advance every type of
customers of one own system).
The system expected throughput time is a representation of an "average" customer and not of the specific
one!
The throughput time varies according to the variation of the work load in the system and to the variation of
the input.

NODES BALANCING TO FLOWS CALCULATION

NB) We don’t have loss of material or of components in queues. I can have scrap in the system, but I’m not losing
materials. This is the base of the nodes balancing (similar to Thermodynamics).
NB) The output in queuing system is ALWAYS given, it’s fixed. According to a given output, we have a complex
system to model.

Example:
The test phase of a stage is a stage where enter, without different priority logics, product A and B that form the
productive selection of the company. In particular, the testing machine (the only stage) isn't of the latest generation.
Indeed, every once in a while (sometimes) it could signal as nonstandard even products that actually are good. For
this reason, in case nonstandard products are signalled, they are tested once more. On average 15% of products A
and 15% of products B that the machine tests, have to be tested again. The pieces that have to be rejected are 50%
of the pieces that have been tested twice, both for products of type A and of
type B. The rest of the products instead are good and can be delivered to the
customer.
In case the stage could work 660 products/hour and in case the company
should deliver at the customer 300 good products of type A and 200 good
products of type B every hour, how much material is it necessary to be put in
the test phase every hour? Shape the system reporting all the significant
parameters.

The test machine is not perfect. It is the only stage: M/M/1. A percentage of the products that are tested the second
time are scratch, but another percentage are good. At the end we will have some scratch that will be eliminated,
because finally they are not good, so we lose some material. How much of product A and B should we put in the
system knowing that there is a loss?
Everything that enters in the node is equal to everything that comes out from it. Nothing is created or destroyed!!
The total sum of the different flows, for a given node, should be equal to 0. This is valid for the single node or for the
whole system.
Everything that enters in the system is equal to everything that comes out. In the system there isn't any flow loss
(the service process works on customers, without creating nor destroying any!)
 lOMoARcPSD

1) First of all: calculate Xa and Xb, which are

necessary different from 300a and 200b!
Each time I have a node, I have some changes.
What is going out of the node is 300+0.15Xa+
0.5*0.15Xa and should be equal to what is
entering.

2) Then I can calculate InputA and Input B.

Da fare a casa: fare la stessa cosa per l’intero
Sistema.

EX 1 - RINASCIMENTO CLINIC

The system average throughput time is around 25 minutes (Ws). But the
patients complain of a very variable time and that the reason behind it's not
clear. Time varies from less than 5 minutes to more than 45. Customers'
complain risks to damage the centre image and to make it lose customers! A
specific analysis of the process is necessary to comprehend the causes of
variability.

Each process is characterized by a certain patients' arrival rate and a certain service capacity.
Both parameters influence queue length and throughput times. Know the system
The system is formed by different entities (different type of patients) who have different behaviours and
paths. Identify different typology/families

1 - MODELLING OF THE SYSTEM: (mapping of what is provided by the text of the exercise)

We have 3 M/M/1 systems and 3 stages (3 families of customers/type of paths in the system):

1) Receptionist + Doc. Carter + out of the system

2) Receptionist + Doc. Carter + Doc. Romano

3) Receptionist + Doc. Romano + out of the system

The service rate is given by the text, while the arrival

rate must to be calculated using the information of the
text. If we know that 30 customers arrive at the
receptionist, we also know that 30 customers should
exit the system definitely. 30 customers/h are received
by the receptionist and 2/3 of 30 (20) are visited by
Doc. Carter and 1/3 of 30 (10) by Doc. Romano. We
have to consider the fact that 15% of Doc. Carter
patients are visited also by Doc. Romano, so 15% of 20, which is 3. At the end we obtain that:

arrival (at the receptionist) = exit (Doc. Carter + Doc. Romano)

30 = (85% * 20) + (10 + 15% * 20)

 lOMoARcPSD

INACTIVITY TIME

First of all we remember that the saturation rate is over , which is in specific situations. The inactivity
probability is the complementary of the saturation rate. Then the inactivity time is the inactivity probability
multiplied by 60 (min in 1 hour).

Each customer who enters the system can follow 3 different paths, so each of them belongs to one of the 3 families:
path 1, path 2 and path 3. Each path can be studied in different stages of the system, in fact to calculate the Wq of
the system we have to calculate Wq for each stage and then summing up (the same for Ws).

NB) When in a system we have more than one family, we have to consider the fact that there can be different
number of customers in each path, especially when we calculate average values. For example, if one family has the
majority of customers while the others have only a small number of customers, the average will not be simply
calculated as always otherwise the final result will be wrong!

2 - CALCULATIONS

a) + d) First of all we calculate:

Ws is given by the sum of Wq and service time.

Why does simultaneously exist an average waiting time and a server’s inactivity time?

Reception (Wq) Doc. Carter (Wq) Doc. Romano (Wq) Wq TOT

Path 1 4.5 minutes 4 minutes - 8.5 min
Path 2 4.5 minutes 4 minutes 26 minutes 34.5 min
Path 3 4.5 minutes - 26 minutes 30.5 min
b) + c)

Reception (Ws) Doc. Carter (Ws) Doc. Romano (Ws) Ws TOT Occurrence Expected time
Path 1 6 min 6 min - 12 min 0.5666
Path 2 6 min 6 min 30 min 42 min 0.1
Path 3 6 min - 30 min 36 min 0.3334
EXPECTED THROUGHPUT TIME 1 23 min

Occurrence = the frequency (percentage) of the single path. In a system, the sum of the occurrences of different
paths should be 1. Total occurrence MUST be not less than 99.98% ≈ 100%.

Occurrence path 1 = 20/30*85%= 0.5666 (85% of 20 [= 100% - 15%, where 15% go also to Doc. Romano]) 56.66%
 lOMoARcPSD

Occurrence path 2 = 3/30= 0.1 10%

Occurrence path 3 = 10/30= 0.3334 33%

Check: 33+10+56≈100

The expected throughput time is given by the weighted sum of the average throughput times of each path.

EXPECTED THROUGHPUT TIME = ∑𝟑𝒊=𝟏 𝑾𝒔 ∗ 𝑶𝒄𝒄𝒖𝒓𝒓𝒆𝒏𝒄𝒆 =

= 12*0.5666 + 42*0.1 + 36*0.3334 = 23 min

Similar to 25 that we suggested at the beginning.

The system expected throughput time refers to an "average product". It doesn't have to be taken as an
absolute benchmark.
It's crucial to know the system and to map it: which phases? which products? which problems?
For each type of product, the variability is much lower: it is explained a part of the initial uncertainty
If it's possible to assign the type of product at the beginning of the process, the uncertainty decreases and
it's possible to better align performances to expectations, therefore it's possible to start effective
countermeasures.

NB) Check to auto-evaluate: check the balancing of the system at the beginning (if flows/nodes/calculations are not
correct).

Possible mistakes:

1. Wrong modelling of the system (flows): it can be right with our assumptions, but not in the reality
2. Wrong balancing of the nodes of the systems
3. Fail in calculating the occurrences: it can be a signal.

THEORY: PRIORITIES

We can have priorities only for M/M/1 system. We consider a M/M/1 system with assigned priority to a customers’
class. It is a system formed by just one server who is able to serve two classes of customers (1 and 2) that have a
different service admission.

HP: customers of class 1 have a service admission priority higher than customers of class 2.

1) PREEMPTIVE PRIORITY

If a customer of class 1 enters the system while a customer of class 2 is being served, the service to the customer of
class 2 is interrupted in order to give immediate precedence to the customer of class 1. This is done, for example, for
customers who represent 90% of the turnover, so it is mainly done for economical and service reasons; there is
always a business motivation behind. Another reason might be the presence of customers with a certain degree of
variability (usually high) which cannot be afforded. For these customers are adopted special treatments.

REMEMBER: Different priorities, same service rate, different arrival rate ( 1, 2, etc…)
 lOMoARcPSD

2) NON PREEMPTIVE PRIORITY:

If a customer of class 1 enters the system while a customer of class 2 is being served, the service to the customer of
class 2 is terminated and then the customer of class 1 is taken care of. (in this case the server first finishes to serve
the customer who is being served of his class and then he takes care of the newcomer customer who belongs to
another class.)

Back to the exercise: now we consider a system with priority. We have changed only the Ws for Doc. Romano, so the
path (type) 1 hasn’t changed.

The alteration of the priority logic doesn't change the system expected throughput time. Only the throughput time
of each single path changes, but not the average one.

However, in some cases it's essential to increase the customer's satisfaction degree and to improve customer
service.
 lOMoARcPSD

31/3/16
Typical exercises of the exam: the approach/reasoning/method is the same requested at the exam
NB) There is only one possible mapping of the exercises. Always do a draft while reading the text.

EX 2 – STERILIZATION CENTRE “A.O. CURIAMO TUTTI”

READING THE TEXT:
There is just one flow entering the system, which is the sterilization centre.
1. The first stage is represented by the nurse (M/M/1). She receives urgent and not urgent kits, where urgent
kits have pre-emptive priority on not urgent kits.
2. The second stage is divided in 4 different specialised departments: Orthopaedics, General Surgery,
Gynaecology and Ophthalmology. Each of them has single queue configuration. In the Ophthalmology area,
special kits have non pre-emptive priority in the normal ones.
3. The third stage is characterized by 2 specialized operators OSS working in parallel (multiple queue
configuration), where we have 2 M/M/1.

At the beginning it is important to reason with variables and then we quantify them, for example we name INPUT
and we do not use numbers.

QUESTIONS:
a) Mapping the system in a detailed and precise way, putting all the important parameters.
b) Calculating the average time between an instruments kit enter in the sterilization centre, until it is ready to be put
into the sterilisation machines. Throughput time
c) Which kind of improvements you would like to propose in order to improve the whole system? (examples:
decreasing cost without increasing queues, reducing queues without increasing resources etc.). You are required to
motivate each proposed improvement.

Framework to solve the problem:

1. Shape the system and define the main parameters (μ, λ, queue’s types, ...)
2. Understand different kinds of ‘’customers’’ and their paths
3. Compute the frequencies of the paths
4. Calculate throughput times
5. Decide the strategic actions to improve the system

A) MAPPING OF THE SYSTEM AND DESIGN ACTUAL CONFIGURATION

INPUT

We already have the service rate, so we have to

calculate the arrival rate of each area, .
 lOMoARcPSD

B) ACTUAL STATE table

It always contains: , , , type of priority, type of queue

= (60 min/h)/ (# min/kit), where # min/kit = # min/# kit

How to choose the best node to calculate the variables? We should choose the simplest as possible, according to
experience. We can usually choose according to the system.

IDENTIFICATION OF PATHS

First of all, we have to take into account that there are two kinds of kits:
Nurse urgent kits
Nurse not urgent kit
Then we also need to take into account priorities: all the paths in the system should be multiplied by 2, because
urgent and not urgent could have the same path.

6 different paths:
• Nurse Opth (special: with priority) OSS
• Nurse Opth (normal: without priority) OSS
• Nurse Ort OSS
• Nurse Ort Sur OSS
• Nurse Gyn OSS
• Nurse Sur OSS

In total we have 12 paths, because we multiply them by 2. Without priority in the system we would have had only 5
paths.

Using formulas is more complicated, but it is easier to check possible mistakes.

PATHS’ OCCURENCES: how are the products distributed among the different paths?
 lOMoARcPSD

We start with urgent, then for non urgent is the same, we just need to change the arrival rate: u and nu.
We start using u = 5.
We calculate the occurrences starting form the left to the right of the system. Every time we have a node, the
frequency (occurrence) changes.
URGENT
1. Nurse Ort OSS
(5 * 0.3 – 0.15 * 0.3 * 5)/15 = [5 * (0.3 – 0.15 * 0.3)]/15 = 8.5%
2. Nurse Ort Sur OSS
(0.15 * 0.3 * 5)/15 = 1.5%
3. Nurse Sur OSS
(0.2 * 5)/15 = 6.67%
4. Nurse Gyn OSS
(0.3 * 5)/15 = 10%
5. Nurse Opth (special: with priority) OSS
(0.2 * 0.6 * 5)/15 = 4.0%
6. Nurse Opth (normal: withot priority) OSS
(0.2 * 0.4 * 5)/15 = 2.67%

NON URGENT
7. Nurse Ort OSS
(10 * 0.3 – 0.15 * 0.3 * 10)/15 = [10 * (0.3 – 0.15 * 0.3)]/15 = 17.0%
8. Nurse Ort Sur OSS
(0.15 * 0.3 * 10)/15 = 3.0%
9. Nurse Sur OSS
(0.2 * 10)/15 = 13.33%
10. Nurse Gyn OSS
(0.3 * 10)/15 = 20.0%
11. Nurse Opth (special: with priority) OSS
(0.2 * 0.6 * 10)/15 = 8.0%
12. Nurse Opth (normal: withot priority) OSS
(0.2 * 0.4 * 10)/15 = 5.33%

The total number must be equal to 100% (99.98 max, 99.90 is wrong). YES

THROUGHPUT TIME CALCULATION (Ws)

1. Nurse:
NB) E(S1) and E(S2) formulas must be
multiplied by 60 min/h.

2. Ophthalmology
 lOMoARcPSD

NBENISSIMO) Per calcolare Lq quando ρ non risulta preciso come i valori della tabella, bisogna impostare
𝜌−𝑦1 𝐿𝑞−𝑥1
l’equazione della retta passante per 2 punti: 𝑦2−𝑦1 = 𝑥2−𝑥1
3. Orthopaedics
Queue system, M/M/3 type
ρ = 4,5 / 4 = 1,125
Lq (1,125; c=3) = 0,073
Ws= Lq/λ + 1/μ = 0,073/4,5 + 1/4 = 15,97 min

4. General surgery
Queue system, M/M/2 type
ρ = 3,675 / 3 = 1,225
Lq (1,225; c=2) = 0,744
Ws= Lq/λ + 1/μ = 0,744/3,675 + 1/3 = 32,15 min

5. Gynaecology
Queue system, M/M/2 type
ρ = 4,5 / 5 = 0,9
Lq (0,9; c=2) = 0,229
Ws= Lq/λ + 1/μ = 0,229/4,5 + 1/5 = 15,053 min

6. OSS:
2 queue systems, M/M/1 type
For each queue: λ = λOss /2 = 7,5
Ws=1/(μ -λ)=1/(12–7,5)=13,33min

NB) when we have multiple queues (2 M/M/1, two servers with independent queues), we have to apply the formula
of Ws for M/M/1, but is obtained dividing OSS by the number of servers of the stage (2).

C) ACTIONS TO IMPROVE SYSTEM PERFORMANCES:

1. Single queue system for OSS (40% Ws reduction for OSS stage), because it is more efficient!
2. Understand the main causes of re-working for kits processed in orthopaedics:
Ad hoc process or tools to avoid re-working
More training for orthopaedics’ operators
The loop is something that can be avoided in the long term, to make the system more efficient.
3. Improve service process in orthopaedics: 2 kits in 25 minutes instead of 30 minutes. Ws is still about 16
minutes (16,1 minutes in front of 15,93 minutes), but it could move one operator to other areas.
4. Try to leave one resource and see what happens: for example, using only 1 M/M/2 in the OSS stage and see
how Ws changes.

REMEMBER: Good saturation is between 70 – 80%. Useful for benchmark.

GOAL OF QUEUEING THEORY:
The analysis let us to understand:
 lOMoARcPSD

• Where were the problems and the inefficiencies

• The possible impacts of changes in the system.

EX 3 – VERDI SPA
READING THE TEXT:
NB) Automatic stage is not considered as a stage of the system, a part from the case in which there is a testing
machine.
After assembly two there is a node with different options:
1. 0.05 * B are products damaged
2. Products slightly damaged in stage 3 are 0.15 * B
3. Good quality products are 0.8*B: they have non pre-emptive priority on slightly damaged products
All assembly stages are M/M/1.
Packaging is M/M/2, with input 500 p/h and working 300 p/h, FIFO rule.
The input of 500 p/h, which must be equal to the output is a constraint.

QUESTIONS:
1. You are required to model the system, highlighting all the important parameters.
2. Calculated Ws of the system (not considering scraps paths).
3. Comment the impact of using different priorities in stage 3.
4. Calculate average WIP before each stage of the assembly line.
5. How much is saturation in stage 2?

1. MAPPING
We have 3 variables:
Input
A = number of pieces that
pass through assembly 1
B = number of pieces that
pass through assembly 2

We need a system with 3 linear

equations to find the 3 variables.
We use nodes balancing (What
enters should be equal to what
exits).
NB) For the Input, we can also
consider the whole system as a
node.

2. THROUGHPUT TIME

Relevant parameters:
1 = A = 657.895
2 = B = 526.316
3 (p1) = 0.8*B = 421.05
3 (p2) = 0.15*B = 78.95
p = 500
 lOMoARcPSD

+ Ws CALCULATION: (best complete table we always need to build)

M/M/1: Ws = [1/ ( - )]*60

M/M/c: Ws = Lq/λ + 1/μ

PATHS IDENTIFICATION:
At the beginning we consider all the paths, also those with scraps. Here we have 3 with scraps and 4 without scraps.
1. Input Scrap
2. Input Ass 1 Ass 2 Good (mkt1) Packaging
3. Input Ass 1 Ass 2 Slightly damaged (low priority: mkt2) Packaging
4. Input Ass 1 Ass 1 Ass 2 Good (mkt1) Packaging
5. Input Ass 1 Ass 1 Ass 2 Slightly damaged (low priority: mkt2) Packaging
6. Input Ass 1 Ass 2 Scrap
7. Input Ass 1 Ass 1 Ass 2 Scrap

At the exam: we have to calculate occurrences for all the paths (also those with scraps). Then we use the occurrence
of only those without scraps, because scraps do no impact the throughput time!!!
The rule is to calculate the occurrences from left to right and every node must be considered very carefully.
For example, the loop encounters the main flow in the first node, but we consider it only when it divides from the
main flow, in the second node.

OCCURRENCES:
We can divide flows in 3 categories:
1. SCRAP: A1
A1 = (0.15*input)/input = 0.15

2. WITHOUT LOOP: A2, A3, A6

The quantity of pieces that moves directly from Assembly-1 to Assembly-2, avoiding the loop, is 85% of input
decreased by 20% of A. The population of pieces that follows the loop is different, isolated from the population
which does not enter in the loop.
(0.85*input – 0.2*A) = 394.7325
 lOMoARcPSD

A2: [(0.85*input – 0.2*A) * 0.8]/input = 0.51

A3: [(0.85*input – 0.2*A) * 0.15]/input = 0.096
A6: [(0.85*input – 0.2*A) * 0.05]/ input = 0.032

3. WITH LOOP: A4, A5, A7

To find the occurrences of A4, A5 and A7 paths it is necessary to consider only the items that entered in the loop of
assembly 1 stage. The quantity of these items is 0,2*A.

A4: ((0,2*A) *0,8)/input = 0,17

A5: ((0,2*A) *0,15)/input = 0,032
A7: ((0,2*A) *0,05)/input = 0,010

Resume:

Pieces/hour: numerator of occurrence

We consider only those paths without scraps and the adjusted

occurrence (occurrence adjusted with the ratio of
input/output). In calculating the adjusted occurrence of those
paths without scraps we divide directly with the output and
not the input.

Output = 500

THROUGHPUT TIME CALCULATION

EXPECTED THROUGHPUT TIME = ∑ 𝑾𝒔 ∗ 𝑨𝒅𝒋𝒖𝒔𝒕𝒆𝒅 𝑶𝒄𝒄𝒖𝒓𝒓𝒆𝒏𝒄𝒆

Average throughput time = 3.92 min

3. DIFFERENT PRIORITIES IN STAGE 3

????
 lOMoARcPSD

4. WIP BEFORE EACH STAGE

5. SATURATION IN STAGE 2
P0 = 1 - = 1 – (526.316/600) = 0.1228
0.1228 * 60 = 7.4 min/h

?????
 lOMoARcPSD

4/4/16
Mid-term exam
1 ex: queue
2 ex: lean

EX 4 – MECCANICA SPA
MODELLING OF THE SYSTEM

B’

C’

In general, when there are 2 machines not completely identical (in the flow, while they are temporally identical), for
example for Milling, it is important to write down 2 blocks. In general, when machines or servers are identical (both
in term of flows and time), it is enough to write down only 1 block with M/M/1 or M/M/c characteristic.

Starting from left:

There are 3 types of output for the milling machines: inputA, inputB and inputC
Drilling machines inputs are: drilling 1 inputA, drilling 2 inputB and drilling 3 inputC
Then there is the testing machine
There are 2 loops after testing machine: products B (0.2 inputB) and products C (0.1 inputC)
CHANGES IN THE INPUT OF D1 AND D2.
B’=1.2 inputB
C’=1.1 inputC
After the testing there are also scraps: 0.15*0.2*inputB and 0.15*0.1*inputC
At the end, there is: output A, output B and output C

NB) When there is a loop, goods are reworked only one time (no more than 2 times). This is a reasonable
assumption we can always make.

Now, we should do a map for each single product: A, B and C. It’s better to consider each single product separately,
because each of them has a different map. Then we should do the aggregated mapping, used to sum up everything.
We can do all calculations together starting directly from the aggregated map, but it is easier with the other
approach, more detailed.

CALCULATION OF UNKNOWN VARIABLES

 lOMoARcPSD

We make reference to nodes balancing.

PRODUCT A
There are no loops and no scraps.

PRODUCT B
There is a loop after the testing
machine, so we consider that node.

NB) 1B + 0.2B = 1.2B

PRODUCT C
It is the same as product B.

We can summarize everything in the following map:

 lOMoARcPSD

RELEVANT PARAMETERS

IDLE TIME=INACTIVE TIME

Few messages for system modelling:

1. Pay attention on how to choose variables.
2. Pay attention to the text point of view! In this case the focus of percentages is related to the input (it might
be related to other points, for example the number of pieces processed by a stage).
3. Node balancing as a control method: everything enters must exits somehow.
Remember:
1. Isolate flows and single products
2. Create the aggregated mapping with all the flows, to understand the arrival rates (which are the sum of all
the single arrival rates)

MAPPING PHASE IS FOUNDAMENTAL TO UNDERSTAND THE SYSTEM AND ITS INEFFICIENCIES

 lOMoARcPSD

EX 5 – MASTER SPA

M/M/2

We know that after the WC there are scraps (0.05B) and some products need to be reworked by A machines
(0.15B)
Testing machine works alfaA and alfaB products.
After testing, both A and B products have loops equal to 0.15*alfaA and 0.15*alfaB. From those products
which have been reworked, 0.5*0.15*alfaA and 0.5*0.15*alfaB are scraps.
We know outputA=300 and output=200

Questions:
1. Modelling and parameters
2. Expected LT without scraps.
3. Calculate the raw material input reduction if the defective products percentage at testing
machine drops to 0 (in the second test all the products are good). What are the effects on the expected
throughput time?

MODELLING OF THE SYSTEM

PRODUCT A
We know A, the output given, but
we do not know alfaA and Xa.
 lOMoARcPSD

PRODUCT B
We have 3 variables: alfaB, Xb and B.

ARRIVAL RATES

PARAMETERS

THROUGHPUT TIME

PRODUCT A
1. RC – finishing – drilling – testing – outside
 lOMoARcPSD

2. RC – finishing – drilling – testing –testing – outside

3. RC – finishing – drilling – testing –testing – scrap
PRODUCT B:
4. RC – WC – testing – outside
5. RC – WC – scrap
6. RC – WC – testing –testing – outside
7. RC – WC – testing –testing – scrap
8. RC – WC – finishing –drilling – testing – outside
9. RC – WC – finishing –drilling – testing – testing - outside
10. RC – WC – finishing –drilling – testing – testing - scrap

OCCURRENCES
Product A:
We have Xa until the loop. When there is a loop, we should subtract it from the flow:
1. (Xa – 0.15*alfaA)/Xa = 0.82
2. (0.15*alfaA – 0.5*0.15*alfaA)/Xa =
3. (0.5*0.15*alfaA)/Xa =
Product B:
We have Xb until after the WC.
4. (0.8*Xb – 0.15*alfaB*0.8/0.95)/Xb = 0.659

We should correct 0.15*alfaB because with this term we would subtact more than the necessary. We should
multiply by:
# B pieces coming from WC 0.8∗Xb 0.8∗Xb 0.8
= = 0.95∗Xb = 0.95
# B pieces coming from finishing/drilling B′′

B’’ = 0.15Xb + 0.8Xb = 0.95Xb

5. ECC

AVERAGE TP

Decrease of scraps, so less pieces at the beginning as input are needed THROUGHPUT TIME DECREASES
 lOMoARcPSD

6/4/16
INTRODUCTION TO SERVICES AND SERVICE PROCESSES
Why the service industry grew a lot recently?
- Social and demographic trend
o Evolution of needs
o Welfare: the global GDP is growing and people are becoming more and more rich, the spend of the
middle class is growing
o Entertainment
- ICT: it is an important enabler to deliver many services. Many services are based on data and on delivering
info to customers.
o Services to distribution and communication
- Globalisation: it has a strong impact
o Transport, tourism: in 24 h today we can reach the other side of the world. The world is getting
smaller, because the speed of travels and information is increasing a lot
o Services for industrial companies
- Outsourcing: why outsourcing means increasing of services?
o Services for industrial companies: bigger companies decided to create a logistic company, spin off
like Number1.

SATISYING DIFFERENT NEEDS - Maslow scale

Moving up means moving from physical products to services.

WHAT IS DIFFERENCE BETWEEN A PRODUCT AND A SERVICE?

Unique characteristics of services:
Intangibility: a product is a physical object, we can touch it, services no.
Creative advertising: advertise is how we visualise the product, but how to visualize the service?
Satellite TV is a service, we can use the decoder, but it is not enough.
No patent protection: it is much easier to copy services than products, but not all: for some of them
a technological competence is needed. For restaurants is easy, because it is difficult to protect pizza.
Sometimes there is a short time advantage.
Importance of reputation: when we buy a car it is easy to benchmark it, but for services the main
tool used is reputation, based on shared experience of previous clients, like TripAdvisor for tourism,
hotel, restaurants.
Mister // Alibi // Services
Customer participation in the service process: attention to facility design but opportunities for co-
production. Many times companies produce the product without interaction with the buyer/user of the
product. For services the influence and interaction of customers is really important.
A significant part of reputation for services is made of interactions with the customer, rather than the quality
of the service. For example, real estates directly develop strong interactions with their clients.
Simultaneity: opportunities for personal selling, interaction creates customer perceptions of quality.
A service is produced at the same time it is required by the customer. When demand goes up and down,
services should be produced according to this trend. For example, call centres have to face this important
 lOMoARcPSD

issue. Because it is very expensive, companies found a way to adjust it, like using technology to smooth and
level the demand (call centres use also chat services to reduce the nr of phone calls).
Perishability: cannot inventory, opportunity loss of idle capacity, need to match supply with demand
o If products produced are put in a stock, this means saving the production time for the future
demand, anticipating it. The stock is considered as the “stock of production time”, so companies can
produce products whenever they want. About services, if we use chats instead of phone calls, the
time spent directly with phone calls is used for chats, a way to decouple the operation and “stock
time”. Services cannot be put in inventories, only time.
o Seasonality in services, like transport and airplanes
Heterogeneity: customer participation in delivery process results in variability

SERVICE PROCESSES
The process is one key element to define a service delivery system. It can be classified according to 3 pillars:
1. Visible vs Not visible to the customer: the process to deliver the service is transparent to customers or
customers perceive only the result
1. Volume vs Variety: high volume and standard type (standardization) of services or high variety of services
(like doctor, any patient has a different need: low volume, high variety)
2. Processing: people, objects, information
Sometimes, designing a good service delivery process is much more important than the service itself. The risk is that
the final service can be a commodity, so the competitive advantage is based on how we deliver the service.
Processes have many differences that are the base for competitive advantage.

OPERATIONS SYSTEM CHARACTERISTICS

INTERACTION DEGREE WITH CUSTOMER
• Volume to handle vs Variety offered
• Variability (of demand, of capacity) and uncertainty

Front office: the delivery system is in touch with the customer. In a supermarket the cashier is a front office, with
a lot of work behind.
Back office

They are affected by operations. Operations affect also volumes and variety:
- High volumes need a very good operations management: like waiting at a restaurant means inefficiency
- Variety means that you need to have a wide set of competences and skills, and it affects a lot the delivery
We need to design two different kind of systems according to volumes and/or variety
- Variability of both demand and capacity (in services)
- Uncertainty

FRONT OFFICE focused operations

• the experience part is very strong: the experience of people sitting in the front
office is very important
• customer management is essential
• important outcome: it is a very important part of the value creation
Ex. Cinema, Entertainment park (Disneyland, Gardaland) are based on front office.
Typically, when the outcome is taken for granted and it’s not differentiated
(commodity), more attention is given to the experience, like for example, the
competence of the people in delivering a nice entertainment is fundamental to make
the difference.
 lOMoARcPSD

BACK OFF focused operations

• the experience part is less important: there are less interactions. The
experience in the front office does not exist, or it is very limited
• the outcome is essential: because it is the thing seen by customers.
Ex. Postal service

CENTRALIZATION
Decoupling Back-Office activities allows to centralize them.

Since many times we have both of them and due to the fact
that BO is not perceived/seen by customers, many times we
can centralize BO to improve the performance of BO,
maximising its efficiency and reducing costs (increasing the
margin). by red

EX) Banks: for them it is really important to be spread in the city, more offices there are, better is for the customer.
The BO could be centralized by reducing the BO operations performed locally, moving as much as possible of them
centrally somewhere: in this way banks can exploit economies of scale, reducing costs.

Going digital allows us to create central BOs, like a centralized call centre (when we have a centralized call centre,
we do not mind where it is placed).
11/4/16
Pros:
1. No interruptions (efficiency): in a centralized situation, creating the back up pays back, on the contrary not.
2. Fast decrease along the experience curve:
x) volume/nr of times we perform an activity
y) any parameter measuring the performance.
The curve normally goes down, so the time needed to perform a specific task is lower across time, because
we get used to it and we have more experience. If the volume produced is high, in the same interval of time,
the people in the BO perform many times the same activity, so they get more skills and become specialised.
This is another reason to centralise!
3. Specialization of activities (thanks to volume): more skilled people
EX) Post offices decided to reorganize the front office, splitting the desk in two main parts: some to deliver
traditional post services and others to deliver banks services.
4. Lower manpower cost: if you centralize you need less people than before (economies of scale)
5. Possibility to follow personal attitudes: not all employees are the same. If the system is not specialised
everybody has to do everything.
6. Less volume variability: if in each BO there is volume variability, the capacity is never saturated, because it is
designed according to the peak of volume. If we centralize, the peak of volume is the sum of the 3 peaks,
which is lower, so the capacity can be designed in a different way.
If we have 3 normal distributions, independent. If we sum them up, N1+N2+N3, the sum is like a normal
and the average is the sum of the averages, while the variance is the sum of the 3 variances and it is less
than 3 times the variance of each single variable.
7. Thanks to high volume:
Economies of scale
Automation, with advantages in terms of cost and quality
Cons:
1. Possible gap with the Front-Office: the Bo is far away, so communication (oral and physical) becomes more
difficult. This could have an impact on the performance of the BO.
2. Activities overlapping: the risk is to have not fully designed and synchronized activities, which probably are
performed 2 times.
3. Longer lead times: if we centralise a post office delivery system, the risk is to have a letter from village A to
village C, which are very close, that is send to the centralized distribution system and then send to the final
destination, which is not efficient. It works well looking at the overall performance, but the performance of
single cases it may not be efficient.
 lOMoARcPSD

4. Greater rigidity: we need to redesign the system which is not always simple and probably changing the
customer experience
5. Inability to keep a product base approach, therefore switching to an activity based approach

VOLUMES TO HANDLE vs VARIETY OFFERED

VOLUME-VARIETY MATRIX

H variety: it means that you do whatever your customers are asking for.
High flexibility
High level of customization
High contact time
High capability focus: the quality depends on variety, which is already high, so in this case it depends on the
capability of the person delivering the service.
Front office oriented.

L variety: we can standardise (low customization), we do not need people, so low contact time and product focus.
Back office oriented

SERVICE SECTOR:
1. H volume and L variety: MASSIVE SERVICES
The company gains money by increasing the productivity, because the price is given by the market (the
variety is very low). It is easier to find key performance indicators.
High volume of transactions for each server/unit
Standardised process
Short interaction with the customer (focus on back-office)
Attention to productivity
Attention to conformance
Automation / informative systems
Competences embedded in the system
Process innovation: fundamental to increase the productivity

2. H variety and low volume: PROFESSIONAL SERVICES

Reputation is very important here. Not standard or well defined process.
EX) Doctor, consultant, personal advisor, lawyer
• Low volume of transactions for server/unit
• Not well defined/standard process
• Longer interaction with the customer
• Attention to provide solutions
• Product innovation
• People’s competences and abilities are a critical asset
 lOMoARcPSD

3. PROFESSIONAL SERVICE SHOPS

Professional Service Shops are the development of the Professional Services as dimension increases.
Group of people delivering services. Any time the volume increases the shop becomes a large organization
which starts to specialize themself according to particular services. When the knowledge in a particular field
becomes very high, you cannot be expert in any topic.
Request of increasing efficiency without losing customization
Request of knowledge sharing (embedded in the system)
Development of a “house style”
Decreasing of the discretion degree (the quality becomes a standard)
Creation of semi-professional roles
Development of roles dedicated to interactions with the customer

4. MASS SERVICE SHOP

What do you do to increase margins? Move from a pure standardize situation, to some level of
customization. Mass Service Shops are the development of Mass services towards a broader service offer.
Increasing of mix offered (trying not to lose the strict control over the process)
Increasing of the front office discretion degree (trying to not worsen conformance)
The Front Office people need to develop the ability to understand customer’s needs (in order to
offer the right service)
EX) Book shops can differentiate themselves starting selling food or technology, changing the experience of
selling a book. They are still book shops, offering something more and different.

This is a continuum, it is impossible to cluster shops or companies in a strictly way like this.

What use can be done of this classification?

1. Verify inner coherence: coherence between levers
2. Verify external coherence: with the value offered on the market
3. Compare different Units / Businesses
4. Manage change
This method is used to evaluate if companies which deliver services know where they are positioned and where they
want to be and if they are coherent in designing the product/service with their values and the values of the market.

Could we have within the same company different solutions? YES. It could have different BUs, some of them
concentrated in a type and others in others.
EX) Lufthansa: Swiss and Brussels lines are own by Lufthansa and they are positioned lower than Lufthansa.

UNDERSTANDING VARIETY
To better understand the variety degree that the process needs to manage, it’s possible to divide service requests
as:
Runners: requests that always need the same operations/activities. Often foreseeable, and in remarkable
volumes; automation opportunity and process review needed. (like MTS, make to stock)
Repeaters: requests that refer to known activities, but clustered in a different way. Not so much foreseeable
and in medium/low volume. They are expected occurrences, but not frequent.
Activities that you have to do to satisfy the request are known, but you do not know the mix of request
for which activities are different. (like ATO, assemble to order, where you have all the modules ready and
known in advance, but the combination is not known until the customer order arrives).
Strangers: requests that need the design of new activities. Often little foreseeable. They are the exceptions:
NOT expected occurrences.
 lOMoARcPSD

Often there is a simultaneous factors presence. Where to focus:

VARIABILITY (of demand, of capacity) AND UNCERTAINTY

In 24 h the number of calls received can vary a lot.
Variability is the gap between the actual value and the average
value (peak and the minimum)
Uncertainty is the gap between the actual value (real value
experienced) and the expected value. When we are not able to
forecast the exact level/volume of service needed.

We can plan the variability and the uncertainty.

MANAGE VARIABILITY
Variability not explained causes uncertainty
Corollary: uncertainty can be reduced by explaining it, thus reducing it
to variability.

Variability conceals phenomena. There are 2 type of variability:

Determined by the company control/remove it
Outer variability limit the effect
Many times, variability is not only the result of external factors, but it can be the result of phenomena caused by the
company itself. This should be tackled properly to decrease the variability per sale.
20% of the products/services generates 80% of variability. If it’s not possible to remove variability, isolate it.

In a service system variability has to be managed at the front office.

The greater the variability, the greater the front office’s competences and discretion needed.
A system based on command and control is very inefficient (and often even ineffective) in managing variability.

EX) Weather condition: it affects the variability or the uncertainty of a service?

BOOOO
 lOMoARcPSD

18/4/16
MANAGING CAPACITY AND DEMAND

ALIGNING CAPACITY AND DEMAND

We can work on the demand and on the capacity side. Working together on both sides is called Yield Management.

Demand management: how we can adjust the demand to mitigate it.

Segment the demand: we can segment the demand, because maybe not all customers want the same things,
have the same needs.
Fix price incentives: we change the price through which we sell the service, suggesting customers a certain
price (discounts) asking them to use the service in a different time window (to move the demand).
Promote demand in low season: leveraging on seasonality to divert the demand (offering lower prices in low
seasons, like September for holidays)
Development of complementary services: coming in rush hours means no service, so people are providing
complementary services to move the demand to non peak hours. This is more expensive to be delivered, but
sometimes there are no necessary additional cost needed to deliver the service.
EX) Hotel with swimming pool and spa and golf. Peak season is very expensive, with extra money asked to
pay entrance for these services. In low season, the hotel offers these services for free to customers, with no
additional costs, because they are already fixed costs.
Development of a reservation system: it is a method to communicate with customers, to spread the demand
on a wider time window, using slots.
Capacity management:
Increase customer participation (in delivering a service): there are many examples, like in supermarkets
where there are less people working thanks the self paying system or the automatic system at the gas
station, where the customer is part of the service delivery.
Schedule shifts: for example, in Milan restaurants have two shifts to use the same table twice during the
same night by different group of people.
Share capacity: with competitors or with other companies working in the same market. For example,
airplanes share capacity moving in different areas, to distribute the capacity in peak and low seasons.
Employees cross-training: employees are trained to deliver more tasks, to avoid specialization which can
have problems about saturation (fluctuated demand)
Use of part time employees: in low hours companies use less number of employees.
 lOMoARcPSD

MANAGING DEMAND

DEMAND SEGMENTATION
You rarely serve only one single market segment. Customers’ needs and behaviours are often different. Proper data
analysis allows us to identify opportunities for segmentation, differentiating the service and thus improving
performances.

EX) Rental cars, different cars for different price and segments: how to understand how many cars? Analysis of the
demand of past years, if you don’t have the cars that you want there there is an upgrade, you can have type B car at
the same price of type A, without losing customer and no additional costs in offering a different segment. The cost of
the car is fixed cost and if you don’t use it, it is a loss and you lose the profit.

DEMAND MANAGEMENT AT HEALTH CLINIC

Doctors complain for the excessive workload. Patients complain for excessive waiting.
Average daily visits: 200
Variability: 150-240
Demand segmentation:

There are more visits at the beginning of the week, because people get sick during the weekend. Also people have
different behaviours according to different characters.
On Friday the number of visits goes up, because if during the week usually work a lot, at the end of the week they
cannot postpone anymore their visit to the doctor.

To reduce peaks and drops, we should try to segment the demand. Instead of making people going to the hospital,
they go to the reception declaring their disease and they get assigned different colours according to different waiting
times.

SET PRICE INCENTIVES

In periods of low demand, lower prices will be offered. This will move price sensitive customers to periods when
demand is lower, thus levelling the request.
EX) Train operators make ticket prices that are differentiated by day of the week and time, especially on Monday
morning and Friday evening.
EX) Flight tickets along a single day are different, like early in the morning or late in the evening

PROMOTE DEMAND IN LOW SEASON

Looking for other services to be offered.

DEVELOPMENT OF COMPLEMENTARY SERVICES

Complementary to do more and better businesses
Complementary for synergies/pairing
Complementary for seasonality

DEVELOPMENT OF A RESERVATION SYSTEM

With a reservation you almost have an early sale. The advantages are significant for customers (reserved seat,
welcome, ...). Reservation is perceived as an additional service, like the pre-booking online for cinema tickets, with
higher price than normal tickets.
 lOMoARcPSD

MANAGING CAPACITY

INCREASE CUSTOMER PARTICIPATION

When demand increases, also capacity increases: therefore, fewer resources are needed to deal with
increase in demand
The degree of customer participation can be varied according to demand
EX) If the supermarket is empty you would choose the employee, while with the supermarket full of customers you
would choose the self payment service.

MAKE A PART OF THE CAPACITY, VARIABLE

Flexibility solutions with outsourcing
First class - economy distribution: if business and economy capacity are fixed, problems may emerge.
EX) Booking a car at the airport, it can happen that the company has no more that car. What to do? Outsourcing.

CAPACITY SHARING
Some airlines make their own aircrafts available to other companies, at times of low demand (they also
change the fuselage decorations)
Multi property apartments
Sharing of resources between branches, also in term of knowledge sharing

EMPLOYEES CROSS-TRAINING
Multi task people: having some operators with extensive expertise allows to answer local peaks of demand for
specific activities.
Supermarket (eg. TESCO)
Bank
..

USE OF PART-TIME EMPLOYEES

If variability is predictable enough you can use part-time staff
In Italy, part-time is still used very little compared to other European countries, but it has great potential

SHIFTS SCHEDULING

It is a call centre. During night people sleep so there are almost no calls. How many people to have in different shifts
to saturate the capacity? We need to manage the capacity and the scheduling of employees for each day of the
week.
Second graph: we have defined the shifts in term of hours and workers per shift.

DAILY SHIFTS SCHEDULING

What is the approach to solve this problem? What are the info we need? We need to:
1. Demand forecast
2. Calculate the needs of operators (limit is on the level of service, and queuing theory)
3. Identify the types of possible shifts (national contracts, union agreements)
4. Shifts scheduling (eg. minimize Σ xi, xi number of people working in the shift i)
5. Assign people to shifts
 lOMoARcPSD

SCHEDULING WEEKLY SHIFTS WITH REST DAY

1. Define the type of shift to implement (eg. 5d work followed by 2 of rest)
2. Model the problem with Integer Linear Programming
3. Solve the ILP problem
4. Assign people to shifts

Objective function: minimize the overall number of people.

Constraint: having enough people to deliver the service.
EX) Sunday: equation which says that on Sunday we need at least 3 people working, because from the info of the
graph 3 in enough. For example, shift type 1, x1, is not working on Sunday, in fact it is not present in the equation.

Solution: we need 1 worker for shift type 1 (resting Sunday and Monday) and so on.

QUESTIONS:
How does the system change if you also introduce the following possibilities?
1. 1 shift with Sun and Thur as rest days and the others as working days
2. 1 shift with Sun as rest day and the others as working days
3. The solution will be better, worse, or the same as the previous case? Why?
 lOMoARcPSD

21/4/16
YIELD MAMAGEMENT
In 1978 the “Airline Deregulation Act” defined the price liberalization in the aerial field.
After that there was the uprising of new companies with lower prices: traditional companies were in danger due to
new competitors that offer very aggressive prices.
Aerial field widens, including new customers that before weren’t willing to pay high prices for transportation and
therefore choose trains or cars.
In this contest, American Airlines is bound to high fixed costs that make very difficult to reduce prices.
Competition started and companies started to ask questions about saturation problems: having a plane landing, the
fuel cost paid to move the plane is not affected (or only in a limited way) by the number of customers present on it
or not. So what is the goal? To sell all the seats at the max possible price. Companies cannot work on costs which are
fixed, but on turnover. The idea is to saturate seats’ capacity by offering some of the unoccupied seats at lower cost.
It was necessary to determine seats’ optimal allocation between economy and business classes in order to avoid the
cannibalization effect, i.e. having customers willing to pay high prices who buy low costs seats.
To avoid rate cannibalization: Service diversification (e.g.: rate with restrictions..)
The biggest difficulty was to allocate capacity to the different classes offered per flight, avoiding airplane’s under or
overestimation. Informative system and allocation model are fundamental to keep track and determine rates and
allocations
What could happen?
Normally low cost passengers tend to book in advance the seat and business people tend to book late, because
many times the book due to emergencies with no matter about costs. Low costs passengers are the first to be
served. The risk here:
1. Too many economic seats sold at the beginning: we are
potentially losing business people who would buy seats
at the last hours before the landing. We would lose their
WTP.
2. The opposite risk is keeping many business seats but at
the end no one wants to buy them, so we have empty
seats loosing the potentially economic buyers

Revenues increase > 1.5 bill. Dollars in 3 years from the introduction of this management system

Yield management applies to this kind of situation!

YM exploits the information on customers’ behaviour that Operations get while delivering the service, so to improve
competitiveness of the company.
YM acts in an integrated way on demand and capacity in order to “sell the right capacity, to the right customer, at
the right time at the right price” so to MAXIMISE PROFIT.
YM systems refer to strategies and tactics used by a certain number of companies to manage the allocation
of their capacity to different rate’s classes in order to maximize their profitability
Main objective is to maximize the capacity utilization rate in order to reach an objective in terms of
profitability that is as close as possible to the potential target of maximum profitability
Therefore, YM is a systematic approach to maximize profitability by offering prices differentiation to
potential clients. YM is an operations management concept applied to the theory of microeconomics; in fact,
we know that different customers have different WTP for different prices for the same service or slightly
different service.

APPLICATIONS:
Various successful implementations, in different fields:
Hertz (1-5% annual profit increase)
Royal Carribean Cruise Line (20 mil.$ increase per year)
Oriental Oberoi (asian hotel company, 230 mil. € profit increase)
Chevy Mexican Restaurants (5% annual profit increase)
World Cruise Line (USA and EU cruises, annual profit increase >180 mln€)
Application fields are constantly increasing.
 lOMoARcPSD

When the service delivery is based on fixed costs, companies adopt Yield Management, like:
Transport
• Airline companies, naval transportation, railway companies
• Rent a car companies
Entertainment
• Tour operators
• Cruise boat
• Golf course, Movie theatre, Advertising
Hotels, restaurants

IDEAL CHARACTERISTICS FOR YIELD MANAGEMENT

Fixed Capacity
Ability to Segment Markets
Same capacity for different Market segments
Perishable Inventory
Product Booked/Sold in Advance
Fluctuating Demand
Uncertain Demand
Low Marginal Sales Cost and High fixed costs
High Capacity Change Cost

POSSIBLE PROCESSES
We have different possible processes which need to be managed
differently:
1. BOOKING – PURCHASE TICKET (PAYMENT) – SERVICE
DELIVERY: pretty rare case, but it can happen sometimes
with flights (in an overbooking situation). You book your
service and you don’t pay the ticket immediately, but after
some time. After other time, you are served according to
the specific service. If during the service delivery the client
doesn’t show up the company has no problems because it
ha already collected money.
2. BOOKING – PURCHASE OF THE TICKET AND SERVICE
DELIVERY: you book your service, like theatres, events and service and you do not pay at the time of
booking, but at time you use or you benefit form the service. Pretty usual situation, like with Booking.com.
3. BOOKING AND PURCHASE OF THE TICKET – SERVICE DELIVERY: booking and purchase and then the service
delivery, like booking an hotel room with the credit card (once you give the credit card it is a cost for you).
EX) Rayan air, Easy Jet, Booking.com
4. PURCHASE AND SERVICE DELIVERY: Simplest case, like going to the cinema or to a restaurant. Directly
purchase of the service, no booking, and immediate delivery of the service.
(90% of the time we refer to the 3 option, the most traditional one)

YIELD MANAGEMENT TOOLS

There are 2 main strengths in YM (situations we will face in exercises):
1. Capacity allocation problem: for example, if we have a capacity of a cinema and we want to understand,
according to different categories of customers, how to organise the number of seats to reserve for which
type of customers and the booking system to max profit.
Price policies
• Demand forecast
• Protection policies (e.g. protection level): sometimes hotels say that they are full just to protect their
profits.
2. Overbooking: like airlines companies and hotels usually let people book more rooms/seats than the one that
are actually available, because they know that some customers will not show up to benefit from the service,
according to some statistics. This situation generates costs for the service provider who has to assist the
client not satisfied. Not doing overbooking means not having extra costs, but at the same time losing
 lOMoARcPSD

possible profits if some customers do not show up (loosing opportunity costs). Yield management means
understanding what is the optimal level of overbooking to max profits.

Example: we usually have to understand what are the different customers, the timing of their booking and purchase
and what kind of services to link to the single offer to exploit the potential of each single customer willing to pay a
given price. In this case we have a deterministic situation with many different group of customers with different
amount of money they are available to pay for the same service (WTP).
The number of seats available are 110 (capacity). We are not able to saturate all the seats with just one category of
customers. How to assign all these clients:
Not to loose most valuable customers
Not to book the seats just for the more valuable customer and leave places empty for cheapest clients

Linear programming approach:

Objective function with constraints: nr of ticket sold to a class should be
equal or less than the number of people in that class. The number of
tickets sold should be equal or less than the capacity.

We are sure that all the clients will show up and in general there are no
problems.

In general, it happens that:

X asix: time
Y axis: nr of sales
Capacity constraint
We consider for simplicity 2 categories:
Sales to low cost passengers: tickets are the one sold at the beginning of the booking or purchase process
(first step of the saturating process). It is natural
Sales to high cost passengers

CAPACITY ALLOCATION AND DEMAND FORECASTING

To implement YM we need two important things
1. IT systems
2. Data:
High detailed level (origin, destination, rates, day of the week, timetable)
Quantities to keep track of:
Demand per each categories/rate/period
Demand elasticity
What cost and solution should be implemented to satisfy detailed data collected.

Forecast difficulties:
Seasonality
Trend
Special events
Competitors actions
Small variations in a variable can have large effects on others
 lOMoARcPSD

PROTECTION LEVEL SIZING

Sequence of events we should consider for YM, in a
chronological term:
1. Demand for discounted seats: in a
discounted situation, we offer the same service with
the whole price and with the discounted price.
2. We can accept or reject the demand:
everyone will want to purchase the service with the
discounted price, so among them we should put a
limit to this number (not all can purchase with the
discounted price) in a temporal window.
3. Demand for full price seats

Protection level (S1): nr of full price capacity units

that I reserve for full price customers.

How to set the protection level? There are 2 approaches:

Marginal Analysis
Heuristic Expected Marginal Seat Revenue (EMSR): it is extracted from practice (experience of people) and
not from theory.

MARGINAL ANALYSIS APPROACH

We consider 3 variables:
X1=Demand for full price unit Demand for full price customers. The demand for discounted tickets can be
in general considered as much higher than the capacity that we reserve for that category, so we need to
study this demand in order to asses the real capacity given a certain full price.
o Since X1 is a stochastic variable there are some probabilities related to this variable.
Cu=cost to underestimate the demand of a full price unit Cost of underestimate the demand for full price
unit. After the forecast, I may discover that the demand for full price units is higher and not low as I though.
Co= cost to overestimate the demand of a full price unit

Demonstration:
The probability for underestimation [P(X1 ≥ S1)=0.999] in
this situation (extreme situation in which I don’t have a
capacity for the full price customer) is much higher than
the probability for overestimation [P(X1 < S1)=0.001]. It is
not a good situation.
What I have to do is to balance the problem. I have to
understand what is the limit in order to balance the two
probabilities and the weighted costs associated.
The costs are fixed so only probabilities can be changed. I try to put a different capacity for the full price
customers and I see that the probability gets lower (If I reserve 1,2,3,4,5 seats and so on). In this way I arrive
at the moment in which the two weighted costs of the two situations (over and under) are equal, we have to
define the right protection level S1.

EXAMPLE 1:
Situation:
It’s possible to classify clients into clients willing to pay full price and clients willing to pay discounted price
(simplification: only two categories).
Potential clients come from different sale channels and therefore the revenues are different for the Hotel.
The capacity is limited, so we are not able to accept all the request. On the other hand, we know that not all
who have booked will actually show up, with a given percentage between clients.
Goal: profit maximisation for the hotel you work for.
 lOMoARcPSD

Decisions:
Define how much capacity you want to protect
Define how much overbooking you want to have (because we have no show, on the contrary we won0t have
to consider overbooking)
Define a guideline to decide which requests you will accept and which one you will refuse (according to what
logic we are going to accept or reject our clients? At what level we want to stop, since the prices for each
category are fixed?)
For each request, you have to accept it all, or refuse it all

Walking cost: a client comes to the hotel and is rejected

because the hotel is full.

We calculate the under and over

costs: there is no mechanical
formula, because there too many
exceptions to consider each time.

Under cost: in theory if I underestimate, much more customers are available to buy the ticket. S1 of 50 and even 60
customers are available to buy the full price ticket. What I am losing in doing an underestimation mistake: the full
price that I would have sold to a customer, mitigated by the fact that I already sold a ticket at a discounted price. The
place will not be empty, but occupied by a discounted price ticket and not a full price ticket.
Over cost: I reserve much more seats than needed for full price customer, so I reserve 50 but I have 30, so 20 seats
are empty. The cost is the cost for having reserved the room for full price customers instead of having sell them to
discounted customers, so I lose the discounted price.

We start always from the cost and then we move to the protection level.

The probability is more or less 73%. We are in a normal distribution and so far
we have found the cumulative value of having x1 lover than S1.
For our S1 formula, we use the table for the standard normal distribution to
find the (fi) (also called z).

X1= average + *standard deviation

FEW MESSAGES
It would be possible to saturate the hotel selling all rooms at a reduced rate. In this way, however, the profit
would not be maximized. Should therefore bear a risk against a potentially superior margin.
The uncertainty and variability must be managed: overbooking and risk on the fixing of the protection level.
Having information on the past history is essential to set a yield management system, as well as having good
demand forecasts. The information system is central to the success of this management approach.
The ability to segment effectively the different types of customer and know in advance is a crucial point,
which the operations are to exploit (e.g. Customers of Japanese nationality usually do less no show than
customer of Italian nationality).
It's important to pay a lot of attention to the choice of prices.
 lOMoARcPSD

FEW ELEMENTS TO SEGMENT THE DEMAND

Some elements of segmentation to avoid cannibalization of rates, to differentiate the offer for customers:
Restrictions on the purchase: discounted categories
Delete option
Number of units: units that people can buy with one purchase
Purchase volume
Single purchase vs buying group: buying in groups maybe there are discounts, while buying alone you
could bear higher costs. Groups guarantee higher revenues for the service provider
Duration of use
Stay 1-day vs weekly stay: guarantee for higher revenues if people stay in the hotel more days
Customer "Value”

NB)
Protection level is used for all cases when you want to protect some capacity for more profitable customers which
will probably arrive later on.
Other applications:
Number of dresses of a new collection
Number of newspapers to buy to resell
Amount of bread to prepare

EXAMPLE 2:
Describe how you would determine the number of dresses to be ordered for next spring-summer collection,
whereas the expected demand is distributed as a normal distribution with mean 100 and standard deviation of 20,
and that the dresses ordered that are not sold at the end of the season, are all sold by lowering the selling price by
50% from 300 EUR / unit to 150 EUR / unit (the purchase cost is 160 euro per unit).
We are protecting the full price from the risk of finally be obliged to sell the dresses left with a lower price.

Cost of overestimating, the cost associated with having ordered an item in excess of those that were sold (cost
associated with having to sell at a discount the excess unit that I bought):
Overestimation cost: 160 € / unit - 150 € / unit = 10 € / unit cost of purchase mitigated by the lower price that
I am able to provide to sell it

Cost of underestimating, the cost associated with the non-sale of an item that I purchased for having
underestimated the demand (lost margin for stock-out of the article).
Underestimation cost: 300 € / unit - € 160 / unit = 140 € / unit lost selling (with the full price) mitigated by
the fact that I didn’t spend the money to buy the product (cost of purchase)
 lOMoARcPSD

OVERBOOKING

Overbooking can be used in two situations:

• Overbooking, due to the fact that not all booking become a sale.
• Overbooking of sold tickets over capacity (Oversaling), due to the fact that not all the ones who bought the
ticked then use the service.

The overbooking can be applied to 2 decision points:

• Overbooking on the number of
reservations: accept a number of reservations
higher than the service is able to provide
in order to protect themselves from the
reservations that do not turn into purchases
• Overbooking on the sales: selling a number
of units that is bigger than actual capacity in
order to protect from the effects of no show

The application of the types of

overbooking depend on the process followed
by each services company
In the following we will refer to process
type 3, but it is valid also for process type 2
and, if applied in sequence, also to process
type 1
STATIC METHOD:
Analysis on the mean values
Analysis with calculation sheet
Marginal analysis
Setting the level of service (not YM)
DYNAMIC OVERBOOKING: continuous balancing

AVERAGE VALUES APPROACH

OVERBBOKING BRUTE FORCE APPROACH

We know the nr of overbooking rooms and nr of
no show and we have the costs for both cases.
If we have 0 overbooking and 0 no show:
zero costs thanks to perfect balance.
And so on with 1 and 1, 2 and 2, etc.
If we have more overbooking than no
show, we have working cost
If we have more no shows than
overbooking, we have empty room costs
We calculate the costs for each solution and we
choose the cheapest one: 723€, with 3 rooms in
overbooking.
 lOMoARcPSD

Pros:
Easy to use
It considers opportunity costs
Cons:
It does not make easy to understand the underlying logic
Data on probability of no show are not reliable

OVERBOOKING WITH FIXED SERVICE LEVEL

You establish a probability of being in overbooking that
should be within a certain limit, L (that I define)

MARGINAL APPROACH
We do not establish a protection level, but the nr of overbooking to be taken into account according to the situation.
We continue to accept reservations until the expected margin on the last booking accepted is greater than or equal
to the expected loss on the last booking.
The model changes:

Ovb: overbooking
NS: no show

Same concept as before: we add the nr of rooms until

the balance. Solving the equation, we get a number
to define the nr of overbooking to protect yourself
from no show.

EXAMPLE:

The average is multiplied by the average no show 20%.

We did an assumption that the discounted category was distributed as a normal.

From here on we will just use the binomial distribution, easier. It will assign to each number of rooms a given
probability.

OR protection level approach OR overbooking approach.

 lOMoARcPSD

In reality, overbooking costs are not linear:

With small overbooking we are able to solve problems in a
quick way, but as far as the nr of customers in overbooking
increases, the problems increase in an exponential way.

DIFFICULTY IN THE IMPLEMENTATION OF YM SYSTEMS

Customers’ reaction
Difficulty to understanding
Attempts to fool the system
Conflicts of objectives among different business areas
Cost / Time Implementation
Cost of the information system
Difficulties/time/cost in data collection

YM IMPACT
Examples of improvements achieved: lot of savings
• American Airlines (1.4 bln.$ increas in turnover in a period of 3 year)
• Hertz (turnover increase 1-5%)
• Marriot Hotels (additional revenues 25 mln.$ in one year)
• Royal Caribbean Cruise Line (20 mln.$ additional revenues in one year)
Sectors adopting (and customising) these approaches are fast increasing. E.g. golf court

HEURISTIC METHOD (EMSR: expected marginal seat revenue)

We have 3 or more categories of prices, where the marginal analysis doesn’t work.

Fix the level of protection for each of the classes / rates:

 lOMoARcPSD

EXAMPLE:
Determine the protection level of classes 1 and 2.

Fi=100*30 + 80*30 / 30+30+50

 lOMoARcPSD

FUTURA SPA
Questions:
1) Calculate the protection level for full price tickets
2) What happens to the protection level if company introduces last minute tickets at a price of 150 €?
Data shows that usually no-show related to people that have purchased tickets at discounted price is equal
to 10% (the ticket is not refundable)

Resume of the costs and timeline: first thing to be done.

We always need to specify to which kind of price costs are
related to and if the company sustain these costs before or
after the event (preventive or final cost).
In a chronological perspective, the protection level (PL
DECISION) should be defined before all the events and costs
of the system, when we are not sure about anything.
Then there is the EVENT, which divides the time in PRE-EVENT
and POST-EVENT: in this case all costs are sustained by the
company before the event.
Sometimes we could have post event costs like the cost of the
buffet for only those who participate to the event.

Buffet: FULL, PRE. We pay without knowing the effective number of full price tickets (we will know the
number only the day of the event). The actual and effective demand for full price ticket is important.
Flyer: both FULL and DISCOUNTED, PRE. It is provided by another external company, so it not a cost
sustained by Futura, we do not consider it.
Program: both FULL and DISCOUNTED, PRE.

Marginal analysis
X1: demand for full price tickets.

1) Protection level definition

Co = company paid for h.q. program and buffet for a seat that remains empty. Moreover, it lost the opportunity of
selling that place at a discounted price (with related costs).
The day of the event I discover that there are some empty places. Their cost is the overestimation cost, that we
should suppose in advance, before the event. Overestimation cost is the lost related to the income that would have
come from of the selling with the discounted price (320) + pre costs related to full price tickets already sustained by
Futura (costs are positive and savings negative) (15+40) – costs not sustained for discounted tickets not sold (25).

Co = 320 + 15 + 40 – 25 = 350

Cu = company lost the opportunity to sell a ticket at full price (not sustaining also related service costs) but gained
the rate of discounted ticket (and paid for its service costs)
The day of the event the seats are full and I think that I could have sold more full price tickets, but I didn’t,
because I sold discounted tickets. Underestimation cost is the lost related to the income that would have come from
the selling with the full price mitigated by the selling of discounted tickets (1000-320) + costs not sustained for the
full tickets that I would have sold (-15-40).

Cu =1000 – 320 + 25 – 15 – 40 = 650

 lOMoARcPSD

First of all we need to calculate the standard deviation of the distribution,

, using the formula.
This is needed for the protection level formula.
Then we calculate the value of Z for the formula of the protection level. We always round up the value of the
protection level. At the end we can define also the number of tickets that we sell with the discounted price.

2) No-show phenomenon of discounted tickets (10% of probability that who bought discounted ticket doesn’t turn-
up). Introduction of last-minute ticket, price = 150 €.

We could have empty seats due to:

- Overestimation
- No show
Tickets are not refundable under any circumstance
Last minute tickets: no further service offered than the ticket itself (any service, any cost)
The day of the event I find empty seats and I sell these seats at a very low price to some last minute participants.
Last minute tickets apply for both empty seats and no-show phenomena!!!!!

NO SHOW CAN BE ON BOTH CATEGORIES! In this case is just for discounted price category.

Co = company paid for h.q. program and buffet for a seat that remains empty. Moreover, it lost the opportunity of
selling that place at a discounted price (with related costs) but sold it at last-minute price.
In an overestimation situation, with empty seats, we should subtract the income that would have come from of
the selling with the last minute price (last minute ticket would have all been sold because of the very low price, like
discounted one) (150). I could have sold also the empty seats to discounted participants and because of the fact that
no show is only for the discounted one (10% discounted), I could have sold last minute tickets also to the discounted
participants that do no show (10%*150).
Empty:
- Mistake of overestimation: last minute
- One of the discounted do no show with 10% of probability: last minute to them (10% last minute)

Co = 320 + 15 + 40 – 25 - 150 + 0,10*150 = 215

 lOMoARcPSD

Cu = company lost the opportunity to sell a ticket at full price (without sustain related service costs) but gained the
rate of discounted ticket (and paid for its service costs)

Cu = 1000 – 320 + 25 – 15 – 40 - 0,10*150 = 635

In this way the protection level for full price tickets is higher, so better for
the company!

Costs that are not differential for PL decision are:

• Cost paid by an external source
• Fixed costs
• Costs to be paid in advance (PRE) for both (together) Full and Discounted fares: for example a buffet to be paid
for all tickets.
These costs will never appear in Co and Cu!!!!!

EXERCISE 2

If the probability is lower than 50% (like 0.485) we find the value on the table (1-0.485=0.515 0.04) and we put a
minus before it.

EXERCISE 3
We immediately eliminate the fixed cost and the cost for
both prices. If the same object has high quality for full price
(5 euro) and low quality for discounted price (3 euro) we
need to consider this cost, because we have different prices
for different categories of price.

1)
Co = opportunity cost to sell a discounted ticket
We do not consider the cost for full price tickets because it is POST the event and due to empty seats it is not
sustained by the company
Co = 200
 lOMoARcPSD

Cu = company lost the opportunity to sell a ticket at full price (400) (without sustaining related service costs) but
gained the rate of discounted ticket (200)
We have saving according to the costs for the full price tickets that we do not sustain because they are after the
event (12). Only the 60% of the full price tickets participate to the buffet for which the company has to pay. Only the
(100-10=90%) of the full price tickets will show (10% of no show for full price category)

Cu = 400 – 200 - 0,6*0,9*12 = 193,52

S1= FULL PRICE

2) Probability perspective, estimation of the future. The profit related to our system will not be deterministic and we
consider it related to the normal distribution of the demand of the tickets.
Some costs will be deterministic.

Profit Revenue – Costs = Profit

- Rent of arena = 50000 €
- Profit target = 45000 €

Revenue = X*400 + (500 – 298)*200 = 400*X + 40400

400 full price multiplied by X (effective number of full price customers, we calculated only the protection level until
now)
Costs = 500*3 +500*6 + 0,6*0,9*12*X +3000+50000 = 6,48*X + 57500

158 is the data that I have to put in the usual formula for the standard distribution, to have at least 158 full price
tickets.

3) Last minute ticket + no show for both categories

Introduction of last-minute ticket, price = 175 €

Co = opportunity cost to sell a discounted ticket (and additional last minute revenue in 10% of cases) but at the same
time, company sells a last minute ticket
Co = 200 - 175 + 0,1*175 = 42,5
 lOMoARcPSD

Cu = company lost the opportunity to sell a ticket at full price (without sustaining related service costs) but gained
the rate of discounted ticket (for both tickets there is an additional last minute revenue from no-show phenomenon)
The situation does not change because we have no show for both of them

Cu = 400 – 200 - 0,6*0,9*12 + 0,1*175 – 0,1*175 = 193,52

Costs are the same, but the revenues can increase thanks to last minute tickets.
0.1*[X+(500-410)*175+ (410-X)*175
410-X: empty seats for full price which can probably buy the last minute ticket
X+(500-410): number of effectively presents customers

EX 1
The difficulty is that:
discounted ticket: the services (700€ + 60€) are included in the ticket
full price ticket (170,000): the services are not included, but can be sold if requested (70% of full price
customers will ask for these services).
Other option: remaining cars are sold to another channel with fixed costs equal to 30,000€.
COSTS (not all important): many of them are fixed costs, not to be considered

London Motor Show

participation (fixed fee to 10.000 €
be paid to organizers to
participate)

Stand preparation
50.000 €

15.000 €
Staff (hostess,stewart)

Precious box of brier-root 100€/box

60 €/ticket 1. Protection level

Opening “Le mans series”
 lOMoARcPSD

ticket
COSTS:
Access to "Brands Hatch" - Access to "Brands Hatch" circuit =
700 €/access
circuit 700 €/access DISCOUNTED, POST
- Opening «Le Mans series» ticket = 60
Advertising banner (not €/ticket DISCOUNTED, PRE
200 €/banner
related to tickets) - Precious box of brier-root = 100
€/box DISCOUNTED, FULL, PRE
Registration “Le Mans - Advertising banner = 200 €/banner
20.000 €
Series” championship DISCOUNTED, FULL, PRE
- Production cost = 70,000 €/car
Gadget e merchandising DISCOUNTED, FULL, PRE
5.000 €
for events - Cost unsold car = 30000 €/car
- Margin access to “Brands Hatch"
circuit = (900-700) €/access
- Margin opening «Le Mans series» ticket = (100-60) €/ticket

Not all fixed costs are present in the list.

Discounted and full PRE with the same amount of money are not considered in Co and Cu: precious box,
advertising and production cost (yellow ones). Only timing and services change in selling a product that is
the same for all customers.
We consider only the first two costs, only for discounted: because they are already included in the costs of
the ticket for discounted and their profit is already included in the price of the ticket.
70% of full price customers will buy these tickets: the price 900 for customers should be diminished by the
cost to be paid for the service 700. (900-700) + (100-60) are margins.
The fixed cost of 30,000 is a cost sustained in a case of overestimation, when there are unsold cars.

Co = cost related to have dedicated too many cars to after London exhibition sales, company wasn’t able to sell all
cars and has to pay additional costs to sell them through other channels.
DISCOUNTED POST ARE TREATED AS DISCOUNTED PRE IN OVERESTIMATION

Co=150.000–60–700+30.000–170.000–0,7*[(100-60) +(900-700)] = 9.072 €

Cu = cost related to have allocated too cars to discounted rate, so company loses the opportunity to sell cars at a
higher price after the London exhibition.

Cu = 170.000 + 0,7*[(100-60) +(900–700)] + 60 + 700 – 150.000 = 20.928 €

2. After-exhibition demand is described by a normal distribution with mean 150 and probability to sell a
number of cars between 110 and 190 equals to 76%.

76% + LEFT PART: (1-0.76)/2 (BECAUSE IT IS SYMMETRIC) + 0.76 = 0.88

 lOMoARcPSD

ASTORIA HOTEL

1. How many reservations should Villa Reale hotel accept?

Cu a room remains empty and I could have booked it to another further customer. Underestimation of
overbooking.
Cu = 145 € - 15€ - 3€ - 5€ = 122€

Co company has to assure a room to the customer that booked a room. Overestimation of overbooking.
60% same chain
40% another hotel

Co = 0,6*(15€+70€+25€) + 0,4*(20€+70€+120€) – 145€ = 5 €

2. Adding costs:

Cu a room remains empty

Cu = 145 € - 15€ - 3€ - 5€ = 122€

Co company has to assure a room to the customer that booked a room

Co = 0,6*(15€+70€+25€) + 0,4*(20€+70€+120€) – 145€ + 250€ = 255 €

 lOMoARcPSD

NB)
Lower number of rooms with further costs, worst consequences.
 lOMoARcPSD

09/05/2016
LEAN PRINCIPLES
Lean manufacturing, services and supply chain is a method becoming popular and applied to every process to
improve its performances.
Lean: design a new product is made in lean principle. Not only with IT but in order to reduce the time of the product
development.
How to make money? Increasing sales or reducing the costs, in order increase profitability (doing both also):
- Increasing the sales: it is not always necessary and it means that
o if the market is not growing and it is mature, the quantity sold to the market is given, so you have to
fight against the competitors and it is not too easy
o if the market is growing, it is easier to increase the sales (when the market is booming everything is
easy)
- Reducing the cost does not depend on the other players, but it is a pure internal issue, it’s up to you to have
knowledge, competences and resources to reduce the costs. How to reduce the costs?
o We could reduce the features of the products: no value if a customer is able to willing to pay more,
even if you have a good idea but anyone will invest in it there is no value and it means that there is
no possibility to get profit
o We could reduce the material used: leveraging on sustainability (lighter and smarter products). Less
material you need, less machining time to produce the product
o We could reduce the processing time: with two sources: if you have less material you need less
processing time, the other is the result of lean manufacturing (find new ways to do the same tasks
you have to do but in a more efficient way)

NEW VISION
We have to be able to recognize those activities that are valuable and
produce value for the customers, while there are some activities that
are not valuable for the customers, called waste (money, time,
resources, material, energy), we have to analyse and identify the waste
and eliminate them (a way to apply lean principles).

TYPES OF ACTIVITIES
There are three types of activities:
Value adding activities are those ones that the customer demands, so valuable according to the customers’
perspective. You have to ask: What is the customer willingness to pay? This is the only criteria to define the
value, not using your point of view
Not value adding activities, which are those ones that the customer does not demand and are unavoidable
in the short term (we could work in removing them but we need time, effort and money)
Not value adding activities that are avoidable in the short term (we have to work on them because they can
easily be removed in the short term)

Examples of waste (which are a cost for the company, but I can reduce or change the way to do them because I don’t
add a value for the customers). Sample activities for a Machine tools producer:
- Insert a ball bearing on its shaft add value
- Test the product before sending it to the customer: in order to be sure that everything is produced well, so I
do it if I’m not sure that what I did is correct, but the customer does not pay for the test, so it is a waste
(today I want to make the process reliable 100 % in order to do not need to have the test at the end of the
process, changing the approach) no add value
- Fix a rough material to the fixture of a vertical workstation add value
- Bring the electric motors from the warehouse to the place where the machine tool is assembled no value
for the customer

1
 lOMoARcPSD

- Urge the supplier for an overdue order no add value

- Plan internal and external personnel’s’ activities for the next week no value for customer. Logistic means
pay people to move product; it is a cost for me not a value for the customer.

WASTES CLASSIFICATION
We can identify 7 types of wastes:
Transportation: moving product, internal and external transportation
Inventory: it is a waste because it is a capital cost, you need a warehouse to stock the inventories, with
space wasted and people to manage it, how can you reduce it? I have to be able to produce the products
whenever I need it
Movement: some movements around the machine during the processes (like assembly)
Waiting: product waiting in the queue, people waiting to do something, with a long queue customers are
not willing to pay for it, you need a space to “store” the objects, which means costs
o “Waiting is the easiest to generate and the hardest to correct, because time does not pile up on the
floor like material wastes” (H. Ford)
Overproduction: you produce more than what you need
Over processing: working or re-working product twice, more than once
Defects: problem in quality, throwing away the products, the energy to produce the product is gone, the
money is gone

UNDERSTANDING THE
PHILOSOPHY

At the base we have STANDARDIZED WORK, that means trying to design processes and activities and trying to have
them standardized. Everything is decided before, so you have to solve the problem and provide a solution to all the
company. Then there is also KAIZEN, that is continuous improvement, which means never stop to looking for new
solutions, if you stop doing it is not lean manufacturing, there is always something better to do.
To develop we use two main tools, pillars:
Just in time (JIT): producing only the quantity that you need when you need (hopefully the right one), doing
the right components when needed and in the right quantity (no inventories).
o To implement it we use continuous flow (instead of producing huge quantities we make the machine
working all day long in a continuous flow), pull system (let the market decide the flow and the
quantity without pushing them), takt time (which is no the cycle time)
Jidoka: looking at the quality
The results are obtained in achieving the best quality, with the lowest costs, in the shortest lead time.

Lean manufacturing wants to reduce all the activities that do not add values and eliminate the wastes (called
in Japanese MUDA), and you have to identify the problems, the weaknesses, the defects and the
inefficiencies in order to solve them (If you hide them you cannot solve because the problem still remains.
Eliminate the cause of the problem, rather than avoiding that it gives troubles).

2
 lOMoARcPSD

THE SMALL BOAT AND THE CLIFFS

When you have problems and inefficiencies (the cliffs in the image), the possible solutions are:
The inventory is a way to hide all of them: with inventories you are always able to send the products to the
customers (you need all the blue space), but it means a cost for the company

Reducing the stocks in order to arise the problems and solve them: if you force your employee they are
initially disappointed, then they learn and find a way to solve the problem and produce the products, in a
long-term way to find the solution

It is an endless process, without stopping it (the focus is on the process, not the result).

From Toyota document (result of implementation of this approach in a company, number one for using this
approach):
“We achieve brilliant results from people with average capacity that operate and improve brilliant processes
(Lean approach). Our competitors achieve average results from brilliant people that work on flawed process
(Western approach). When they meet some problem, they try to hire more brilliant people. We will beat
them”
This is a though statement against your employees
Do not look for the result to overcome the product, but look to the process (simple and efficient)

LEAN PRINCIPLES
There are 4 lean principles (flow, pace, pull, zero defects) and 5 steps in applying them:
1. Identify the value, that is defined by the customer in term of all that the customer is willing to pay.
There is a huge change of approach, one of the major changes of the last 30 years: the shift to an external
point of view. From
cost + mark-up = price (accounting approach)
to
price (value) – mark-up = cost (firstly understand the value for the customer)
2. Map the value stream, which means understanding where the value is created (most of the companies does
not know). The value stream is the set of activities that are made on the inputs to give to the customer a
product or service, with three main flows:
Development and introduction of new products and services
Managing information from customer’s order to delivery (information has a value)
Manage product or service realization from the inputs elements to the final output
(Identify the activities and the resources involved)
3. Create flow
Anything that stops in the flow is a waste (even if it is a small event)
You have to identify what stops the flow and remove it

3
 lOMoARcPSD

4. Establish pull
Pull means that we have the market (which is our boss, customers…) and the whole production
systems and all of the activities are pulled by the end and not pushed from the beginning
Forecasts are always wrong by definition (make it in advance means that maybe you produce more
or less than what is needed, you need it for designing the system not for managing the system)
Many of the systems we develop are too complex to be completely controlled (FMS, MRP, ecc.)
Instead of trying to make a better forecast, we should focus on having a better response capacity
5. Seek perfection
o When quality is sufficient? When cost is low enough? When wastes are low enough?
o Looking for the best quality, no mistakes at all (if perfection is the aim, we focus on the journey and
not on the destination)

POSSIBLE RESULTS
If you implement the lean you can achieve incredible results, all with a very limited capital investment:
• 50% improvement of work productivity in the whole system
• 90% cut of production lead time
• 90% reduction of stocks
• Halved process defects and wastes
• Halved accidents
• Halved Time to Market

- Low costs, low resources needed

- With traditional approach, there is a little improvement but a great effort and stress (same NVA, low VA),
while with lean approach you maintain the same VA reducing NVA (great improvement, little effort and
stress), working better.
EX) Supermarket on Saturday afternoon: you spend 2 hours without adding value apart from retrieve the
product from the shelves on your bag (10-15 %), waiting in all the queues that are formed is not adding
value activities
Why is it not already done?
• Mainly for a cultural aspect
• Because 50 years ago it wasn’t perceived so urgent
• Because it’s necessary to completely change the organization and human resources managerial vision
• Because the heart is the process and not the final product/service
This can be an advantage for the service sector, where there is not a tangible output, it’s not possible to
verify before the delivery to the customer and focusing on the process is a must.

LEGGI ESEMPI SLIDE 25-35

4
 lOMoARcPSD

11/05/2016
VALUE STREAM MAPPING
All we are really trying to do in lean manufacturing is to get one process to make only what the next process needs
when it needs it. We are trying to link all processes -- from the final consumer back to raw material -- in a smooth
flow without detours that generates the shortest lead time, highest quality, and lowest cost.

Value stream mapping is a pencil and paper tool that helps you to see and understand the flow of material and
information as a product makes its way through the value stream. What we mean by value stream mapping is
simple: follow a product’s production path from customer to supplier, and carefully draw a visual representation of
every process in the material and information flow. Then ask a set of key questions and draw a “future state” map of
how value should flow.
Doing this over and over is the simplest way - and the best way we know - to see value and, especially, the sources of
waste.

Just remember that the point of getting lean is not “mapping,” which is just a technique.
Mapping helps you see and focus on flow with a vision of an ideal, or at least improved,
state. The system can be:
As is: as it is now today the system
To be: how will be the system in order to be better performed in the future

The first step is drawing the current state, which is done by gathering information on the
shop floor. This provides the information you need to develop a future state. Notice that
the arrows between current and future state go both ways, indicating that development of
the current and future states are overlapping efforts. Future-state ideas will come up as you
are mapping the current state. Likewise, drawing your future state will often point out
important current status information you have overlooked.

We can analyse a manufacturing system with different levels of analysis

(details):
Process level: directly in the machine, manufacturing process for
instance, and we could improve the process and the performances
(shorter set up time, faster machines). We need technicians and
mechanical engineers who are directly connected to this knowledge
Single plant: a single plant is composed of different machines.
Originally, the value is created on the door to door, so the value
stream usually begins at plant level
Multiple plants: supply chain of the company
o Sometimes the value stream can be created in this level of
analysis, the whole supply chain of the company
Across companies

The lean approach is an approach which is based on a number of small improvements, that are somehow a good
sense (we do something we think that is good for the improvement of the company), so it is not a mathematical
approach (there is not a multi-objectives model based on lean manufacturing). We just want to identify the problem
and solve and reduce its impact, in a flexible and continuous way (step-by-step approach).
This means running a solution in a spiral approach: we do something, see if it works, doing something else and
continue, according to multiple recursive steps. You do not do all the decisions only once, you go straight away
taking decisions every moment, in a continuous improvement approach.
It is different from the waterfall approach, that goes from the top to the bottom, only in one way/direction, with no
second chances.

First step: identify a family

Value stream mapping means walking and drawing the processing steps (material and information) for one product
family from door to door in your plant. Your customers care about their specific products, not all your products so

1
 lOMoARcPSD

you will not be mapping everything that goes through the shop floor. Unless you have a small, one-product plant,
drawing all your product flows on one map is too complicated.

A family is a group of products that pass through similar processing steps and over common equipment in your
downstream processes (the most important ones, according to ABC curve concept: the majority of my profit is done
by only few items in my portfolio). Write down clearly what your selected product family is, how many different
finished part numbers there are in the family, how much is wanted by the customer, and how often.

Lean manufacturing is made by many simple tools, because normal (average) people have to use them.
This is a simple tool that considers steps for assembling the products and different products. The matrix matches the
two dimensions. It is also a visual tool, used for ordering and at the first round of the spiral approach we start
eliminating the products that we do not need (like E, F, G), focusing our attention on A, B, C products.

Equipment, technologies and competences needed are located on the

columns, while products represent the rows. In this sense, X indicates
that a product requires this specific equipment.
We can observe how A, B, C can be grouped in the same family.
Obviously, the overlapping in not total, because B and C require the
addition of one resource. However, the more products are similar, the
more they will require the same productive processes.
We can also observe how at the beginning there may be a non-perfect
match between products of the same family, but when we redesign the
product/service or the process, then the matching can be more precise.

Second step: focus on main streams only

We must keep in mind that it is not required to map all the possible process, but to
focus on main stream only: in fact, a map should have as few elements as possible; all
the rest must not be there in order to simplify.
The longest line, which represents the main flow.

The main KPI is time, which can be declined in:

1. Cycle time: it is the interval time between one exit from a line and the following one
o The slowest station (the bottom-neck) gives the cycle time to the whole system
(it does not matter where the bottom-neck is, it is the same if it is at the
beginning or in the end, you need to improve the performance of the worst)

Is there any reason why the first 8 stations should work faster if the 9th
station works slower? NO. If the last station is the bottom-neck, all the other
stations should work as the last one to avoid WIP and other consequences. The
conclusion is the fact that the position of the bottom-neck does not matter!

o We can improve (reduce) the cycle time removing the bottom-neck or reducing
the time of working of it (balancing the working time between different
stations), in order to have a higher production rate

2. Value added time: real time that each station works on the product, time of loading for
each station (it has to be smaller than the cycle time by definition, if it is higher it means
that it is the new cycle time, the bottom-neck)

3. Lead time: it is the time spent by the product from the beginning of the process to the end of the process.
If the cycle time is 10 seconds, what is the lead time (you have 8 stations)? 80s.
o There are two types of line: one where we have a buffer between one station and the other, and
another one where we do not have buffer.

2
 lOMoARcPSD

When we have a buffer upstream and downstream, the movement between different stations takes
time (maybe 1 seconds) and the time to take the movement is not dealing with the activity of the
workers in the stations (9s + 1s = 10s).
We need to see if there is buffer, and we add the movement time!!
If the product stays in the buffer for 2 hours, the waiting time should be included in the lead
time!!
(NB) In general: VA < C/T < LT

DRAWING THE CURRENT STATE MAP

Developing a future state begins with an analysis of the current production situation, as we will do using a simple
example factory we call Acme Stamping. Mapping begins at the level of the door-to-door flow in your plant, where
you draw process categories like “assembly” or “welding,” instead of recording each processing step.
In particular, the company wants to recur to the VSM in order to evaluate the structure of its production
system, identifying the potential critical points of the system and implementing further developments.

DESIGN OF THE VSM

As one of principles of Lean management is pull, the starting point for the design of the VSM is the customer (the last
action), adopting a customer point of view and having a whole system from the market.
Mapping starts with the customer requirements. We’ll represent the customer’s
assembly plant with a factory icon, placed in the upper right-hand portion of the map.
Underneath this icon we’ll draw a data box recording the requirements of Acme’s
assembly plant customer.
o We need to fulfil the demand and know how many pieces in a time interval
(pieces/hours, pieces/day, pieces/week, pieces/month, pieces/year) we need to
produce.
There are 2 different components (L and R), two families of products
The next mapping step is to draw the basic production processes. To indicate a process,
we use a process box. The general rule of thumb for a door-to-door map is that a process
box indicates a process in which the material is flowing and it stops wherever processes
are disconnected and the material flow stops.
Material flow is drawn from left to right on the bottom half of the map in the order of
processing steps; not according to the physical layout of the plant. At Acme Stamping we find
six processes in the steering bracket material flow, which occur in the following order (with
their relative number of workers):
1. Stamping, where in addition to the operator, includes a press of 200 tons. It is important to
underline that, unlike the other stages where the resource is dedicated; in this case, the press has to
realize more than two products of the products mix.
2. Spot-Welding Workstation 1
3. Spot-Welding Workstation 2
4. Assembly Workstation 1
5. Assembly Workstation 2
6. Shipping

3
 lOMoARcPSD

As you walk this flow on the shop floor, you need to collect data that is important for deciding what the
future state will be. So a data box is drawn under each process box. At Acme Stamping, we have the
following information to record in the data box under each processing step:
o Current cycle time (C/T) because we are looking at the current situation (time that elapses between
one part coming off the process to the next part coming off, in seconds). the operator process a
piece for X seconds;
o Change-over time (C/O), the set-up time, so the time to switch from producing one product type to
another (in this case switching between L and R products). But also the time needed to position
products in the right way in the machines;
o Uptime, the % of the time in which the machine works over the total time that it has.
Availability is a % over a long time of observation (we look at the system that can be repaid
by changing the component and not the component), so it is the % that the system was
properly working out of the whole time of observation (if I have a car and the availability is
98 % it means that 2 % of the time I use it, it does not work)
Reliability is the probability that a component (a non reparable component which is thrown
away once it goes down) is working after X time (it is a function of the time. It is a function
that goes down, but the shape depends on the system and the technology).
The relationship between availability and reliability can be described by:
If we consider the mean of the uptime is the mean time to failure (MTTF), which is the time
between one fail and another one.
R(t) function of time. The graph depends on the component that have a continuous value
(ex. electronic components, you switch them and they broke, others work for years).
The mean (MTTF) is the area divided by the two extreme (integral between 0 and t),
𝑇 𝑅(𝑇)
calculated as: ∫0 𝑇−0 𝑑𝑡
Any system which can be repaired has up-time and down-time (when the machine does not
work, you have to take time to understand the problem, buy new components).
The length of the down time can vary depending on the time of the repair, so we can write a
similar function R(t) and calculate the mean which is the mean time to repair (MTTR,
average time to have a reparation)

𝑀𝑇𝑇𝐹
𝐴𝑣𝑎𝑖𝑙𝑎𝑏𝑖𝑙𝑖𝑡𝑦 =
𝑀𝑇𝑇𝐹 + 𝑀𝑇𝑇𝑅

o Available working time per shift at that process per day (in seconds, minus break, meeting, and
clean-up times). If we have 8 hours/shifts, it means 28,800 seconds, and 27,600 seconds of
availability means that 20 minutes are taken for the workers for rest.
o Buffer, and we have the quantity that we have in the buffer for each family of products.
As you walk the product’s material flow you will find places where inventory accumulates.
These points are important to draw on the current state map because they tell you where
the flow is stopping. We use a “warning triangle” icon to capture the location and amount
of inventory. (If the same inventory accumulates in more than one location between two processes,
draw a triangle for each location.) At Acme Stamping there is raw material inventory, finished goods
inventory, and inventory between each process. The observed amount of inventory is recorded
below the triangle, in quantity and/or time.

NB) Notice that available work time divided by cycle time multiplied by uptime percent is a measure of current
process capacity, if no changeovers are made.

4
 lOMoARcPSD

The third view of the VMS is understanding the physical/material flow (input and output).
After Acme’s last assembly workstation, products are staged in the shipping area according to the daily
shipping schedule and delivered daily by truck to the customer’s assembly plant. A truck icon and a broad
arrow indicate movement of finished goods to the customer.
At the other end of the map, we’ll represent the steel supplier with another factory icon. We use the same
truck icon and broad arrow to show movement of material from the supplier to Acme.
The steel supplier receives a weekly order from Acme and ships twice a week. We record in a data box that
the supplier’s pack size is a 500 foot roll of coiled steel. (In other words, the supplier cannot deliver less than
a full coil but can deliver any number of full coils, as requested). Once at Acme, the coil steel is taken to a
storage area, as shown by the inventory triangle.

The fourth view of VMS is understanding the information flow in the whole system. To do this we’ll need a
few more icons and arrows, in particular a narrow line to show information flows. This line is modified with
a lightening-like wiggle when the information flows electronically (via electronic data interchange) rather
than by paper.
The mapping icon for push movement of material is a striped arrow. At Acme Stamping, only the shipping
department is connected in any way to a “customer.” Each of the other processes is producing according to
a schedule, so the transfer of material from one process to the next is occurring via push. A push arrow is
drawn between each process.

5
 lOMoARcPSD

Information flow is drawn from right to left in the top half of the map space. In our Acme Stamping example,
we draw the flow of information back from the State Street Assembly Plant to Acme’s Production Control
department and from there to Acme’s steel-coil supplier. Notice there are separate lines for the forecasts
and daily orders, as these are different information flow.
o We have a weekly schedule, every single station says what it has to produce, giving them
information which are independent from the other station (the MRP runs every week, once in a
week, and gives the information to the stations)
o The MRP (Material Requirement Planning) is calculating also the raw materials
o The market issues every day an order, which is used for managing the quantity to deliver every day
with the shipment
o We need forecast because otherwise we are not able to understand the raw materials that we will
use and need

The timeline under the process boxes and inventory triangles is done to compile the production lead time,
which is the time it takes one part to make its way through the shop floor, beginning with arrival as raw
material through to shipment to the customer.
The shorter your production lead time, the shorter the time between paying for raw material and getting
paid for product made from those materials. A shorter production lead time will lead to an increase in the
number of inventory turns.

Why do we have inventory there?

We do not have inventory for the differences in cycle times, but inventories are justified for:
o FP, to react to unexpected order from customer;
o The batch production;
o Breakdown for maintenance;
o Setup time;
o Variability in the demand of the different processes;
As a consequence, stocks are born from the necessity to decouple each stage from the impact of the others
(for example for different schedules). In fact, different stages have different constraints (C/O, uptime), so
stages have different schedules.
We can understand immediately that one of the main drivers of the presence of inventory is the process of
production planning, supported by the MRP. Due to the fact that each stage is measured in term of
productivity and adherence to the established plan, the more it is able to produce exactly what it was
required (not more, nor less) in the right time, the higher will be its evaluation.
As a consequence, decoupled stages are necessary to give the possibility to each stage to realize what was
planned, independently from what could happen downwards, and to be run autonomously only on the basis
of its constraint, without limits imposed by the next or previous stages.

6
 lOMoARcPSD

Little’s Law: the value of the queue (stock if we are in a warehouse, the number of pieces waiting in the
buffer) is the demand (daily, hourly, weekly…) times the lead time. Higher is the LT, higher is the working
process, while demand is given and we cannot change it.

o We can use it to calculate the queue and then how many time a piece is in the queue, so the waiting
time, and the consider the processing time and the total production lead time
o Ex. Supermarket: every hour customers are leaving the supermarket. The flow is 90
customers/hours. If the customer remains in the system for 1 hour and half it means that we have
155 costumers and the flow is still 90.
o We are able to calculate the time that a product needs to be processed and wait in the queue? YES.
The queue is the value. What is the flow? Where we get the information? Product per month divided
for the working day and we have the product per day and this is the result.
Queue is a waste and also lead time is a waste.

SYSTEM PERFORMANCE
How to measure the system overall performance?
𝑣𝑎𝑙𝑢𝑒 𝑎𝑑𝑑𝑒𝑑 𝑡𝑖𝑚𝑒
:
the highest value is equal to 1 (target value), so if it is 1 the lead time is equal to value added time,
𝑡𝑜𝑡𝑎𝑙 𝑙𝑒𝑎𝑑 𝑡𝑖𝑚𝑒
which means pure flow, no stocks, no wait. If it is worst (lower than 1), lower the value higher the waste.
𝑝𝑟𝑜𝑐𝑒𝑠𝑠𝑖𝑛𝑔 𝑡𝑖𝑚𝑒
We can approximate it by (flow index), because processing time is the time in which we have no wait.
𝑡𝑜𝑡𝑎𝑙 𝑙𝑒𝑎𝑑 𝑡𝑖𝑚𝑒
The flow index measures the capability of the system to be as close as possible to an ideal one, where all the
processing time is a value added one and there are not wastes.
In other words, it is an indicator of the percentage of inefficiencies of the system, because it puts in evidence the
degree of wastes inside the process.

DRAWING THE FUTURE STATE MAP

To draw the future state map, we follow a standardized procedure to design the system in order to reach a certain
result. We have to answer the following 8 key questions:

1) TAKT TIME
Takt time: we take the net available working time (removing
breaks, rest of the workers…), we divide it by the customer
demand (taken from the first view of the VMS) and the result
is the takt time.

7
 lOMoARcPSD

Every 60 seconds we should be able to produce the product in order to be able to fulfil the demand of our
customers, so the takt time is the time at which the system should work in order to be able to fulfil the demand
coming from the customers. It is the target cycle time.
Takt time is used to synchronize the pace of production with the pace of sales, particularly at the “pacemaker
process”. It helps to see how you are doing and what you need to improve. On the future-state map, takt times are
noted in the data boxes.

EXAMPLE:

Therefore, it is evident how, in our case, there are some stages (pre-welding, final welding, final assembly)
characterized by a processing time lower than the takt time, while another (pre-assembly) that has a processing time
higher than the ideal value.
The principle of redesigning the flows derives exactly by the necessity to uniformity the working time at each phase,
in order to be aligned with the market’s requirements.
The answer to this issue does not lie on the creation of stocks; this represents a short term solution, but in the long
term, it will represent a waste and even a useless approach, since it will force to recur to overtime.
A first alternative could be the line balancing, which implies to redistribute workloads in a uniform way, in order to
respect the takt time.
However, looking at the lean principles, this is not the right approach: balancing and so trying to be as close as
possible to the market is not the correct solution to the problem, because we will progressively saturate all the
phases.
In addition, it takes the perspective of the resource, and not of the product (which represents the customer), like in
lean management.
In this way, the remaining part would be assigned to the last resource.
We can say hat this approach is focused on the product and on the client, rather than on the resource.
The right procedure is therefore the following:
1. We need to define all the activities required, in order to realize the product in the best way possible;
2. We have to compute the takt time;
3. We assign an amount of activities equals to the takt time to the first resource, and so on for the other ones,
until we reach the required workload. Obviously, as in the reference case, it is easy to assign to the last
resource as residual workload, obtaining in this way a non-saturated resource.
The main advantage of this procedure is the ability to put in evidence
all the wastes: in fact, we are able to separate resources and
activities, dividing the value added from the non-value added one. In
particular, it emerges how the first three resources are fully saturated
while the last one does not (we need to note that, in the previous
configurations, each phase presents inefficiencies, while in this case
only one – the spare time is still the same).
In addition, it is easier to resolve a problem that is circumscribed to
only one stage, rather than extended to all of them; but also if only
one phase is characterized by an idle time, it is possible to plan other
activities (for example the assembly of another product, maintenance
and cleaning).

The goal is the ideal approach of Kaizen, trying to find a way to do the same product by using less workload.

8
 lOMoARcPSD

In fact, if we look at the first alternative and if we are able to increase the
efficiency of the first three departments, we can think to remove the
unsaturated resource, moving it under the responsibility of the other
ones.
Obviously, this transformation is not lacking of difficulty; in particular, the
main constraint is represented by the redefinition of the resource’s skills,
which will imply the inclusion of additional activities (for example welding
is a very complex process and requires lots of training). It is necessary to
understand that barriers are not linked to the inability to execute the
operations, but to the cultural, image and heritage limits of the
organization and employees.
Lean approach revolutionizes the traditional wisdom, especially if there is
the possibility to modify the relationship with resources, passing from a
hostile to a collaborative one.

If those barriers are overcome, we have the opportunity to obtain the wanted improvements: in particular, the ideal
situation consists in working with the minimum number of people, indispensable to follow the pace of the market:

Minimum number of people = work content/takt time

In this exercise 7 seconds is the waste. This indicator has a double meaning, because it measures the degree of
waste in term of human resources, compared to the actual state.
It is important to note that if we buy a more sophisticated tool, another fixture, we can save one operator. With line
balancing we cannot think about this possibility.
However, it is mandatory to protect the resource: if one operator is spared, it will not be fired.
We can use it to increase volume because we are able to reduce costs and prices.
We can insource something that before was outsourced, saving additional money because resource is a fixed cost.
It is more convenient to dedicate time to other activities rather than waiting.

The key point of these consideration is the following: when we design a lean system (therefore compatible with the
takt time required by the customer), it is necessary to be able to structure it in a flexible way in order to adapt to
unexpected changes.

12/5/2016
2) SUPERMARKET OR SHIPPING
How a supermarket works? Goods are on the shelves and you pay to the cashier and then you leave. Every time the
cashier passes a good, this is no more on the shelves. There are employees that look at the shelves and see where
the level of products is lower, so they replenish the goods on the shelves. Today this happens in a IT way, the
employee looks at the computer.
At this level, we need to comprehend how, in the new configuration, it is more appropriate to produce. The solution
could be dual:
Direct shipping: we produce and we deliver the order activating the
process after the acquisition of the order itself.
The customer in the direct shipping is issuing the KanBan directly to
the assembly which is activated in the production due to the
KanBan. The assembly is not producing anything if there is no a
KanBan which activates the process. The KanBan is the switch that
activates the production. There is no WH and stocks, but we have
only a physical space to prepare the shipping. It is a kind of ATO and
MTO.

KanBan: traditionally is a piece of paper, but today we have also electronic

KanBan, where we have information about the item and a particular
quantity of that product and it is a very simple way to send information upward.
EX) All the KanBan are the same for the same product, for example 1 KanBan is 20 pieces of product A, so if we need
40 pieces of product A we will need 2 KanBan of product A.
9
 lOMoARcPSD

Why we use it? Because it is a simple tool and it avoids to make mistakes.
EX) McDonald: the kitchen and the cashier are divided by shelves, with one row for each type of burger and they are
normally empty apart for some type of products (the cheapest ones). The kitchen is not producing anything if not
activated by an order.

Supermarket: MTS raw materials are already in the warehouse, so once the order has been received, the
finished product warehouse prepares it, colleting a certain number of pieces. At this point, it is
communicated to the upstream stage that there was a pick of material, activating in this way the entire
system; as a result, all the stages will start to work, in order to replenish the inventory level.
It works exactly as a supermarket. Customers are ordering the quantity needed to the shipping process,
without interacting with the assembly (where there is a different information flow).
The shipping goes to the shelves and the quantity needed to be delivered is taken by the shelves, so the
products are ready. The assembly is activated by a KanBan, because each tray is one KanBan (it is associated
to the stock). The tray goes to the shipping area and the KanBan is sent to the assembly to inform it to
produce pieces.

If you reduce the number of KanBan you stress the

assembly. Why? What is the lead time for the line? The
system has to be quickly to react. According to the Little’s
law, the lead time has to be lower and due to the fact that
the queue is the number of KanBan. If we reduce Q also LT
reduces, while the demand remains the same. This is the
way to force the assembly to be quicker.
Overall daily quantity that needs to be sent (black
KanBan).

The main drivers that influence the decision are:

Value density of goods: obviously, if this element is quite relevant, it will be appropriate to avoid the make
to stock configuration, in order to do not immobilize too much capital. In addition, the larger the number of
finished goods, the more we go toward a make to order. Perishability, obsolescence and depreciation lead to
MTO;
Demand variability: the higher the variability, the higher the necessity to absorb this uncertainty with
stocks. This could be not required if the system is highly flexible;
Throughput time: it will be possible to operate on MTO/ATO only if the process and the system are
configured in order to meet delivery lead times required by the market (delivery lead time > lead time).
However, with reference to this last point, we need to keep in mind that the MTO logic will be sustainable only if the
arrival rates of the orders are constant and less variable along time, with the aim to guarantee a controlled level of
queues. We can absorb variation with a constant flow of orders, provided by a buffer at the beginning of the system,
that allows a high saturation and so efficiency. However, the drawback is that we have a queue, that must be
controlled and limit to a maximum level. If we overcome this level, we will recur to overtime.

3) CONTINUOUS FLOW
There are often spots in the value stream where continuous flow is not possible and batching is necessary. There can
be several reasons for this including:
Some processes are designed to operate at very fast or slow cycle times and need to change over to service
multiple product families (e.g. stamping or injection moulding)
Some processes, such as those at suppliers, are far away and shipping one piece at a time is not realistic.
Some processes have too much lead time or are too unreliable to couple directly to other processes in a
continuous flow.

Resist the temptation to independently schedule these processes via an independent scheduling function, because a
schedule is only an estimate of what the next process will actually need. Instead, control their production by linking
them to their downstream customers, most often via supermarket-based pull systems.
Number of trucks per day: If we have one truck per day it would be better to use supermarket, because the
black KanBan represents the overall daily quantity that needs to be sent, while with more that one shipping

10
 lOMoARcPSD

per day is better to use the direct shipping (Toyota: multi model line where the customer is close so the
KanBan is directly activated by the supplier).
The smaller is the KanBan, the better is to have as less as possible pieces in the KanBan.
The ideal situation would be to have 1 KanBan with 1 piece, but we do not do like this, because to prepare a
batch of products (even as small as possible) is more efficient due to the presence of set up times and so on.

Continuous flow refers to producing one piece at a time, with each item passed immediately from one process step
to the next without stagnation (and many other wastes) in between. Continuous flow is the most efficient way to
produce, and you should use a lot of creativity in trying to achieve it.

If we have 3 separated stations, how can we connect them? Suppose to have a product made of 3 components,
assembled by 3 workers.
Phases are not coupled: A produces following its schedule, retrieving materials from the upstream buffer
and putting the final piece in the downstream one; so do the B and C, referring to their relative upstream
warehouses.
On the contrary, in a continuous flow, the system is not decoupled, there are not stocks or intermediate
buffers or batches: inventories are located only at the beginning of the flow
The separate workstations generate stocks at the end of each station, without creating an efficient flow. With the
continuous flow we put the workers in the same line and station and we can reduce the stocks by a factor of 3. The
production rate is not changed, the throughput is the same, but in this case there are low stock, low costs and low
waste.

LINKING PROCESSES
What are the criteria that guide the choice to one configuration
rather than another?
Looking our example, we can recognize some reference criteria, such
as set-up time, probability of breakdown, dedicated resources or not
(for example we can adopt it for assembly and welding). However,
the shift from one logic to another is not “on-off”, but there is a
continuum of configurations:
Push: all stages are independent one from another and each
of them follows the established schedule, being valued on
the basis of the adherence to that plan.
Push the product: it means that I am in the upstream station
and I produce pushing the product downstream and as soon as the other station receives the product, I
continue producing. The speed of the system depends on the speed of the pushing: if the second station is
slower, the stock will arrive later. The first station keeps producing without caring.
Pull (supermarket): the upstream station is activated by the KanBan, so in this case we use the KanBan logic.
Pull implies to see the production process as operations that goes from downstream to upstream and works
pieces only when they are required (so when to replenishment order is received and there was a consume in
the downstream stage).
Inevitable in this configuration, phases are linked to each other: if the downstream stage stops, also the
upstream one will be not able to proceed, after having worked off the queue. In a push system, we cannot
put a limit to inventories; instead in supermarket we can control it, although the sequence in production is
not constrained, so the stage upstream can replenish the shelf in a different way respect what was done by
the upstream one.
11
 lOMoARcPSD

In the supermarket we can change the sequence, because what is produced depends on what is
requested/consumed.
FIFO: we have balanced the line and synchronized the stations, where each station is working with the same
sequence. An assembly line is FIFO by definition. It is characterised by:
o An information flow that maintain the right sequence between the stages;
o The limitation of the length of the queue: after having reached the maximum value, it is no more
possible to introduce further material in the system;
o The coupling of phases, which is even stronger: only when the downstream phase pick a
piece/product/component, the upstream one is activated.
As a result, the more the coupling is relevant, the shorter the queue.
One peculiarity of this solution respect the previous one is the capability to maintain and respect the right
sequence in each phase.
OPF: it is the extreme of FIFO, one-piece flow, in which the length of the queue of the FIFO approach tends
to zero: the logic behind is “make one, move one”, where stages are exactly in sequence and completely
coupled (they are coupled in term of overall amount of inventory, sequence and production rate).

As highlighted by the image, we can observe how, moving from the first to the last configuration, we are reach the
ideal system, to which the lean approach tends.
Given that, it is therefore possible to put in evidence the main criteria that allow to move in this direction; they are
the so-called DeCAF conditions:
Dedicated: the resources assigned to each stage but be dedicated to them, but also specified for a particular
product family;
Capable: stages must be “capable” both in term of quality and quantity: there must be enough capacity to
meet demand in the right interval, in the right quantity and in conformance to customer’s requirements. If
this could be impossible, it would be necessary to introduce buffers in order to manage these problem/lacks;
Available: down-town (time? must not have an impact on the system. If one stage has a low availability and
the decoupling stocks are small, there may be greater repercussions on the system in case of stops;
Flexibility & change-over: set-up times of each stage have to tend to the minimum value, in order to be able
to produce recurring to unitary lots. Wider lots would imply, at every product change-over, the necessity to
stop all the phases, for a time equal to the maximum time required by the resources.
The more they are in this way, the more we are able to couple them.
The more we want to couple the more we need to improve the system.

It is important to underline how these conditions must not be intended as a obtacole to the implementation of the
lean approach.
The right way of thinking is: “what should we change in order to have a lean system?”. It is a different perspective
respect “can we implement lean?”, basing our consideration only on DeCAF criteria.
Only in this way, we are able to achieve the real advantages of an interconnected system: it is not necessary to
respect the limit of all the DeCAF criteria, but to improve them along time; only in this sense, we have the
opportunity to make problems visible and we are obligated to resolve them. The main issue is to define the
minimum target of each dimensions, that we want to address; in particular, these minimum target act as
intermediate goals that, once they are achieved, must be improved, with the aim to increase the overall
performance of the system.

Where it is possible to use a continuous flow? Not in the stamping station,

because it requires a specific machine, but maybe in welding and assembly,
redesigning the system and training the workers.
With set up time this does not work and the ability of the workers should be
more or less the same, because we need to balance the line.

The main constraint is to remove stocks, we need to remove all the problems (we
have seen how in system physics). What is the minimum change needed to
achieve a goal? We need to opt for supermarket when we cannot achieve 100%
DeCAF: for example, in stamping, where we have to put shelves in order to allow
people to pick product from the shelf.

12
 lOMoARcPSD

4) SUPERMARKET

In a simplified supermarket configuration, there is customer and a supplier, where the customer has built inventories
of boxes, each of them containing ten pieces.
When the customers open a box, he/she detached from it a ticket, called kanban, with all the description about the
products; for instance the number of units, the supplier from which it comes from and the customer.
This ticket is then sent to the supplier. And so on for the other boxes, in this way the supplier has an idea of the
inventory of the customer, and follows the kanban to start production.
This is called kanban loop, because the tickets are sent again to the customers, once the order is fulfilled.
In this case we are not constrained by the sequence, but we need only to avoid that customers stop us.
We can classify order on the basis of their urgency.

It is important to keep in mind that this process can be replicated to all the other stages of the production system,
and that the supermarket system should be placed in proximity of the supplier, due to the fact that stocks represent
the inability of the supplier in following exactly the demand.

The situation is more complex if we consider a two - KanBan system or supermarket pull system, because here
there are two KanBan processes:
CUSTOMER PROCESS goes to supermarket and withdraws what is needs when it needs it. (“withdrawal”
KanBan)
SUPPLYING PROCESS produces to replenish what was withdrawn (“production” KanBan)
PURPOSE: Controls production at supplying process without trying to schedule. Controls production
between flows.

Customer, when opens a box, detaches a ticket called withdrawal kanban that triggers the transport and not the
production. From the storehouse, the transport kanban is substituted by the production kanban that then is sent to
the supplier in order to activate production. As a consequence, we can recognize two different loops: one for the
withdrawal and one for the production.

A “production” KanBan triggers production of parts, while a

“withdrawal” KanBan is a shopping list that instructs the material handler
to get and transfer parts.

The purpose of placing a pull system between two processes is to have a

means of giving accurate production instruction to the upstream process,
without trying to predict downstream demand and scheduling the
upstream process.
Pull is a method for controlling production between flows. Get rid of
those elements of your MRP that try to schedule the different areas of
your plant. Let the downstream process’ withdrawals out of a
supermarket determine what the upstream process produces when and in
which quantity.

These are several icons associated with a supermarket pull system.

Before you decide to use any supermarket pull systems, be sure that you
have introduced continuous flow across as many process steps as is
possible. You don’t want to set up pull systems, with their supermarkets
of inventory and the extra material handling requirements, between
processes unless you must.

In the continuous improvement logic, the supermarket system represents only a temporary solution; in fact, in the
long term, it will be necessary to plan proper intervention, in order to guarantee the transition from this
configuration to a more lean one.

13
 lOMoARcPSD

Where is the best position for the supermarket in the Acme VSM? The only way to
decouple the system in two parts is to use a supermarket, which needs to be placed
downstream of the stamping machine, where the set up time is very high due to
high changes (the high set up is not a good thing). The lower the set up time, the
lower the number of KanBan, so the machine can go from one product to another
in an easier and quicker way.

Moreover, Decaf is not possible between stamping and welding due to the fact that
stamping is not a dedicated resource. As a consequence, we have to opt for a
supermarket configuration, which allows to use upwards productive sequences and
management logics that are partially independent respect what it is required by the
downstream stage. It is evident that, if the cell starts producing with the same
sequence of the stamping phase, it should inevitably manufacture left products for
two weeks, and only right products for the other following weeks.

The future state after, having placed the supermarket, will look like: with two shelves for the 2 products and with
one shipping per day, without changing the value of the stock.
Takt time is now 60 seconds, and C/T is different. C/O is zero now. The total work has changed.

5) PACEMAKER PROCESS
The greater portion of wastes and inefficiencies in the current state derives from the fact that the production
planning assigned to each stage a specific schedule, strictly followed without a systemic vision. The result was the
creation of stocks and the misalignment of the schedules themselves: in fact, each plan is designed in order to
optimize the work load of each stage, protecting it from any negative event. However, these schedule derive from
forecasts, which are highly influenced by variability.
All of these have to disappear in the future state:
How future state works?
We receive the customer order, we sent it to the shipping department from which the order is delivered. Then
shipping communicates to weld+ass that it is necessary to replenish the shelf.
So they start producing the amount of product that were delivered and they put them in the finished product
warehouse of the supermarket.
There should be one single scheduling point, on the basis of which the other stages have to adapt and react. Only in
this way, we have the chance to minimize the probability of misalignment between the phases, building a system
that is highly flexible and easily controllable: in the moment in which the central stage is obligated to review its
plans, also the other ones have to consequently react, realigning to this change.
14
 lOMoARcPSD

The yellow one are the changes to implements in order to achieve these results. (ACME future state picture)

The more we reduce the lead time, the more changes we are able to achieve, the more the EPE and the
batch size will be reduced.

The pacemaker is where we schedule (scheduling point) and it gives the rhythm to the system.
By using supermarket pull systems, you will typically need to schedule only one point in your door-to-door value
stream. This point is called the pacemaker process, because how you control production at this process sets the
pace for all the upstream processes. For example, fluctuations in production volume at the pacemaker process affect
capacity requirements in upstream processes. Your selection of this scheduling point also determines what elements
of your value stream become part of the lead time from customer order to finished goods.
Note that material transfers from the pacemaker process downstream to finished goods need to occur as a flow (no
supermarkets or pulls downstream of the pacemaker process). Because of this, the pacemaker process is frequently
the most downstream continuous-flow process in the door-to-door value stream. On the future-state map the
pacemaker is the production process that is controlled by the outside customer’s orders.

CASE 1) The first one is directed by the finished products

warehouse: the entire process produces for the final buffer,
which means that programming is done on the basis of the
requirements that come from the customer. As a consequence,
it is necessary to put the scheduling point in correspondence to
the last stage: every pick to satisfy demand activates the
previous stages in order to obtain the replenishment of this
pick in the case of MTS or supermarket.
EX) If number 1 is controlled by number 2 and number 2 is
controlled by number 3 and number 3 is controlled by number
4, the scheduling point is at station number 4.

CASE 2) The second options consists in fixing the scheduling

point in the phase that receive directly the customers’ orders.
In this case, it is evident that the downstream stages will react
to and follow the sequence established by the scheduling point, but also the upstream supplier will act in turn in
reaction (in the moment in which the production will start, it will have to restore consumptions). In this way, there
will be not the possibility of misalignment: in fact, there will be the exact maximum amount of stocks, permitted for
a FIFO logic when we differentiate our products

The second station pushes the other stations (controlling them) so the scheduling point is at station 2. Process 3 is
activated by process 2 and process 4 is controlled by process 3, so both 3 and 4 have no freedom.
The schedule point is placed where there is the highest degree of freedom.

Anyway, we need to keep in mind that the adoption of one configuration or of the other requires that the stages
that have to react should be characterized by a high level of flexibility, due to the fact that they have to react
immediately to changes. In addition, the higher the flexibility, the less the necessity to schedule again.

Acme VSM: the schedule is downstream the supermarket.

We have obtained a better and more efficient system respect the original one; a further proof is given by the fact
that with this configuration it is possible to operate with a lower level of stocks.

LEAD TIME IMPROVEMENT

15
 lOMoARcPSD

6) HOW TO LEVEL THE PRODUCTION MIX

Levelling the product mix means distributing the production of different products evenly over a time period.
For example, instead of assembling all the “Type A” products in the morning and all the “Type B” in the afternoon,
levelling means alternating repeatedly between smaller batches of “A” and “B”.
The more you level the product mix at the pacemaker process, the more able you will be to respond to different
customer requirements with a short lead time while holding little finished goods inventory. This also allows your
upstream supermarkets to be smaller. But be warned that levelling the mix requires taking some pains in assembly,
such as more changeovers and trying to keep all component variations at the line at all times (to eliminate
changeover time). Your reward is the elimination of large amounts of waste in the value stream.
The icon for levelling is the following symbol, which is inserted into an information flow arrow.
How will we level the production mix at the pacemaker process? How do we level production?
We can level in term of volume (same amount every day) or level in term of mix (it is more complex and it means to
change the perspective).
The traditional approach implies to change as less as possible, from large to average to small.
The lean approach, instead, says to produce the same every day: we reduce the impact from changing from large to
average to small, but we have the advantage to produce the same sequence every day.
In other words, it means to product in a short interval the average mix/proportion over a month. We replicate the
same proportion on a shorter frame. In order to do that, we can rely on the load- levelling box.

LOAD-LEVELING BOX
A good place to start is to regularly release only a small, consistent amount of production instruction (usually
between 5-60 minutes worth) at the pacemaker process, and simultaneously take away an equal amount of finished
goods. We call this practice a “paced withdrawal”.

We call the consistent increment of work the pitch (pacemaker interval), and often calculate the pitch increment
based on packout container quantity (the number of parts one finished-goods container holds), or a multiple or
fraction of that quantity.
For example: If your takt time = 30 seconds, and your pack size = 20 pieces, then your pitch = 10 minutes (30 sec x 20
pcs = 10 minutes).
In other words, every 10 minutes:
give the pacemaker process instruction to produce one pack quantity;
take away one finished pack quantity.
So in this case pitch means multiplying your takt time upward to a finished-goods transfer quantity at the pacemaker
process. This then becomes the basic unit of your production schedule for a product family.

A load-leveling box has a column of kanban slots for each pitch interval, and a row of kanban slots for each product
type.
Pitch, which is the interval of control and the lot of
information that are sent by the pacemaker (the
scheduling point).
In this system, the kanban indicates not only the quantity to
be produced, but also how long it takes to produce that
quantity based on takt time.
In fact, at this point it is not only required to define the right
production sequence, but also its speed: provide orders at
smaller intervals means communicating to proceed in a faster
way, while diminish the quantity means proceeding in a
slower way.
Kanban are placed (loaded) into the leveling box in the desired
mix sequence by product type (see levelling box illustration).
The material handler then withdraws those kanban and brings
them to the pacemaker process, one at a time, at the pitch
increment.

If there are 3 types of products (A, B, C) and three rows and one KanBan means one tray, you take A and you
produce A, letting empty the column, so you go to the second column. The process is in sequence from left to right.
16
 lOMoARcPSD

In this case pitch is equal to 20 min: this guy, every 20 min, goes to the levelling box and takes all the KanBan which
he sees in the first column (in this case A, while in a more generic situation we could have A and B and so on).
If there is more than 1 KanBan, the peacemaker will take more than 1 KanBan per pitch.
One column one pitch, every 20 min. This is the way to synchronize the system. The pacemaker is a point and a
physical person.

EX) Suppose that a kanban corresponds to a tray of 20 pieces and that the takt time (final demand) corresponds to 1
piece every minute; this means that one box will be required every 20 minutes. If we decide that this is the control
slot, each column will have at maximum one kanban, due to the fact that one ticket is equal to 20 minutes. Each
phase has to produce one kanban every 20 minutes and not more or less; otherwise, we will incur in overproduction
or we will not be able to meet demand.
This is the levelling in term of volume, because every 20 minutes one tray has to be produced.

In this ambit the resource that is responsible for the respect of the takt time is called runner; the operator moves
inside the factory, paying attention to the flow of kanban, as shown by the following figure:
Shipping picks a tray and detached a transport kanban;
The transport kanban is then given to the runner,
which brings it to the cell, allowing the star of the
production;
If the pace is respected, the cell should provide the
exact amount of products to replace the previous
consume, which then is transferred to the
supermarket;
Subsequently, the runner will take a new
transportation kanban and will start another tour.
He/she goes through the tour in a defined interval of time: if the frequency is equal to one kanban at a time, the tour
will be equal to 20 minutes. It represents a sort of feedback and a check of the production pace of the stages, acting
as a real metronome (in fact we can understand the degree of delay).

Obviously, we can deal with more sophisticated situation,

as shown.
In the case we plan the levelling, the transportation kanban
is not directly transferred to the cell, but to the resource
that manages the load levelling box, which will insert it in
the correct column. Finally, the runner will pick the kanban
located in the first column on the left side, bringing it to the
relative stage.

It is fundamental to keep in mind that we need to schedule and level production always on the basis of the target
determined by the takt time: if the demand will be higher than this value, the surplus will be allocated to the
following day, in addition to the demand that it will occur in that day. In other words, we produce on the average
value, leveraging the inventories in case of variability.
We need to pay attention because these considerations are not valid in the case of a lower value of the demand
respect the average. In fact, without kanban, we are not allowed to produce.

Since the redesign of the process looks at a system that produces an entire family of products, it is appropriate to
comprehend which are the logics that lie behind the levelling and sequencing of the mix.
In an ideal system, we should produce the exact mix that the customer requires. However, it is possible to identify
many different alternatives.

17
 lOMoARcPSD

The first one coincides with the no-levelling: the customer sends the
order to the warehouse, which picks the required boxes and prepares
them for the shipping. Picking a tray, one or more transport kanban are
detached and sent to the cell (as an input for the replacement) without
levelling (the cell sees the arrival of the kanban on the basis of the same
logic of picking in the warehouse).

The second option requires to level the kanban before

they are transferred to the cell: the system does not
receive the entire amount of batches, but one at a time, on
the basis of the levelling logics.
This guarantees that on single kanban for each column of
the load levelling box is produced, regardless the type of
picking in the warehouse.
As a result, the cell is not able to decide to produce a batch
that is different from the one defined by the kanban
(increasing in this way the batch size).

The final alternative coincides with possibility to level upward the

finished products warehouse, so before the scheduling point:

20 number of units for each kanban

OX icon for levelling, mix and volume

B we level demand and mix

C we level the withdrawal kanban, we do not produce

In other words, we do not sell to the warehouse the order but a

production kanban: in this way, we are able to level production
even before the execution of the shipping, which receives as an
input a levelled scheduling.
The main benefit of this solution is to minimize the inventories in the finished products warehouse.

7) What increment of work will you consistently release and take away at the pacemaker process?
What increment of work are you releasing?
How often do you know your performance to customer demand?
This is referred to the management interval, so the pitch. It is normally smaller than 1 hour, because it gives the
rhythm of the system and we should decide where to put it.

We have seen that the “runner”, also called “water spider”, is the figure that has the task to collect kanban with a
certain pace. It is evident that this pace depends on how the control interval and the transfer of the kanban have
been defined. As a consequence, it is possible to schedule in order to transfer a kanban every 20 minutes, so 3
kanban per hour (the amount of information and workload that we bring to the pacemaker, and we bring back, is
one hour worth) and so on.
However, change the interval means also organize in a different way the columns of the load levelling box, due to
the fact that a column could correspond to 20 minutes of the customer demand (extremely punctual), or to 3 hours
of demand (with an higher aggregation).
18
 lOMoARcPSD

It is obvious that a more punctual solution allows to monitor strictly the pace of the production: on the one hand, we
are able to check what we have produced every 20 minutes; on the other hand every 3 hours. There is a very
relevant difference.
With a control that is exactly equal to the kanban, the check is more stringent and efficient; however, we must keep
in mind that this advantage implies a higher waste in term of runner’s movements. But, the more the management
interval is punctual, the more the production works without problems, respecting the pace. In fact, working in a so
small interval, if there would be problems, we will be unable to resolve them.
As a consequence, it is appropriate to proceed gradually and having a more strict control only if the system is able to
improve.
The concept of pitch represents this dimension; it is the managerial interval of
control (how often the runner starts his tour and check production).
We do not have to forget that the pitch is correlated to the production volumes
and workload: the more the product is complex and volumes high, the more the
control should be stringent.

In some case one day is more than enough, in other they work by the week (the
due date is set in weeks). This depends on the business.

The larger this amount the more difficult is to level demand.

It is much easier to produce 24 trays in one day than 1 batch every 20 minutes.

8) What process improvements will be necessary for the value stream to flow as your future-state design specifies?
Considering also the levelling of the demand we can discover
that the lead time decreases to 4.5 days.
From push to pull first improvement
Level demand second improvement

EXAMPLE: HOSPITAL
In the hospital there is a patient that is considered as a raw
material. He needs to be processed and then he will leave the
hospital with a higher value. He has to go through different
processes, so different areas of the hospital with their queues.
Any time there is a process we can represent it with VSM and
apply the concepts seen to reduce the wastes and move away the no value added activities (NVA) and focus on the
value added ones (VA).

19
 lOMoARcPSD

19/5/16
SYSTEM PHYSICS

It was introduced by some physicians, using some quantitative approaches and applying some physics theories to
develop a very structured approach for lean management. We start considering basic concepts.

PRODUCTION CAPACITY
It’s possible to measure production capacity in different ways and to focus on different levels (single resource,
department, company). At single resource level, the fundamental element is theoretic capacity (eg. unit/hour).
Theoretic capacity refers to the conditions when everything goes well
Eg. Unit cycle time: 6 min, theoretic capacity: 10 unit/hour

DEF) Cycle time is defined as the frequency with whom a production resource (stage, station or the whole plant)
exits some elements (output).
We can define the cycle time for each activity in the world. Each time an activity (equal to the others) is finished and
produces an output, we have cycle time.
EX) Each 5 sec a station produces a finished product (one piece, product) C/T = 5 s/p.
EX) Machine that works with a batch of 5 pieces. Each 5 s it will produce 5 pieces C/T = 5 s/p / 5 p = 1 s/p

To determine the actual capacity, it’s necessary to also take into account other factors...

AVAILABILITY
A resource may be not available to produce. Part of working time during which the resource is available for
working is named availability and its symbol is A.
Causes for unavailability can be:
• Breakdowns
• Interruptions (for problems, of for calls)

Actual Capacity = Ct * A

For example, if availability is 80%, a resource with a theoretic capacity of 10 units/hour, in the mid/long term will
produce only 8 units/hour.
It is the effective time that you have excluding break downs and other drops of other organizational issues. It is a %
of the theoretical capacity.

SERIAL SYSTEM
What is the capacity of the system?
We start considering this simple system, where each phase has a different C/T, that is a measure of the its capacity.
In a system like this, we do not have any indication about stocks. We should consider what is the worst C/T, that is
the worst rhythm that affect the whole production capacity of the system. In this case it is 6 min/pc.

o Capacity is limited by the slowest phase (medium-term viewpoint): C/T = 6, C = 10 u/h

Categories of systems:

DECOUPLED SERIAL SYSTEM

In this case we have stocks between each phase. Each phase can work at its own rhythm without any problem.
By respecting the pace of their own phases, stocks are going to be accumulated between phases.
So stocks are used to make each resource work alone, on its own.

o System capacity is 10 u/h

Queues between each phase do not balance out different C/T

We should consider the availability: the effective production capacity of each

single resource changes. If we consider phase 1 with 90% of availability, the
production capacity changes.
 lOMoARcPSD

The constraint of the system is represented by phase 3, with 8.4 u/h, so the bottleneck changes in the system.

COUPLED SERIAL SYSTEM (LINE)

Phases work together, respecting the same schedule, so they are no more
independent as in the previous case. Each time a phase stops for any reason, all
the production is interrupted.
We have different availabilities and C/T. Each time that one machine stops also
the others stops, so we have a joint probability.
The availability of the whole system is given by 90% * 80% * 70% and C/T is
defined by the slowest C/T (bottleneck) between the different phases, which is
the one that affect the work of the other phases, in this case it is phase 1.

o Queues disappear
o System availability is the product of the single availabilities
A = A1*A2*A3 = 0,5
Capacity is therefore 60 * 0,5/6 = 5 u/h

Why this system? What are its advantages?

There are no stocks: a stock is a queue, because stocks are waiting to be processed. If we do not have stocks, we do
not have queues inside the system and waiting time, so maybe the throughput time is lower.

Adding queues
If we add queues to the previous system, we choose the worst one as the
bottleneck of the system (the worst C/T) and the production capacity of the
decoupled system becomes 8.4 u/h, so it is higher than 5 u/h.
• It seems like queues/stocks bring an improvement, but it’s not true:
they are just an increase of waste!
• Eliminating stocks, we can decrease significantly the waiting time,
being more efficient.

IDEAL COUPLED SERIAL SYSTEM

With this system we can reach a capacity of 12 u/h, more than any
previous system. Going towards an ideal system, the coupled system
becomes more effective and the same happens for the decoupled
system, reaching both 12 u/h (there are no differences between the
ideal coupled and ideal decoupled systems).
If we balance the C/T (5 for all phases), we improve the availability
(100% all phases), we eliminate the activity related to inefficiencies and
we improve all the parameters of the system, the capacity of the system will tend to ideal, both for coupled and
decoupled systems. There will be less waiting time and lower throughput time.

Decoupled system is better for throughput time, but to adopt it we have to work a lot on the inefficiencies.

SET-UPS AND SYSTEM FLEXIBILITY

Set up is not included in availability, but it is considered as an inefficiency of the system, because it is a non value
added time. We have a system with these characteristics:
The system works 8 hours/day and daily demand is 60 units (constant in volume).
The system has a range variety of 10 products, and they are all requested every day (60 products per variant)
Is there any problem in delivering to order the requested quantities with a delivery time of 1 day?
 lOMoARcPSD

C/O (change over) is a synonym of set up. Let’s calculate 3 components:

• Tp: Time to produce (only to process)
Tp = Average demand (pc / day) * (C/T) / A
• Tsu: Time to Set-up
Tsu = Number of set-ups (set-up/day) * C/O
• Ta: Available time (daily)
(Tp + Tsu should be compared with Ta)

Calculations:
• Tp + Tsu = 60*6/0,9 + 10*10 = 500 minutes
• Time Available: Ta= 480 minutes
The company cannot satisfy market demand in make to order…Thus it has to intervene in order not to lose profits!
What can be done?

BATCHING
The company does not produce the variety requested one day, but with a wider window of the demand.
In order to have more time to produce, companies usually batch production of each single product.
EX) Doing set up each day, if we have to produce 60 units for 5 days a week, we will produce 300 units weekly and if
we have 10 set ups per day, in a week there are 50 set up. The quantity to produce remains the same (60 300),
but if I consider a weekly demand, the time to set up is lower. The same can be done for 1 month.
DEF:
Quantity Batching defines the exact (sometimes minimum) quantity to produce every time a product is to be
produced.
Time Batching, defines the times a product is produced over a defined time horizon (e.g. once/week)
If Time batching is used, and each product is realised every two days (instead of being produced daily, so 5 different
products/day instead of 10):
Tp + Tsu: 6/0,9*60 + 5*10 = 450 minutes
Ta = 480 minutes

That implies:
– Advance of unrequested production: I am able to produce on a reasonable time window, but some of the
amount of products produced are not still requested by customers, unrequested production. I will have a lot
of stock and capital invested, so a lot of queues and waiting time higher throughput time.
– Having inter-operational/finished products stocks: they are affecting the waiting lines and the final
throughput time.
– Increase of system response times: the system is full of stocks that decrease the flexibility of the system
– Increase of throughput time variability, system congestion (need of space) and coordination needs
(between product types)

The main problems are:

• There are stocks (need for space, costs increase)
• Production in advance of demand not yet requested!
• Risk to keep stocks of not requested units! (for example, when demand has variability)
• Flexibility decreases (if today code F is requested and it is not in stock nor scheduled for today’s production,
it could be produced only adding a rush order that creates inefficiency).
• Increase coordination and supervision needs in production process

Batching vs reducing set up time. What enterprises do?

• Batching is THE way companies adopt, because it solves the flexibility problem of the system, even if it
entails some disadvantages and it’s NOT FOR FREE!
• It’s a lot better to pursue the reduction of set up times!! They should try to improve the parameters
(availability and set up time) that affect the throughput time of the system. Just reducing set up time from
10 to 8 minutes, it’s possible to produce every day each product in the requested quantity! (without having
to adopt batching)
 lOMoARcPSD

EXAMPLE 1
o The following stocks decoupled system operates 8 hours per
day and has a daily demand of 60 units (constant in volume).
What is the minimum batching for each of the 3 phases?
o The system has a range variety of 10 different products that are requested every day. Each machine
produces 10 different products (with set up at every product change).

Decoupled system. The minimum batching is a measure of the flexibility of the system (KPI). It is calculated for each
production unit, which are decoupled. The final objective is to improve the production reducing the quantity of the
batch, so producing more frequently with a lower quantity.

MINIMUM BATCHING (DECOUPLED SYSTEM)

To calculate the minimum batching, it’s necessary to understand how many sets up can be done in each phase every
day.
o Phase 1: number of sets up per day?
6/0,9*60 + X*10 = 480 minutes 8 set-up per day (this is the flexibility of the system, given by its fixed
characteristic)
The average batch quantity for each product is 60/8 = 7,5 units (I cannot produce batches with a lower
quantity, it is the limit of the system)

o Phase 2: 4/0,8*60 + X*15 = 480 minutes 12 set-up per day (that is, if orders are grouped per day, there is
no need to produce more than the ordered quantity). The concept of minimum batching can be applied, but
it is not important, because in this phase the company is able to produce the quantity and variety asked by
clients.

o Phase 3: no constraints on the batch, because there is no set up (C/O = 0). I can work with batches of 1 unit
(A,B,C,D,F and so on repeating)

MINIMUM BATCHING (SERIAL COUPLED SYSTEM)

The system is coupled.
If a phase stops for set up, so do the others.
It’s therefore important to understand how much time the line is stopped to make set-ups (total time of stop
for the line is the total time of stop for set-ups)
In the case that for each product change corresponds a set up at each phase, when a phase does the set up, so do
the others.
The time for a set up is 15 minutes (longest set up time, if setups are performed in parallel. It is the sum of setup
time if setups are performed sequentially. E.g. by one single person)
o System’s Cycle time is 6 minutes
o System’s availability is 85,5%
Therefore:
6/0,855*60 + 15*X = 480 about 4 set-ups per day

MINIMUM AVERAGE BATCH: 60/4 = 15 units (that is, 2,5 days of customer’s demand for each code!!)
In this current situation batching is needed.

EXAMPLE 2
Daily demand 60 units/d (constant in volume), time available for production
8 hours/d and variety 10 different end products (each one is requested
every day). The product rages change between phase 1 and 2:
There are 5 different product types at phase 1.
At phase 2 there is a further differentiation leading to 10 different products types.
What is the minimum batching of the system (coupled)?
What would it be the minimum batching of the system if de-coupled?
 lOMoARcPSD

COUPLED:

Each time I have to do a set up in a phase, also the other phase does set up with its rhythm (there is a parallel set up)
and who finishes before waits for the other. Then we move to production and after some time, there is another set
up needed for phase 2 (7 min). The cycle continues. We have to do this for each product, so 5 times for phase 1 and
10 times for phase 2. Each time I have to do a set up in one phase, the other phases waits.

o TOT set up time (10 products) = 85 min

o Average for each single type of product = 8.5 min

The average time to change from a finished product type to another: 8,5 minutes

o Cycle time: 6 minutes

o Total availability:85,5%

6/0,855*60 + 8,5*X = 480 around 7 sets-up per day

MINIMUM AVERAGE BATCH: 8,6 units (that corresponds to the average demand of 1.43 days for each type of
finished products!!)

SET UP (FLEXIBILITY)
Production change requires set up.
The longer the set up time is, the less frequent the production changes will be
The longer the set up time is, the greater the risk of producing something that is not requested will be
This has obviously impacts on performances, increases costs (stocks, necessity of coordination…), increases
throughput times variability and decreases system flexibility

PRODUCTION CAPACITY
• Theoretic production capacity indicates the number of products realized in the time unit that the resource is
able to produce under optimal conditions
• Availability takes into consideration resource’s stops that decreases the theoretic production capacity
• Sets up decrease the time available for production and system flexibility. Batching is used to increase single
resource’s efficiency, but it creates problems at a system level

COUPLED VS DECOUPLED SYSTEM

• A system decoupled by stocks allows to make production phases more independent, but it’s a system that
requires stocks to work, thus showing that is a system with problems.
• A serial coupled system links the production phases. This highlights problems and allows to see and address
them.
• The objective is to remove problems (remove the causes of the problems), so that the system improves.
 lOMoARcPSD

26/5/16
VSM-1
EXERCISE 1
1) Current state Map
MATERIAL FLOW
Start from the material flow: all the physical flows of the company.
We start from the modelling of the different production stages, which all have only 1 worker. Then we add the
necessary information.
Then there is SHIPPING (stand alone stage) and then we identify customers and suppliers. This is the standard
procedure.
CUSTOMERS: 3 necessary information and the picture of a truck with the frequency (daily). The same for
SUPPLIER, the picture with the frequency.
Between each stage we position INVENTORIES (black and white arrow means PUSH).

TYPICAL QUESTION: why we have a number of inventories at the beginning? It is random, so an engineering
approach to define a realistic picture in time of the company, without a specific logic.

CALCULATIONS: first of all, we calculate the takt time (the rhythm that we should respect in the production to have
a constant and regular production). It’s like the rhythm of the pacemaker. It is not necessary the rhythm of the
production, but the time that we should respect.
In order to respect the daily demand of customers, and considering the constraints of the opening time, we should
produce with 120 sec/pc, so this is the standard that we should respect.
The important thing of takt time is the fact that with takt time we can give a temporal value to inventories, because
here we want to calculate the timeline of a general piece that enters the system at the beginning and exits, so its
throughput time.

We can calculate the amount of inventories at the beginning of the system:

(10 dd * 900 min/dd * 60 sec/min) / 120 sec/pc = 4500pcs

Weekly plan: delivery to each production stage

Daily plan: delivery to the shipping stage
They have linear arrows, so information provided by paper communication

Customer forecast, daily orders

Supplier forecast, weekly orders
They have non linear arrows, so information is provided by electronic communication
 lOMoARcPSD

TIMELINE:
o NVA activities, related to
inventories. We have to
translate the number of units
of inventories in days (using
an analysis of units of
measures)
(1500*120) / (900*60) = 3.33

o VA activities, related to the

production activities. In this
case, we have just 1 operator
and 1 machine, so the
working time is the same
working time of the pieces
(cycle time)

At the end we calculate the sum of

the times needed in each phase.
2) Improvements to obtain EPE = 1
day?
They produce with MTS and push, with batching. They produce more in an inefficient way, creating too many
inventories, but satisfying the demand. The objective is to produce just what is needed, to improve efficiency and
flexibility.
We are directly considering the less flexible stage (C/T = 95).
Every part every is not calculated when C/O is equal to zero, it doesn’t make sense.
EPE it is not an unknown variable, but in this case it is equal to 1, so we multiply by 1 both the production time and
the available time. In this case the variable is the set up time, while in the case in which EPE is unknown, we should
multiply by X production and available time. 68 is the number of products per reach range.

Improvement: reducing the setup time by SMED (Single Minute Exchange to Die): reducing the set up time from 10
min to 2,2 min, we can reduce C/O (NVA activities), improving efficiency and flexibility.

EXERCISE 2
1) Current state map
New things:
1. In this case there are two parallel activities, with their own inventories both before and after the process.
Both press and cutter are not dedicated at the production family.
Talking about a line, we make reference to a product family, because each line is dedicated to one family. It
can happen, that some resources of the line, usually the first ones (with more general production aim) can
not be dedicated 100% to one family. We include them in the line, but considering this aspect.
2. To calculate the time to put in the timeline in the case of parallel activities, we have to put only one value,
according to the critical path: we consider Press because it has the critical path (the one with the higher TT,
throughput time). Otherwise, in cutter we know that for each pressed thing we have 2 pieces coming from
the cutter that should be welded on. (SEE PICTURE BELOW). Once define the critical path, we put all the
value of that path. If the critical path was Cutter, we should have put 20s instead of 4s.
3. Starting from the picture below, I know that for each pressed piece I have to provide 2 cut pieces. The values
of inventories are related. (20 000 pc and 21 600 pc)

WELDING
 lOMoARcPSD

PRESS

CUTTER

2) EPE?
EPE is calculated for each stage and then the EPE is the worst EPE of the system, because it conditions our
production.
What is the minimum EPE for the company?
It's necessary to calculate the time, in days, necessary to restore the product range, under the constraint of available
times and the demand required by the customer.
T. production + T. setup ≤ T. available
o T. production = C/T * demand (for a period equal to EPE)
o T. Setup = C/O * n of products in the range
o T. available = t. available (for a period equal to EPE)

3) Future state map

Key 8 questions (first 5 the most important ones because they condition the system, 8 is fake it means just to write
the future state, 6 and 7 do not mean calculations, so probably not requested in the exam)
1. Takt time of the family

2. Supermarket or shipping; how to organize the production?

These drivers are the main characteristics of the system, so we should use these to provide an answer. We
must no copy the characteristic of the current state. In the exam, the answer will be unique depending on
the information provided that will guide you in a direction.
 lOMoARcPSD

o Delivery time to the customer: short delivery time is better to work with MTS. We can do MTS and
MTO and being lean anyway
o Product features (product’s value, obsolescence level, standardization level of the product):
If the economical value and obsolescence risk are high, working with MTS makes no risks
The more the product is standardized, the more we can work with MTS.
o Demand predictability: the more the demand is predictable the more we can work with MTS
o Demand stability: the more the demand is stable, the more we can work with MTS
3. Where to put the flow? What are the flows that respect DECAF.
Dedicated: machines first of all should be 100% dedicated to the production family. Only those machines
should be considered.
Capable: the cell should be capable. Capability: C/T of the cell can work at takt time or lower value. If C/T
of the cell is higher than the takt time (obsolete conditions) there are big problems.
Available: is the cell available? Yes, in this case
Flexible: it is related to C/O. Usually the answer is NOT. But lean is continuous improvements, so we
should propose some improvements in order to be able to produce at the takt time.

We calculate EPE of the cell (0.17 < 1) without having to propose improvements to set up time and C/O because the
system is already flexible enough (EPE < 1). When EPE > 1 you should propose improvements.

2 operators are needed: the total work load of the cell is 45s (10s + 12s + 23s) To produce 1 product in the cell we
need 45s of VA activities. C/T is 23 (the worst). If I know that the total work load is 45s, I do not need to produce
more than the takt time and so just 2 operators are needed for the cell. 1 works with C/T similar to 10s and 12s and
the other one with 23s. Yes, but the C/T are different! Only if C/T are reasonably similar we can do like this.
POSSIBLE IMPROVEMENT: let’s divide equally the total work load of the cell between the 2 operators: 1 operator
works with 22.5s and also the other one.

4. Pull supermarket? The drivers to be followed are:

High production batches: the supermarket is suggested between the resources which works with
high batches and the following
Low machinery reliability: if a machine is not reliable, after the machine we put the supermarket to
have decoupling and batching strategy
Shared resource: if a resource is not dedicated, like Press and Cutter, we should put after them a
supermarket to decouple
In this case we put supermarket after Press and Cutter.
When we put a supermarket, we have to propose improvements to the resources before the supermarket.
PRESS STAGE: reach EPE = 1 day (very usual). We use the same equation, with set up time as
variable.
MILLING STAGE: reach EPE = 1 day.
 lOMoARcPSD

FINAL CONFIGURATION: working with make

to stock (MTS)
We put the cells
Supermarket
Still push logic (OXOX)
Kan ban added
Symbol: the shipping is not using kan
ban but simply taking from
supermarket what needed (pull
production, circle arrow)
Pacemaker: only production plan is
put soon after the last supermarket
stage (from left to right, rule)
Shipping

If we work with make to order (MTO)

FIFO queue instead of
supermarket
Pacemaker on the cells
The throughput time from 1 day
becomes 0,5 days, due to 2xday
shipments

RESULTS COMPARISON
Results by just reducing NVA activities:
 lOMoARcPSD

VSM-2
EXERCISE 1
1) INITIAL CONFIGURATION
There are 2 methods to calculate if we have enough time to produce our daily production.
We have 2 operators working in parallel (together) on the same product.
Each production cell receives every day units belonging both to family 1 and to family 2:
o Trying to balance them, but producing only 5 tractors a day, it happens to make 2 of family 1 and 3 of family
2 or vice versa.
o It should be verified that there are no problems in meeting demand even producing three tractors of a type
and 2 of the other

Method A: Workload vs Capacity method

We are assigning a variability just to the production phase and not to the set up phase.
We take the worst case and we calculate Tp and Tsu:
o On average, the work load for a tractor in the worst case is 152 min.
o On average, the set up load for a tractor in the worst case is 110 min.
SETUP SEQUENCE: this because I start the day with the production and I end the day with the set up.
Each day the production begins with the family of the final product of the previous day.

Productive capacity of the work cell:

7,5 h (opening time) * 60 min/h * 2 operators = 900 min of labour available, in which we should do the total set up
time (110) and work load for production (152*5).
152 * 5 + 110 ≤ 900 No need of overtime!

Method B: Work Pace method (C/T, C/O, ...)

We consider directly the perspective of the C/T of the production balanced between the 2 different operators
(reasonable assumption). So the reference point in this case is one operator.

2) EPE?
See slides

3) FUTURE STATE
1. Takt time:
2. MTO?
Cumbersome: it takes a lot of space to have products and tractors
Infinite range (no MTS) Main constraint
Demand constant in volume, leads us to MTS
Even if one driver leads to MTS, but another one is stronger and against MTS, we should not do MTS.

3. Continuous flow?
We could dedicate lines hence separate the two production families:
Pros:
1. I decrease the C / O because I no longer have to change from one family to another.
2. Exploit specialization: increased efficiency and reduced costs.
 lOMoARcPSD

Cons:
1. The change in mix becomes very critical because it could create imbalance, then you must take into
account demand variability
In this case we can separate them because the demand is constant.
FAMILY 1: in general, the workload of a product of family 1 is 160 min (WKC) and it will be
distributed on a line with 10 operators (20/2). We have to consider also set up time between
different products of the same family.
One operator is not necessary for definition, while on the other hand the set up time is not null and
since we cannot put it equal to zero, we should consider the number of operators and rebalance the
WKC between the operators according to a logic: the exact value 17.78 min for each operator, and
the remaining time is for set up time. DECAF condition used to define the objective set up time.
Capable: 17.78 ok
Available: 100%
Flexible: 0,22 min are the one that on average each operator should spend to do his set up. So from
1,1 for each operator we should pass to 0,22 as improvement proposed to work with family one in
one line.
FAMILY 2: we have to assign to each operator a workload equal to the takt time.
o With this approach, the ideal one, we need only 8 and not 10 operators and the last one works
only for 14 min (7 operators full saturated, the last one no)
o C/O time is not equal to zero: our aim is to reduce it, but if at the beginning we have C/O we
cannot simply put it equal to zero, but we need to understand what is the actual workload that
can be assigned to each worker and the remaining part should be assigned to the set up. What
we do is a continuous improvement approach, to reach the goal only at the end of some
different phases. We use the LEVELLING APPROACH to level the workload between operators.
o Then, we move to the DECAF CONDITIONS: capable (17,5 < 18), availability = 100% and flexibility
is measured according to the value of TOT C/O, which value is equal to 0,5 min. It is possible to
work with EPE = 1 day!

In case we were able to bring the change-over times of the two lines below the defined maximum time:
Material handling simplifies (shift to one point for each line) ... Maybe one handler can also be enough.
The space required in the factory would be reduced (less space for buffers).
The planning and control of production and coordination activities will drastically simplify, as well as the
work of operators (free time to resources, increase labor productivity).
Productivity improves by savings three operators of the line.

4. Where is required to use a supermarket-pull system?

If we were to use a supermarket, the sizing would be:
Supermarket: 1,5 EPE* Ddd [formula only for the exam]
The supermarket is necessary in the following cases:
Department upstream and downstream make setups with different criteria and the combination of
variants is high enough to require reductions in setup time that are not possible yet. (FOR EXAM PUT
SUPERMARKET if setup criteria are different and the number of variants is very high)
Big batches: if the stage that is upstream, like the first one, and the one that is downstream, like the
second one, are doing set up with different logic and high number of variants, it is better to put
supermarket in between. WHY? The actual constraint to couple or decouple stages is the flexibility of the
stages (set up). If I am able to change set up logic as I want, I can couple!
Set up logic: for example, one stage does set up according to length and the other one according to
colours. In this case I can decouple stages in order to avoid blocks. If stages do set up according to
the same logic, they are more synchronised, so it is more easy to couple them
Number of variants: if the number of variants (so the number of set up) is very high and the system
is coupled, the system will not work and stop many times (if one stage stops the others have to
wait). Only with small number of variants is reasonable to couple the stages.
Or
Upstream stage is not dedicated
Dedicated: if the stage is not dedicated, I have to put supermarket because I cannot work with FIFO
logic. In this way I can decouple the flow.
 lOMoARcPSD

5. What only point in the production chain (the pacemaker process) does the company have to plan?
The company plans the assembly stage (single stage of transformation)
6. How should the company level the product mix to pacemaker process?
Demand is leveled day by day. Within the day you level with a Heijunka Box. With regard to the mix, having
separate lines, a line will only receive products of the same family.
7. What should be the increase of work to release in process pacemaker? (not requested) PITCH
8. What improvements to the processes will be necessary to obtain the flow of the value stream described
by Future State? (not requested) DRAWING
6/6/16
EXERCISE 2

There are 4 stages. For standard products we have 4 variants. From stage 2 on, products are actually customised: we
need to do set up when we switch from one product to another.
Stage 2, 3 and 4: standard products with 4 variants are produced together with customised products (one
different from the other, infinite variants and possible combinations). Set up is done when we pass from
standard to customised products.
Stage 1: there are basic working processes. Customised products are given by standard products with some
variants, that are not the same of standard products, but they are the basis and specialised for customised
products. There are 4 variants for standardised products (they are always the same) and 5 variants for
customised products.
The technological basis of the two categories is the same.
Available time: 8 net hours
Special products demand: 15 products per day
Standard products demand: 85 products per day
Standard products variants: 4
Special products variants: each one is a different
variant Production stages decoupled by stocks

1. EPE of stages?
Stage 1: there is only variance. Slide 4 and 5
Stage 2, 3 and 4: the priority is for customised products, because with customised products I can
surely work with orders, so I am sure to work due to the always presence of orders (priority is for
guaranteed orders). With standardised products I am not sure to always work, because they are
related to the forecast. After this I calculate EPE for standardised products, while for customised
products it does not make sense because they have unlimited variations.
EPE is calculated with the available time for standard products (calculated subtracting
special processing time and set up time from the available time)
EPE is calculated for each stage and we pick the worst one.
Min batch size? 1 with no calculations because from stage 2 products are customised.
Other method:
starting from stage 3. EPE is 4 and my range is composed of 4 variants of products (for
standard ones). We have to divide EPE by the number of variants: (4/4 = 1). For standardised
products is: (4/4)*85 = 85.
Stage 4: (0.31/4) * 85 = 6.6
Stage 1: (1.72/5) * 15 = 5.16

FUTURE STATE
1. Takt time for the first 3 stages that work with 2 shifts and the the last stage has a different takt time because
Works with 3 shifts. IT IS POSSIBLE TO HAVE DIFFERENT TAKT TIMES!
2. I can produce for supermarket vedi slides
3. Continuous flow: general rules, suggestions:
With different shifts, stages cannot be put together in a cell
Stages not dedicated cannot be put together
 lOMoARcPSD

Stages that work with different set up logics and different variants cannot be put together in a cell.
They are decoupled with SUPERMARKET or FIFO logic, usually supermarket.
Then we do DECAF analysis: the only possible stages are stage 2 and 3. Let’s see if they can be put together.
o We have to do the weighted average of the C/T: (7*85 + 11*15)/100
o Availability is the multiplication of all the variabilities
o Flexibility? Two conditions (according to the test):
You must guarantee to be able to produce 15 units per day
EPE = 4 days
We try to calculate the EPE, but there is a big problem. I set EPE equal to 4: 4 times the time of production
and 4 times … . The C/O average final value is 1,18, but we do not know the right value for the unknown
variable of the standard and customised products.

4. Supermarket? There are two different logics:

o Standard products I am working with MTS: at the end I am working with a supermarket. Upstream a
supermarket we can have only another supermarket. HUGE MISTAKE TO PUT FIFO LINE
UPSTREAM!
The pacemaker is placed in the shipping phase
o Customised products I am working with MTO:
o Stage 1 is not dedicated so downstream stage 1 there is a supermarket.
o Stage 2+3: in the cell we will place the pacemaker
o Stage 4: FIFO logic
With no constraints, the objective is to put the pacemaker soon after the last supermarket.
For sure, downstream the pacemaker we can have only the FIFO logic.
Drawing the future state putting the two approaches together. VEDI FOGLIO

CUSTOMISED
PSP is used to manage the variability of the orders: in order not create disbalances in the pull line, I release orders
according to the work load of the stage (line balancing). PSP is not FIFO logic.
Variability higher than 30/40%: we have to dimension the pre shop pull in order to manage the variability of
different orders. The dimension is calculated according to a thumb rule: 30 products 2 days.
EXAM: we have to put PSP only where there is the pacemaker and when variability is higher than 30/40, we
dimension it (2 days) and the throughput time is given as 2 times the dimension (another thumb rule) (4 days of
production as throughput time of PSP).

C’è ALL’ESAME !!!!!

VSM-3
EXERCISE 1 (About set up)

We have 40 different products and 30 of them with a particular size and 10 with a smaller size. There is not only the
size problem, but there is also only one dimension to do the set up. Each variant of product has a different shape, so
we have two logics to do the set up: size and shape. According to the different stage, each stage can have both logics
and behaviour in term of set up, according to the process and the machines present in the stage.
The shape changes in stage 1, 2 and 3: each time there is a change in variant (40 times) we have set up (in these 3
stages). In addition, stage 3 does set up when the size changes.
See first slides.

1) EPE of each stage? The most critical?

For EPE calculation, there are no difficulties in stage 1 and 2. In stage 3, the approach is the same, but from the text
there is an indication about the stages: some are manual and other not. In this case we are not coupling anything (as
in the other exercise), but we calculate EPE for each stage. One operator is working and there isn’t any parallel set
up logic: in this case we do the shape set up with its logic and then the size set up with its own logic.
If one operator is doing the whole work from the beginning till the end, it is like a machine, so we have to calculate
like this:
 lOMoARcPSD

One machine does set up with one operator:

SHAPE (5min for each change of shape 40 shapes*5 min*60)
SIZE (we have 2 categories, small and big: for this stage we do also set up for this. 20 min* 2 sizes*60)
We sum them

2) Coupled system, one-piece flow. Min batch?

Worst cycle time, A is the result of multiplication and for the set up time we can apply the parallel activity approach:
Different machines have to do set up at the same time and in this way we can see what is the worst time in the line.

Whole product range: 40 variants.

30 big and 10 small:
1 and 2 stages do set up only for size, in parallel.
3 stage has a different logic: one set up of 20 min and another one of 5 min 25 min

25+29 set*20 min + 25 + 9set*20min = 810

Average=20,25 min/var used to calculate the min batch!

MIN BATCH: here we put the takt time (and not the available time) so we need to be coherent, considering the cycle
time divided by the availability without multiplication by the number of products produced daily.

X=it is directly the pieces per batch

3) Interventions?
A = 100% , one piece flow EPE = 1 day is our goal, so we need to define what to change in the set up!

EPE = 1: C/O = 4 min

This is not the final solution of the question, because stage 3 does set up on 2 variables
 lOMoARcPSD

STARM INDUSTRIES

Drawing (from right to left):

Ta = 900*60=54000 s/d
Stages:
o In case of processes in parallel, identify the critical path and use that time in the preparation of
the timeline. In Starm S.p.A. the critical path is represented by the stage of CUTTING and buffer rods
cut. Consequently, you must use 15 seconds for Value Added time and 5 days for Not Value Added
time.

To calculate the worst path, we need to consider that when a product is composed by 2 or more identical elements
the added value time is the sum of times to process all the components. For each raw material cut, we need two
pieces of raw material to be machined to prepare the fitting. In Starm S.p.A., machining added value time is equal to
40 seconds (2 forgings/product *20 seconds/forging).
The calculation of the critical path is: transform units in days.
CUTTING:
o stocks upstream the stage are transformed in days
o stocks downstream the stage are transformed in days
o second of the C/T needed to work in the stage are transformed in days
MACHINING: ALL TWICE!
o stocks upstream the stage are transformed in days
o stocks downstream the stage are transformed in days
o second of the C/T needed to work in the stage are transformed in days
 lOMoARcPSD

To consider the critical paths:

Cutting:
24.000[pcs]/1.200 [pcs/day] + 15[s]/(15[h/day]*0,5*3.600[s/h]) + 6.000[pcs]/1.200[pcs/day] ≈ 25 days
Mechanical machining:
48.000[pcs]/2.400 [pcs/day] + 40[s]/(15[h/day]*3.600[s/h]) + 4.800[pcs]/2.400[pcs/day] ≈ 22 days

NB) CUSTOMER REQUEST: # pieces/day = 24.000 [pcs/month] /20 [days/month] = 1200 [pcs/day]

The critical path to consider is that passing by the stage of cutting.

o MACHINING: the first fitting is welded 1 while the second one is welded 2. Half of the pieces go
directly to Weld1 and the others to Weld2. We consider the total amount of stocks all concentrated
in the same place downstream the machining stage, no matter where they are going then.
o PAINTING: it is outsourced, so there are inventories both in the painting and the Starm WH. In the
description below we put only the daily delivery, while the LT is used in the Timeline
o ASSEMBLY: there are 6 people who individually assembly alone the product from the beginning till
the end of the process. Each operator assemblies a product in 215 seconds on average. Another
thing is the cycle time of the whole stage, which is on average is 215/6. In the table there are both.
In the timeline for AV I have to put the throughput time perspective of each single operator: 215 sec,
while for LT there is the info in the text (LT = 2 days).
o DEFLASH: in the first 10 seconds an operator is loading the machine and the following 20 seconds
the machine works and unloads the pieces. The operator each 10 seconds finishes an activity and
can start an operation. The total is 30, so we put the total in the timelines, because a piece takes 10
sec to be loaded and 20 to be worked, so in 30 seconds it enters and exits the stage.

WE KNOW THAT the machine produces bars, and the streams are welded on the central and they are called fittings:
fitting 1 and fitting 2 can be different, they are not coupled.
20 different lengths, 2 diameters and 3 types of fitting (not always coupled, can be different): total 120 possible
variants, but we have 2 fittings so the total variant for this product is 240.

AV: ADDED VALUE TIME

LT: NON VALUE ADDED ACTIVITIES

FUTURE STATE MAP

1. Takt time:

2. Supermarket or shipping?
We do not always have all the information on all
drivers.
o We know that customers allow you to produce with a LT of 60 days even if we can produce with LT
of 48 days. We use leans to gain a competitive advantage on the market and save money inside
(decreasing costs – operational objective - and increasing revenues thanks to new market share)
We can produce with no problem with MTO
 lOMoARcPSD

3. Continuous flow?

Cutting and machining are not dedicated, so the options available are the three highlighted.

W1 + W2 + DEFLASH: each machine can work on average 20 sec and operators

10 sec. What if a single operator can do the work for all the machines? We have
a rigid sequence that is the one of machines, but we can put all the machines in
a cell and only 1 operator prepares the machines, each in 10 sec (10 sec the
first machine, 10 sec the second one, 10 sec for the last one = total is 30 sec).
So the operator is able to load the whole cell in 30 sec, which is lower than takt
time.
Freeing the operator from the single machine:
You can organize a cell in U and manage it to a single operator that moves between machines.
o C/T Operator: 10 sec/pcs for machine 30 sec/pcs for 3 machines < TT
o C/T machine: 30 sec/pcs < TT

DECAF analysis:
1. Machines are all dedicated
2. Capable: 30 < 45. C/T of the operator in the cell compared with the takt time of the whole system.
3. Available:
o Connect stages also means linking the availability! You need to check if the availability of the stages is a
restriction to the possibility of making flow: (C/T) / A < TT?
30 / (0,9*0,8) < 45 the machine availability is not a problem.
4. Flexible:
o You can work with an allotment of 50 average products? We work for shipping and must be able to be
flexible enough to meet the customer's request. The client requires 24 different versions (1200/50) of
product per day on 240 available in the catalogue.
CUSTOMER REQUIREMENTS: the customer order range is 25-200 pieces, with 50 pieces on average.
The daily request is 1200/50 = 24 (10% of the whole range)
NB)
o when we calculate the EPE we must consider the whole product range (240 variances) seen by the whole
stage
o when we have to propose improvement for different stages in lean perspective, we do not consider all the
variances, but just the number of variances requested effectively every day by the client. The number is
equal or lower than the total one.

o In this cell it is held set-up only for the length and the diameter so the range that the cell “sees” is 40 (and
not 240), i.e. in the cell there is no set-up for welding different types of fittings.
So of the 24 variants requested, probably only a part depends on the diameter and length.
o Example:
Product requested by Client1: D1 L1 GF SM (diameter 1, length1, great fitting, small fitting)
Product requested by Client2: D1 L1 GF MF (diameter 1, length1, great fitting, medium fitting)
 lOMoARcPSD

They are two different products, but in the cell is not expected to set-up to switch between them. So to
make sure that the cell is flexible does not mean that you should necessarily take 24 set-ups a day.
For sizing we consider, however, the WORST CASE, where 24 set - up a day in the cell are required.
Even if the total range of 40 is used for EPE, the range requested is 24 and the worst case is 24 variants
for the set up.

o In the current situation, without changing any parameter, it is possible to carry out 24 set-up every day
(situation corresponding to work with batches of average size 50 products)?
Available time every day to do set-up: t. available – t. production
15 h/day – [ 30/0.72 * 1200 ]/3600 h/day = 1.11 h/day
In 1.11 hours it is not possible to do 24 set-up. We need to do some improvements to be implemented in the
cell!

PROPOSED CONFIGURATIONS:

C/T operator = 30 < 45. Also, the time required to set-up does not allow the production at the Takt Time: the cell
must produce faster in order to accumulate the production for the time in which it is stopped for the change-over.
Some of the time should be dedicated to the set up We should increase as much as possible the production rate,
avoiding the worker to wait, in order to have time to do set up. If he waits he is not able to do set up, because all
machines would stop (coupled system).
SOLUTION: the operator goes ahead in loading the machines, instead of waiting till the end of the takt time,
accumulating production and having time to do set up at the end.
PROPOSED CONFIGURATION (a regime): C/T = 30 sec instead of 45 of the takt time, to have time to be dedicated to
do set up.

In this way we can save 2 resources (labour), but it is necessary to reduce non-value added activities (set-up).

Reduction of set-up time

T. C/O + C/T * daily demand ≤ 15 h
T. C/O ≤ 15 h – 30/0,72 * 1200 / 3600 = 1,11 h/day

With the objective to work with batches of 50 products and considering a daily demand of 1200 products, the
number of set-up required daily would be equal to 1200/50 = 24 set-up. (WORST CASE, see slide 14).
With this objective, the average time for the set-up on each machine should be equal to: 1,11 h * 60 min/h / 24 s-up
= 2,78 min.
This means that each machine can stop to do one change-over at most 2,78 minutes.
Having one single operator, this is the maximum time that he has to set-up the whole line. If he can do the set-up for
the whole cell in less than 2,78 minutes, then just one operator is needed.
If you can not reach this level, or you require the use of multiple operators:
or you must do a lower number of set-up, or you have to increase the availability of the cell.
 lOMoARcPSD

If we are able to set-up the cell in less than 2,78 minutes, we can get:
- A saving of 6 days in lead time
- A savings of 2 operators that can be assigned to other value-added activities
At the cost of just:
- Reducing non-value added activities

9/6/16
Implementing the flow is not only among different stations but also within a station.
ASSEMBLY:
First of all, we start with some considerations about set up time because we have a problem: in this case the set up
situation is different. At this stage each assembler assembles completely a finished product.
C/T for each operator is 215 sec.
Average order: 50 pcs.
With an average order of 50 pcs, they are required 1200/50 = 24 set-up.
On average they are spread over the 6 operators that work completely finished products. For this reason, each
operator must carry daily 24/6 = 4 set-up.
The point is that, if we decide to implement the line, each operator has to do 24 set up and no more 4. This is the
negative side effect! How to manage the set up now?
FIRST STEP: First, we calculate the time required by each operator to do the set up. Calculation: C/O is 10
min for each variant, 4 set ups per operator (4*10) and the production time is given by daily demand
(1200/6) multiplied by C/T/A.
The total time required daily to each operator is then: 1200/6 * 215/60 + 4 * 10 = 717 + 40 = 757 min =
12,62 h

To put continuous flow in the final assembly it is necessary to put in line the operators.
Benefits: The line configuration has many benefits, including greater efficiency (it is easier to manage) and a
lower supply required (this way, material to supply 6 islands is not always needed).
Disadvantages: The line configuration, however, requires a larger amount of set-up. In particular, each
operator on the line should run 24 production changes every day.
With the current set-up times the final assembly can not be put in flow. We must act on set-up times in order to
reduce them drastically!

NB) EPE calculation with more than 1 operator: or the reference point is the single operator (1200/6) or the whole
stage (see tractors). The best way is to use the single operator!!!
Remember that 15 hours is the opening time of the plant. There are no capacity problems!

SECOND STEP: In the stage of final assembly, C/T is 215 seconds for the finished product. The Takt time is 45
seconds/finished product. You may calculate the minimum number of operators required.
Content of work / Takt time = minimum number of workers
215 sec/pcs / 45 sec/pcs = 4,77 operators 5 operators
In the current situation, with 6 operators working in 6 islands assembly, you can not work with 5 operators
working in 5 islands. In this situation, each worker should do 24/5 = 4.8 set-up on average.
The total time required would be: 215*(1200/5)/60 + 4,8 * 10 = 860 + 48 = 908 min = 15,13 h
The situation changes if we pass to a line situation. Maybe not all 6 operators are necessary, so the
general approach is: we can calculate the WKC for a single product (215) and we divide it by TT (Takt
time = 45). In this way we get a decimal number (4.77) which is the min number of operators we
need to maintain the rhythm of the TT. We need to approximate to the higher unit (5).
0.23 operators is a waste in somehow and the final objective not to have any kind of waste, so we
need to use them to do the set up: this is the target for our set up improvement.
The average set up is 4.8 and we do the same calculation to calculate the total time required (15.13
h). THE SITUATION IS WORSE (15.13 is higher) and we are not able to meet the demand. We need to
find further improvements
THIRD STEP: In the current situation, even putting the continuous flow in the final assembly, it would not be
possible to work with 5 operators. Putting in line the operators means to allocate to each operator a
package of activities with a cycle time of less than but close to takt time.
 lOMoARcPSD

The possibility of putting the continuous flow in the final assembly and the possibility to save an operator is
in any case subject to the possibility to reduce the set-up time.
We assume that activities are formed by uniform 'packages' of activities and therefore we load each
operator as follows: 215/5 = 43 sec. All five workers will do 43 seconds of work on the single piece.
43 < 45 (TT), so 2 sec are used by workers to increase the production and to do the set up (2 sec are
collected by each product worked).
FOURTH STEP: In the situation described, what it is the minimum target of set-up time reduction for each
assembly station?
t. production + t. set-up ≤ t. available
43/1 * 1200 + 24 * C/O unitary ≤ 15 h * 3600 sec/h
52800 + 24 * C/O unitary ≤ 54000
C/O unitary ≤ 100 sec 1,67 min (total time the ENTIRE LINE will be blocked for set-up)

If it is possible to reduce the set-up time for each station to less than 1,67 min, then it will be possible to
save one operator (6 to 5), who may do different tasks.

PAINTING: Supplier takes arms unpainted once a day and returns them painted once a day.
In order to absorb the variability in the rates of withdrawal and deposit of the products in the two stages, and
prevent blocking events of the stages, Starm must decouple the processes using FIFO queues.
FIFO queues allow a constant supply to the stage downstream of the painting and the ability to the stage
upstream of the painting not to be blocked uselessly.
The use of FIFO queues allows to avoid overproduction in cases of problems in the production process and
allows a constant movement of the material!

Until now we used OPF, while now we use FIFO queues. It is not the only solution available, but considering that we
are working in MTO perspective without a supermarket at the end (we want to avoid supermarket as much as
possible, going to the most lean system). The pacemaker could be on the cell, according to experience, so the best
way to let the flow go ahed is to implement FIFO line. With FIFO line we have any particular type of scheduling. From
the cell products are put in a FIFO line, and they are pick up in the order they are received by the painting and 2 days
after they are put again in the FIFO. The adv is to avoid problems in term of:
- Scheduling
- Material supply
Each time the cell produces something, products are accumulated and with the rhythm of the cell the painting stage
work products.

In order to prevent events of starving and blocking of the stages, Starm must decouple the processes using FIFO
queues dimensioned at least on a production day.
For FIFO queues dimensioning:
1. The supplier may have some variability in the times for picking up/ delivering the goods.
2. Production volumes may fluctuate daily. It is not always possible to produce 1200 products per day.
We should slightly over-dimension FIFO queues, for example on 1.5-2 days of production, in order not to lose
production and output. The corrective factor is between 1,5 and 2 and the formula is the same of the supermarket.

One scheduling point (cell)

We need to guarantee that there is no bottleneck! In this case there isn’t!
 lOMoARcPSD

4. Pull-supermarket?
If there are two different set up logics and variant is high, we have large production loss
The only places are:
o Raw materials inventories: require large lots from suppliers, beams warehouse is shared.
o After cutting: it is a shared resource
o After machining:
Different setup logics: set up on the length and diameter and in the other stage (welding)
according to fittings.
Machining also has a very high C/O, while the other stage has lower C/O.
Considerable high number of variants.
It requires large productive lots.

CUT STAGE
The stage of cutting is shared between different families productive.
All production flows starting from the stage of cutting, must be free to take what they want when they want, even
when the cut stage is working on another product family.
The stage of cutting can not therefore be put in flow with the downstream stages.
• At the stage of cutting, for the family of the steering arms, there are 20 different lengths and two different
diameters. Each of the 20 lengths of the bar can in fact be of two different diameter dimensions.
• The machine for cutting works on the family of steering arms for 50% of its total available time.
• The total time of set-up required for re-assort the whole production range is equal to:
2 diameters * 60 min + 2 * 20 lengths * 15 min = 12 h

The reduction of EPE allows increased flexibility of the machine, allowing a reduction of the needed stocks
downstream, thanks to the ability to resort the range much faster.
The reduction of EPE is pursued by means of the goal of reducing the set- up time.
A progressive improvement in the time to set-up allows a constant increase in flexibility and a reduction of
the stocks stored downstream.

MACHINING STAGE
2 hours spent for each set-up, the stage of machining requires very large productive lots. This production stage
works three different forged fittings. With current technical parameters, the stadium has to do batching.

In the current situation, batches of about 4 days are

required, i.e. it is necessary to aggregate the average
demand of about 4 days of production for each product
code.

What said for the stage of cutting is valid also for the stage of machining.
Reducing the set-up time will increase the flexibility and reduce stock levels downstream.

RAW MATERIALS WH
Stocks: 20 days.
Shipping frequency of the supplier: 2 times a month.
LEAN SUPPLY CHAIN: we extend the lean approach also to other players of the supply chain, like suppliers. We can
establish a Kanban system based on supermarket with them, so the electronic Kanban is not inside the system, but
external. In this way it is a PULL SUPPLY, not based on a precise scheduling, in order to reduce the throughput time.
 lOMoARcPSD

Acting on the supplier can increase the frequency of delivery,

reaching (for example) weekly deliveries.
The stock could be organized with supermarket and
controlled by kanbans that will be transferred to the control
of production. This last will send orders to suppliers at fixed
time intervals.

Where we put supermarket, we always MUST propose improvement objectives in the stage upstream, because
the supermarket absorbs the inefficiencies, decoupling, INCREASING THE FLEXIBILITY AND REDUCING THE NEED OF
STOCKS DOWNSTREAM.

THUMB RULE:
Reasonable improvement.
Initial EPE >= 5 days First objective for the improvement is EPE = 2 days
Initial EPE < 3 we can propose EPE = 1 day

SUMMARY OF EPE:
Cut stage: EPE = 5 days.
Machining stage: EPE = (3.6) 4 days (C/O: 3 set ups of 2 hours each)
Raw material WH: we know that we have initial inventory in term of days (20 days). This is the problem, so
with supermarket, we can ask suppliers to improve their supply (their frequency) in order to reduce the
number of stocks in the inventory. We do this with a Kanban system!
EXAM: just let know the concept of increasing the frequency doing some proposal (EX: twice per month
1,2 times per week. Then the amount of stocks will change!)

5. Pacemaker?
Where, in the process, Starm must work with customer orders?
The weld cell+ deflash cell needs to work with customer orders since:
o At this point the customer can choose one of 240 different variants of finished product.
o The mover person picking the material from supermarkets must have an indication of what is
necessary to take.

System configuration at this time. We put the star “Set up” to identify where you have proposed improvements in
kaizen perspective.

SUPERMARKET DIMENSIONING

CUT STAGE
The stage of cut is shared among the various production families and therefore must have a supermarket
downstream. At the stage of cutting, for the family of the steering arms, there are 20 different lengths and two
different diameters. The machine works on the family of steering arms for 50% of its total available time.
• Now the EPE on the machine is 5 days.
To respect the time available is therefore necessary to batch five days of production on the stage of cutting.
To reduce the stocks of rods cut and increase the flexibility of the stage we need to reduce the set-up time.

A first goal of improvement might be to obtain an EPE of 2 days, given the high impact of the set-up in the current
situation. Daily, the time allowed to set-up is:
C/O + (C/T)/A * daily demand * EPE ≤ (50% * 15 h/day) * EPE
C/O ≤ (50% * 15 h/day) * 2 – 15 * 1200 * 2/ 3600 = 5 h
If the goal is EPE = 2 days, the time of C/O target per product would be: 5 h * 60 min / 40 set-up = 7,5 min
In 7,5 minutes, the operator must be able to prepare the machine for cutting for any of the 40 versions of the rods.

To size the supermarket, you need to think about the cycle stocks and safety stocks.
In the current situation, Starm stocks with EPE = 5 days 6000 products (1200 * 5) for 40 variants of rods.
With EPE = 2 days, you will have two days of stocks (2400 products) as cycle stocks.
 lOMoARcPSD

In addition to these cycle stocks, there will be also safety stocks, for example equal to 1200 products (half EPE).

The cutting stage will be managed by means of KANBAN SIGNAL. Withdrawals from the pacemaker will accumulate
kanbans in a special table. When the accumulation of kanbans will reach a threshold level, the stage of cutting will
produce the complete lot of products taken.
The reachment of the threshold of the number of kanbans taken will activate a signal, which indicates to the cut
stage to start the production.
The objective on the reduction of EPE improves considerably the size of the store of rods cut and activates a process
of continuous improvement that aims to raise the regularity of production on the stage of cutting.

MACHINING STAGE
We can assume to have an initial objective of EPE = 1 day for machining.
In the downstream there are different set up logics. The second stage many set up and the first one lower set
up, but the EPE = 1 day, so the batches in this stage are bigger! I cannot produce with a single Kanban, because
the logic is different.
In the supermarket they collect a Kanban to understand how the consume is working, when they reach a
threshold they start the production, due to bigger batches.

System configuration at this time. We put the star “Set up” to identify where you have proposed improvements in
kaizen perspective.

COMAPRISON
 lOMoARcPSD

EXAM EXAMPLES:
- CURRENT STATE drawing
- Different questions to produce the future state (mandatory the first 5, 6 and 7 are for theory, 8 is the
drawing that must be done!!!)
o Future state drawing, identify the improvements with the stars and putting all the symbols
o All the dimensioning (supermarket, fifo, set up) Numbers!!!! (audio 1.08)
o Motivate the choices with 1 sentence

NB) ANSWER 6:
XOXO is levelling with Heijunka box, using both volumes and mix
Quadrato con 6 caselle is levelling
60 A: 3
20 B: 1
20 C: 1
A
B
C

ANSWER 7: PITCH
Takt time is a theoretical approach, but in practice it gives the info about the production of 2 products.

Box with 5 products: BATCH

PITCH TIME: is the real tythem used in order to release the info to the pacemaker (heijunka) = 45 * 5 = 225 sec

Since in the Heijunka box there is one column per each pitch, with on average 20 products, every 225 sec the worker
goes to the column and waits that 20 products are ready, even if the batch is of 5 products.
 lOMoARcPSD

30/5/16
LEAN TOOLS
Lean tools are tools to implement the improvements that are going to be applied in the future state map.
The first tool is related to the analysis of the products that we need to consider in order to make the analysis of the
current state map. In the current state we should focus on the more important family of products, where the highest
part of the value is created. To do this, there are some methodologies that we can use.

Identify a products/service family

• Single product: if volume is high enough to dedicate resources to it. We run an ABC analysis and a Pareto
analysis to identify it.
o As a first attempt we focus on the most important products, then if our solution is good also for the
others we use it also for them, on the contrary we try to adjust it also for them. This because we
have to provide a solution that will be used by average people, so it must be simple.
Production Flow Analysis (PFA): if variety is high and single product volume low, there is need to group
different products together to reach enough volume to dedicate resources.
In this case we consider Family of products: formally they are different products, but if we compare them
they are quite similar between each other, so we can consider them as part of the same family.
o In the past, people were looking at the way to identify the family looking at the design and form of
products: when products were similar according to these drivers, they were part of the same family.
o After this period, people defined that even if some products have similar design, they do not
necessarily have the same production processes, so they belong to different families and have
different production flows.
The similarity comes from the production flow (machines and flows to manufacture the product)
and not from the design. PFA is used to do this.

PFA (Production Flow Analysis)

Product-Process matrix
The letters refer to the products and the numbers refer to the phases
and processes. They are not necessarily equal to the sequence of phases!
For now, we are not able to define similarities, because data is given
randomly. We can assume that D and E belong to the same family, but
for the others it is quite a mess.
We need to rearrange this matrix in order to detect easily if products
belong to the same family. There is another different matrix, with the
same content, but a different order of columns and rows. Products and machines are clustered in a different way:
there is a big family (A-G) where products use almost the same machines. Empty spaces are called VOIDS and
exceptions are called OUT OF LINE. We recognize also that maybe machine 3 is a BOTTLENECK: we can say this not
only because all the products have to pass through it (it is a bottleneck from a logistic pov), but we need to consider
also the time needed for the products to be worked and for this reason we do not know exactly if it is a bottleneck
(so we do not know if it is a bottleneck from a production pov).

RANK ORDER CLUSTERING (ROC: first tool)

How to go from the matrix on the left to matrix on the right? We have to follow a methodology. The algorithm is
called RANK ORDER CLUTERING (we cluster according to a ranking of each line and each column).
1) the product is visiting the machine
0) the product does not need that machine
We need to look at the sequence of numbers (0 and 1) as a binary digit number. We transform the sequence into a
number. EX) Starting from left for A: 27, 26, …, 20 multiplied by (1, 1, 0, …) = 200
 lOMoARcPSD

After having calculated the values for each row, we have to rank the values from the highest till the lowest (A, F have
the same value and their order does not matter). Then we have to do the same approach for the columns.
EX) Starting from left for 1: 25, 24, …, 20 multiplied by (1, 1, 0, …) = 48. After having calculated the values for each
column, we have to rank the values from the highest till the lowest.
We have to continue doing this process till we get a stable solution (no column and row change its position).
The solution is the following: the cells with 1 inside will be along the diagonal. Looking at the matrix, the result is that
we have 3 families, some exceptions (product C requires machine 2, not used for the 1 family, the same for product
E). The conclusion is the fact that we have 3 streams, so 3 independent systems (VSM), each with an independent
family.

ROC: it is one of the possible algorithms to solve the PFA (the problem that we are addressing)

Create flow
One of the principles to build the future state map is the creation of the flow. We do not like what is stopping the
flow and we need to remove them. What is stopping the flow?
• Production change: there is a production change when we do the set up. The problem is not the passage
from product A to B, but the fact that we need to prepare the machine and so we need time. The perfect
situation is without any loss of time.
• Wastes / lack of necessary information / re-entrant loops
• Breakdowns (stops of machines)
• Batch (Quantity)
• Batch (Time)
The ideal situation is lot size 1 and batch 1.

SMED
Setup times reduction: SMED (Single Minute Exchange to Die) used in Toyota.
The goal is to have one digit and not 1 minute. The approach is that every time we need to do something, we need
to identify the activities needed for the set up and we divide them in internal set up and external set up:
• INTERNAL set-up: done when the machine isn’t working
• EXTERNAL set-up: done while the machine is working
The external set up should be done before the machine stops. Internal set up should be moved to external, because
they stop the flow wasting time.

STEPS
1. Identify the activities necessary to the setup and classify them in external and internal.
In this phase, internal and external operations are often interspersed.
Examples of criticality
• equipment is brought at the machine in internal setup;
• tools are placed far from the machine;
• broken tools;
• failure to verify the correct functioning of the equipment, before staring the setup. Or Before starting
production
Do not analyse the procedures, but go at the Gemba (real behaviours): do not trust the people, workers,
staying in the office asking them what they do, because they never tell the truth or they do not remember.
Always go to the flow to look what they really do. Measure all you can, because the measure is not personal.
 lOMoARcPSD

2. Reorganize activities in order to have the greatest number of external activities grouped together at the
beginning of the setup, or at the end of it. Switch the largest number of activities (time) from internal to
external.
3. Simplify and shorten the duration of internal and external operations. Reduce the internal set up time as
much as possible, using tools and approaches.
Not Rocket Science, but continuous improvement of common operations. It may seem trivial, but it’s so difficult that
it creates a sustainable competitive advantage.

EX) Fastening system:

WASTES and REWORKS

Scrap, rework are also causes of stops in the flow and should be eliminated.
It was thought that a better quality leads to higher costs, it was discovered that a higher quality decreases costs.

The purchasing department is normally evaluated according to savings: the less they spend, the higher their
performances, even if the quality of components is lower, in average. Using these kind of components make
problems appear in the factory, so the incentive for the purchasing department (increasing his value) is not balanced
with the problems of the production process (interruption of the flow). There is a TRADE - OFF.

ELIMINATE PROCESS STOPS

The usable capacity is net of breakdowns, failures, and other stoppages for maintenance, etc.
Compared to the potential rate, the system can actually produce much less.

Availability: to improve it we have to increase MTBF (higher reliability of components higher availability
of the system) or we have to reduce the MTTR (we design a system easy to be repaired higher
availability)
The real/actual capacity of the machine is the speed multiplied by the total availability of the machine.

Total Productive Maintenance (TPM)

Methodologies which are moving as much as possible maintenance activities from failure based maintenance, to
scheduled maintenance, to predictive maintenance.
Failure maintenance approach: we wait and see. When the failure appears, we adjust it.
Predictive maintenance: when we do not use the system, we do some activities in order to increase the
availability of the system later on and reducing the chance of failure (without exaggerating).
Notice and act promptly on signs of malfunction and prevent failure.
In addition to generate a greater available capacity. It avoids unforeseen events that cause problems and
inefficiencies.
 lOMoARcPSD

ELIMINATE MIX BATCHING

Since we have a set up, we produce a higher number of products to distribute the set up time over a higher number
of products.
Batch processing: batch - set up - batch - …
AAAAA-setup-BBBBB-setup-CCCCC -setup-DDDDD

Mix model processing: try to mix as much as possible the sequence in order to follow the demand as much
as possible. All the models are mixed together, with any sequence suitable to create the continuous flow.
ABCADCDBACDABACADBACDADCAABDABB

LEVEL THE DEMAND

Market demand is often more levelled than you believe (80% of mix and volume variations are determined by the
company itself). The more demand is aggregated, the more it’s levelled (overall volume Vs single model;
geographical areas, time, ...).

EXAMPLE
We have 4 finished products and their weekly demand.
1. Starting point: this matrix (a kind of bill of material)
shows how many parts are needed for each product
(Part 1 is needed in 2 units for product 1). The average
service factor is the weighted average of component
parts that we have in the products
(2*75 + 0*50 + 1*25)/75+50+50+25 = 0.875

2. We take the weekly demand and we start from the first

product on Monday morning, we produce as much as
possible and we keep going with the production in the
next days, till reaching the weekly demand.
Supplier 1 has to deliver to us 80 units of part 1 and all
the other parts needed to produce product 1, according
to the coefficient. At the end we get the table with
what supplier 1 has to do.

3. We have a levelled demand, but the drawback is for the

suppliers: they have not a levelled demand, but peaks
and drops along the week. How can they implement a
lean approach? If we change the way we schedule or
planning, we can create a levelled production also for
suppliers.
Instead of producing 1 batch, each day we produce 1/5
of the weekly demand as an alternative scheduling.
In this way, also suppliers have perfectly levelled
demand.
This is only PARTIALLY TRUE. In this way we batch
along the day and not along the week. The
consequence for suppliers is the fact that between
morning and afternoon there is no levelled demand.
(1/4 in the morning and 3/4 in the afternoon, which is
balanced for me, but not for the supplier).
There is another option: instead of batching along the
day, we create a mixed model in the morning and
another one in the afternoon.

The system must see the variety that the market demands as it were a single product.
 lOMoARcPSD

ELIMINATE TIME BATCHING

Batching is a very tricky issue. TIME BATCHING is like going to the supermarket once a week instead of going to buy
food every day.
In many offices, and service activities, time batching is frequent:
To process invoices once a week
To process subscription requests every other day
To process design change requests for a product once every 15 days...

Link everything to the customer (Pull)

You should not produce if the customer has not made a request, otherwise you overproduce. Forecasts are always
wrong. To be efficient you have to work on reliable information. Market is not willing to wait. There’s a need of an
extremely fast system.

KANBAN SYSTEM

PULL SUPPLY CHAIN SEQUENTIAL PULL

PULL advantages:
• Simplicity
• Control over stocks
• No overproduction by the supplier
• An easy way to push towards constant improvements
• A single planning point, everything else just reacts
• Limited amplification of production and inventory across the supply chain (Bullwip effect)

Scientific method
Any intervention must be carried out following what is the
scientific method.

Lean and Readiness

 lOMoARcPSD

1/6/16
CREATING CONTINUOUS FLOW
It means putting in place all the possible actions to make a continuous flow, to avoid having the need of stocks,
because LT is the lowest possible and if it is equal to working time, stocks are equal to zero (Little’s Law).

Now we need to go to a higher level of detail, working at process level, looking at the performances of the machines
in order to try to reduce C/O (if we have machines with high C/O of course we will have stocks because we are not
able to have continuous flow).
Process level: How to design a cell?

EXAMPLE
• Apex makes tubular products for different applications:
automotive industries, trucks S, L, A, heavy trucks and
heavy equipment
• The Light Truck families are the most important ones
(those that we consider): S, L and A belong to the same
family of products, like heavy types.
• Current state map
We ALWAYS start reading it from the right, the demand (what we need to produce comes from the quantity
required by the market). If we do not understand the demand, we do no understand what the company is doing (the
factory is only a tool used to answer the demand).
Pieces are delivered using boxes with 30 pc each. There are 2 shifts and 4 deliveries for day. The shipping is done by
the shipping unit.
Before shipping there are all the phases and processes, that can be read from left to right:
• Tube extrusion: C/O of 1 hour is right because of the different dimensions of the tubes.
Since the schedule goes to each phase according to MRP, stocks are needed.

Where to put supermarket? At the end we have 3 types of products to be shipped in boxes with 30 products each,
with the kanban equal to 30 and there are different kanban for different products, so before shipping we need to
put supermarket for definition.
In this way all the stocks in these phases (from bending, with all assembly phases till the end) are eliminated and LT
is shorter. The cell is characterised by:
• 4 operators
• C/O 20 sec (the worst)
• LT 159 seconds (sum of values)
 lOMoARcPSD

NB) Do not pretend to do all the improvements in one shot, because the best way to do it is continuous
improvement.

This improvement is not enough

Final results are not as good as on paper
Redesign the value stream is good but it is
only the starting point
VSM is no solutions, only problems
highlight

Tube bender is a phase with an automatic

machine that needs to be manually loaded with
25 tubes at a time. One worker (worker 1) is
dedicated to assembly 1 and once in a while he
has to go to load also the bending machine: the
flow is interrupted, because if he goes to bending
machine in that moment no one is working in the
assembly 1. To avoid assembly 2 to be interrupted
or waiting, 7 pieces of inventories must be
present after assembly 1. The same is done by
worker number 4, who is dedicated to the tester machine but once in a while he has to go to the crimper to take
products to be worked.

APEX was working as in the current state map, but they decided to do some improvements. Firstly, they added cells.
This is what an external consultant sees after the last improvement: in some cases, the actual production is lower or
higher than the planned production. He has to investigate the reasons.
Some of the reasons are due to the fact that the continuous flow is not there. The sources of not continuous flow
are:
1. Workers have to move between different phases
2. Workers number 4 has to move to get materials to work. Having people walking is a waste
Cells need to be redesigned in order to put stations closer one to the other (1st improvement), to avoid walking and
moving material around. The 2nd improvement consists in changing the way the loading activity is done by worker 1.

Questions: Why there is so much variation?

Is the machine incapable (scrap/rework)?
Is it the machine that won’t run?
Are parts missing?
Is the supplier shipping defective parts?
Who reacts when these problems occur?
We should analyse SIGNALS, possible indicators of variation:
• First operator leaves his/her area every 25 pieces (flow stops)
• Inventories between operations
• Production associates are anchored to their machine
• Very wide U

Is there any better way to organize workers’ work?

We need to eliminate the differences between actual and plant values.
After the first revision, with the introduction of cells, the results are:
 lOMoARcPSD

TARGETS:
1. To produce exactly what is scheduled to be produced
(stability means continuous low)
2. Less workers, less space, less costs
3. We need to identify a pacemaker!

First of all, we need to consider the 3 different products.

We have 2 options:
• Option 1: creating ad hoc cells for each product
• Option 2: creating mixed model cells where the same cell is producing all the products. A and B normally are
better balanced, the flow is balanced and it is easier to create continuous flow.
For example, 2 products are different and maybe with C/T different, but mixing them it is easier to manage
them.

How to choose what items in the cell

• Flexibility: A+B more flexible than A and B
• Variation in total work content: Max around 30%: it is just a rule of thumb (if variation is higher than 30%
we risk to create work not balanced)
• Similarity in processing steps and equipment
• Takt Time: Optimal between 10 secs and 120 secs.
o If it is too low, it means that workers have to repeat the same sequence of operations in a very short
time (more than 6 times in a minute).
o If it is higher than 2 minutes, the drawback is related to a slower learning curve. It could be that
workers’ tasks have inefficiencies due to the fact that in 2 minutes repetition is not enough and
maybe the task is too complex.
• Customer Location: Close to the customer

TAKT TIME: Expected production rate to deliver the product to the customer. Takt time is a goal, a target.

Available time of workers: we need to take into account breaks.

Stoppages of machines due to failures should NOT be
included in the available time, because we do not want the
machine to stop, so we consider it as a non expected event.
It is not a constraint we need to consider!

Setting the pace

TT is based on customer demand (which you cannot change) and available production time (which you can change).
Your levers are:
• Number and type of items in a cell
• Available production time
• Number of cells making a specific item
 lOMoARcPSD

What if demand rate changes?

Takt Time refers to average demand. Changing TT frequently yields inefficiency.
In MTS use Buffer Stock to protect the production system
In MTO use Backlog to protect the production system

Cycling much faster than TT

• May require more people
• Increase the chances of over producing
• May conceal production problems
• Eases tension to improve

PAPER KAIZEN
How to remove all the possible reasons of changes in the production flow and the reasons of actual and expected
productions differences? We should use PAPER KAIZEN, going inside looking at process and activity details
(elementary activities performed by workers). We have to record all the work elements and the time needed to do
them. DO NOT TRUST PAPER, we need to go and measure them!

Cycle (wait): cycle of the machine that supports

the worker.
• Highlighting the machine cycle is used to
understand how much time the worker
has available to do something else

We should detect the activities done by workers and also the activities of the machine which is supporting the
worker in the process.
C/O is done only when we change the type of product, while the cycle (wait) is the sum of the time needed
to do the work of one product.

TIMING TIPS
• Collect real time at the processes (do not rely on standard time)
• Position yourself where you can see the operator’s hand motions
• Time each work element separately (going is an action, taking is another one, etc)
• Time several cycles of each work element (we calculate the mean of different shots)
• Observe an operator who is qualified to perform the job
• Always separate operator time and machine time (because when the machine is working, normally the
operator is not working)
• Select the lowest repeatable time for each element
• Remember shop floor courtesy (be kind)

OPERATOR BALANCE CHART (OBC)

Tool to see in a graphical way if the work assigned to workers is balanced or not.
It is based on the same list of work elements of the Paper Kaizen. On the left we put the station with the highest
value and the others in a decreasing order going right.
 lOMoARcPSD

Purpose:
• See if the work load is balanced between workers
• In each phase there is a gap between the takt time and the work load time, so we can still do improvements
(like adding activities to workers in order to reach the takt time for each of them and maybe not using one
worker levelling).

How much automation? We can automate some activities, but

not all of them.

Level of automation:
• Level 0: only human work
• Level 1: only the cycle of the machine is automated
• Level 2: also the unload of the machine is automated
(first this one, because the level of precision required
by the unload is less than the load, the positioning
requires more precision)
• Level 3: also the load of the machine
• Level 4: also the transfer of the machine
• Level 5: everything is fully automated (cycle, load,
unload, transfer)

How we use the analysis done till now? To improve the situation reducing the time of the workers, trying to:
• Eliminate wait while machine cycles (Operators wait for machine to cycle 21 s)
• Introduce auto-eject at Ass I, Ass II, Crimper, Tester (unclamp, remove, set aside): unload the machine in an
automatic way.
• Convert out-of-cycle work (loading the bender with a batch of 25 tubes) to in-cycle-work of loading one tube
every cycle
 lOMoARcPSD

Starting from the last Paper Kaizen we fill the table. Can your equipment meet the TT?
If not meeting TT, so if the cycle time is higher than the TT, we can select from the list below what improvement to
do:
• Reduce setup time
• Increase availability
• Kaizen the load, start, unload process
• Eliminate waste in the machine cycle itself
• Split apart some of the tasks a bottleneck machine is performing and use more than one machine (select
simpler machines)
• Install 2 machines of the same type in parallel
• Create 2 cells instead of one
Finally, if anything else fails:
• Remove the bottleneck equipment from the cell, decouple it and operate in batch.

TRADITIONAL APPROACH
The pieces are worked in a repeated sequence by w1 w2.
Operators work for the machine. We can revise this approach, trying to find
another solution.

Large vs Small machines

Large: economies of scale
Small:
Reliability
Volume flexibility
Mix flexibility
Trade-off between Machine
utilisation, Material utilisation,
Man utilisation. Human are the
most flexible
 lOMoARcPSD

MULTI-PROCESS HANDLING INSTEAD OF WAITING

One worker loads and unloads the stations while the machine is working, but for
example the first machine has to wait to restart working because the worker is
doing something else in the meanwhile in another station.

CELL LAY-OUT
Arrange machines, workstations, and material presentation devices as if only one operator makes the product from
beginning to end.
 lOMoARcPSD

Applying the guidelines to the APEX cell

Too much walking distance from beginning to end
The inside of the cell is too wide
The leadoff and final operations are far apart
There is ample space for WIP accumulation between machines
The tube bender needs modifications to load one tube at a time
The out tray for the bender obstructs the operator walking path

Dealing with batch-oriented equipment

Separate batch process from continuous flow with a supermarket or a FIFO lane
Keep in the continuous flow if equipement is moving-conveyor type (and operator can drop a single piece at
start, and pick up one at end within TT
Ignore equipment batch capability and use it single piece
Transform equipment from batch to single
Replace batch equipment with one or more single piece inexpensive equipment

Distributing the work

Number of operators

Guidelines for determining the number of operators in a cell

Work distribution among operators

Options:
SPLIT THE WORK
Splitting the work means that each operator is given one portion of the total work content.
In this case there are 3 workers and more stations than the workers. Stations are needed due to the special
equipment. This is a way to organize the layout, splitting the work among workers. Stations need to be close to
communicate and avoid waste of movement. This is a CELL.
 lOMoARcPSD

THE CIRCUIT
Each operator performs all the work elements. Single workers follow the whole
assembling process.
• Generally limited to 1-2 operators: if we want to improve the situation, we
add operators, with a number up to the number of stations in the line,
because with a higher number it would be impossible to move.
• Does not work if a single operation has more than 40% of TWC
• Require skilled operators: the draw back is the fact that operators need to be
more skilled, because they need to be trained to do everything.

REVERSE FLOW
Operators move in the opposite direction of the material flow.

COMBINATION
The criteria here is to balance the workload and keep the cycle time below the takt time.

How will the pacemaker react to changes in customer demand?

When designing cells, engineers should prepare one-up and one-down scenarios for responding to volume changes.
If the demand increases, the takt time decreases and then you need to move part of the work to another operator.