1

Chapter 3
Forecasting
2
Forecast
OM is mostly proactive not reactive
It involves structured planning activities
Planning requires data pertaining to the Ieature
Forecast: A statement about the Iuture
Not necessarily numerical
Weather Iorecasts

Accounting Cost/proIit estimates

Finance Cash Ilow and Iunding
Human Resources Hiring/recruiting/training
Marketing Pricing, promotion, strategy
MIS IT/IS systems, services
Operations Schedules, MRP, workloads
Product/service design New products and services
&ses of Forecasts

REMARKS
W ssume a causal system
Future resembles the past
W Forecasts rarely perfect because of randomness
W Forecasts more accurate for groups vs. individuals.
Forecasting errors among items in a group usually have a canceling
effect.
Extremes in a group cancel each other
W Ex. can forecast the class average from the midterm better than
Mrs. X's individual grade.
Sample variance of {-1,1,-1,1} is 1.
Sample variance of {(-1+1)/2, {(-1+1)/2} is 0.
W Forecast accuracy decreases as time horizon for forecasts increases
W Ex. can forecast this year's class average better than next year's
class average
Elements of a Good Forecast
TimeIy
Accurate
ReIiabIe
Written

Steps in the Forecasting Process

$tep 1 Determine purpose of forecast
$tep 2 EstabIish a time horizon
$tep 3 $eIect a forecasting technique
$tep 4 Gather and anaIyze data
$tep 5 Prepare the forecast
$tep 6 Monitor the forecast
"The forecast"

%ypes of Forecasts
W Judgmental - Subjective analysis of subjective inputs
W Associative models nalyzes historical data to reveal
relationships between (easily or in advance) observable
quantities and forecast quantities. &ses this relationship to
make predictions.
W Time series Objective analysis historical data assuming
the future will be like the past

udgmental Forecasts
W Executive opinions (long-range planning)
%here are factors hard to quantify
W Ex: Effects of November 2004 election on new houses built in 2005
W Sales force composite
Retailer forecasts for the manufacturer
W Consumer surveys
%he guy at the mall who asks if you like cherry flavor in your shampoo
W Outside opinion
Financial and consulting gurus and companies
W Opinions of managers and staff
Delphi method: series of questionnaires developed sequentially

ssociative Forecasting
Based on identiIication oI related variables that can be used
to predict values oI the variable oI interest.
Sales oI mountain bikes in an area may be related to the percentage
oI the young population living in that area.
Sales oI Harley-Davidson motorbikes is related to mid-aged men
population. Average age oI H-D owners is .
Ice cream sales can be related to temperature
Home depot bases sales Iorecasts on mortgage reIinancing rates,
smaller rates imply higher sales.
Changes in Federal Reserve Board`s interest rate leads to certain
business activities
House sales
Industrial investments
Increase in energy cost leads to price increases in products and
services
10
ssociative Forecasting
Find an association between the predictor and the
predicted
Predictor variables - used to predict values oI variable
interest, sometimes called independent variables
Predicted variable Dependent variable
Regression - technique Ior Iitting a line to a set oI
points
Linear regression is the most widely used Iorm oI
regression
The obiective is to obtain an equation oI a straight line that minimizes the
sum oI squared vertical deviations oI data points Irom the line.
11
inear Regression (cont.)
y a bx
Where
y predicted (dependent) variable
x predictor (independent) variable
b slope oI the line
a value oI y when x 0 (the height oI line
at the y intercept)
12
Computing a and b
Given n data points, Iind the intercept a and the slope b to
2
1 1
2
1 1 1

'
+

'

=


= =
= = =
n
t
t
n
t
t
n
t
n
t
n
t
t t t t
x x n
v x v x n
b

=

=
=
n
t
t t
bx a v
1
2
) ( Minimize
line the Irom deviations oI sum the Minimize
errors squared oI sum the Minimize
n
x
b
n
v
a
n
t
t
n
t
t
= =
=
1 1
1
inear Model Seems Reasonable
0
10
20
30
40
50
0 5 10 15 20 25
X Y
7 15
2 10
6 13
4 15
14 25
15 27
16 24
12 20
14 27
20 44
15 34
7 17
Computed
relationship
1
nother inear Regression Example
Variables: Weeks and Sales
t y
W e e k t
2
S a l e s t y
1 1 1 5 0 1 5 0
2 4 1 5 7 3 1 4
3 9 1 6 2 4 8 6
4 1 6 1 6 6 6 6 4
5 2 5 1 7 7 8 8 5
% t = 1 5 % t
2
= 5 5 %

y = 8 1 2 %

t y = 2 4 9 9
(
% t )
2
= 2 2 5
15
inear %rend Calculation
y = 143.5 + 6.3 t
Sales in week t = 143.5 + 6.3 t
a
12 - 6.3(15)
5

b
5 (2499) - 15(12)
5(55) - 225

12495-121
275-225
6.3
143.5
1
inear %rend Calculation
y 1.5 .t
When t 0, the value oI y is 1.5 and the
slope oI the line is .. meaning that the
value oI oI y will increase by . units Ior
each time period. II t 10, the Iorecast is
1.5 .(10) 20.5
Excel example
regression.xls
1
inear Regression
Remember from Statistics
Correlation (r) between variables: The strength and
direction oI relationships between two variables
1.00 means changes in one variable are always matched
by changes in the other, vice versa.
A correlation close to zero means little linear relationship
The square oI the correlation coeIIicient provides a
measure oI the percentage oI variability in the values oI y
that is explained by the independent variable.(0 or
more: the independent variable is a good predictor oI the
values oI dependent variable)
1
%ime series
%ime-ordered sequence of observations taken at regular
intervals over a period of time
Future values of the series can be estimated from past values.
Types oI Variations in Time Series Data
W Trend - long-term movement in data
W Seasonality - short-term regular variations in data
W Cycles wavelike variations of long-term
W Irregular variations - caused by unusual circumstances
W Random variations - caused by chance
1
Forecast Variations
Trend
IrreguIar
variation
CycIes
$easonaI variations
Year 1

99
Figure 3-1
CycIicaI
20
W &ses a single previous value of a time series as the basis
of a forecast.
W Virtually no cost
W Data analysis is nonexistent
W Easily understandable
W Cannot provide high accuracy
f it were true, future will always be the same as the past
Some notation: Forecast at time t is F(t)
Actual observation at time t is A(t)
Today is temperature is F, A(Today)
F(Tomorrow)
F(Day aIter)
Nave Forecasts
21
Stable time series data
Forecast is the same as the last actual observation
F(t) A(t-1)
Seasonal variations
Forecast is the same as the last actual observation when
we were in the same point in the cycle, where a cycle
lasts n periods.
F(t) A(t-n)
Data with trends
There is constant trend, the change Irom (t-2) to (t-1)
will be exactly as the change Irom (t-1) to (t)
F(t) A(t-1) (A(t-1) A(t-2))
&ses for Nave Forecasts
22
Naive Forecasts
&h. give me a minute....
We sold 250 wheels last
week.... Now. next
week we should sell....
2
Nave (Cont.)
Check iI the resulting accuracy is acceptable
The higher the accuracy, oIten the higher the cost.
Do we really need our Iorecast that accurate? Is it
worth the additional resources?
Why do you need Iorecasts Ior? How critical they are
Ior operations?
2
%ime Series Models: Variations
What is random and what is not?
Historical data contain random variations or noise
Random variations are caused by relatively
unimportant Iactors.
What is random? Can we not study everything to negligible
detail? 'God does not roll dices A.E.
The obiective is to remove all randomness and have
real variations.
Minor variations are random and large ones are real.
25
%echniques for veraging
Moving averages (MA)
Nave methods iust trace the actual data with a lag oI
one period, F(t)A(t-1)
They don`t smooth
MA uses a number oI the most recent actual data to
smooth
Weighted moving averages
Exponential smoothing
2
Simple Moving verage
Note the sensitivity of forecasts
35
37
39
41
43
45
47
1 2 3 4 5 6 7 8 9 10 11 12
ActuaI
MA(t,3)
MA(t,5)
ns observatio actual n using 1 - t period in made Iorecast MA :
,
,
1
,
n t
t
n t i
i
n t t
MA
n
A
MA F

=
= =
Averaging (over time) techniques are used to smooth variations in the data.
2
Ex: %hree period moving average forecast
Month Demand
1 2 MA(,) ( 0 1) /
2 0 1..
II A() , then
0 MA(,) (0 1 ) /
5 1 0.00

2
Weighted average
Moving Average
AdvantageEasy to compute and easy to
understand
DisadvantageAll values in the average are
weighted equally
Weighted Moving Average
Similar to moving average
It assigns more weight to the most recent values in
a time series
Idea: most recent observations must be better indicators
oI the Iuture than older observations
2
Weighted average
Compute a weighted average Iorecast using a
weight oI 0. Ior the most recent period, 0.
Ior the next most recent, 0.2 Ior the next and
0.1 Ior the next.
Continuing with the data on the leIt
F() .0(1).0(0).20().10(0)1.0
II the actual demand Ior period is ,
F() .0().0(1).20(0).10()0.2
The weighted average is more reIlective oI
the most recent occurrences.
Month Demand
1 2
2 0

0
5 1

0
Exponential Smoothing
Forecast todayForecast yesterday(alpha)*(Forecast error yesterday)
Each new Iorecast is equal to the previous Iorecast plus a percentage oI
the previous error.
Today`s Iorecast
Depends on yesterday`s (time-wise dependence, strong memory)
But it has to be corrected by Iorecast error
ThereIore, we should give more weight to the more recent time
periods when Iorecasting.
Alpha smoothing constant percentage oI the Iorecast error.
) (
1 1 1
+ =
t t t t
F A F F
Forecast error:Actual Forecast A(t-1)-F(t-1)
1
Exponential Smoothing
as an Weighted verage
dea--The most recent observations might have the
highest predictive value along with the most recent
Iorecast errors. Let us balance them:
1 1
) 1 (

+ =
t t t
F A F
1 t
A
1
) 1 (

t
F
t
F
2
Period ctual Forecast with Error with Forecast withError with
1 42 lpha=0.1 lpha=0.1 lpha=0.4 lpha=0.4
2 40 42 -2.00 42 -2
3 43 41.8 1.20 41.2 1.8
4 40 41.92 -1.92 41.92 -1.92
5 41 41.73 -0.73 41.15 -0.15
6 39 41.66 -2.66 41.09 -2.09
7 46 41.39 4.61 40.25 5.75
8 44 41.85 2.15 42.55 1.45
9 45 42.07 2.93 43.13 1.87
10 38 42.36 -4.36 43.88 -5.88
11 40 41.92 -1.92 41.53 -1.53
12 41.73 40.92
Example of Exponential Smoothing
Forecasts made in a period and the period has the same color
1 1
) 1 (

+ =
t t t
F A F

Picking a Smoothing Constant:

Responsiveness vs. Smoothing
The quickness oI Iorecast adiustment to error is determined by the
smoothing constant.
The closer the alpha is to zero, the slower the Iorecast will be to
adiust to Iorecast errors.
Conversely, the closer the value oI alpha is to 1.00, the greater the
responsiveness to the actual observations and the less the
smoothing
Select a smoothing constant that balances the beneIits oI
responding to real changes iI and when they occur.
1 1 1 1 1
) 1 ( ) (

+ = + =
t t t t t t
F A F A F F

Picking a Smoothing Constant

Sensitivity of Forecasts
35
40
45
50
1 2 3 4 5 6 7 8 9 10 11 12
Period
D
e
m
a
n
d
=.1
=.4
ActuaI
Excel example
exponential-smoothing.xls
5
%echniques for trend
Develop an equation that will describe trend
The trend component may be linear or it
may not
Linear trend:
Y
t
= a + bt
1 2 3 4 5
t
Y
b is similar to the slope.
However, since it is
calculated with the variability
of the data in mind, its
formulation is not as
straight-forward as our usual
notion of slope.

Common Nonlinear %rends

ParaboIic
ExponentiaI
Growth
Figure 3-5

Adiusting Ior Trend with Double

Exponential Smoothing
Simple exponential smoothing with no trend
Add Iorecasted trend
This time trend is also smoothed, note that
previous trend (oI t-1) and current trend (oI t)
appear in the smoothing Iormula:
See Table -2 Ior an exercise
= =
t t t t
T F A F + + =
+
1
1
t
T
= =
1 1
1 ) (

+ =
t t t t
T F F T . .
1 1
and


t t t
F F T

%echniques for seasonality

Regularly repeating upward or downward
movements in time series values
Seasonality: weather variations, vacations and
holidays
Seasonality: Expressed in terms oI the amount
that actual values deviate Irom the average
value oI the series
Seasonality is expressed as a percentage oI the
average amount
seasonal percentages seasonal relatives seasonal indices

Different models of seasonality

Seasonal relative .45 Ior the quantity oI television
sold in August at Circuit City, meaning that TV sales
Ior that month are 5 above the monthly average.
Seasonal Iactor0.60 Ior the number oI notebooks
sold at the UTD bookstore in April, meaning that
notebook sales are 40 below the monthly average.
Seasonal indices are your vehicle to travel between
the seasonal and deseasonal worlds.
0
Use Seasonality Indices
to Deseasonalize and Seasonalize
Deseasonalize historical observations
Divide them by seasonal indices
Make the analysis Generate Iorecasts
Seasonalize Iorecasts
Multiply them by seasonal indices
t t t
Inputs
Analyze Output
Excel example
seasonalforecast.xls
1
Forecast ccuracy
Measurement is the Iirst step to improve an
activity
What value oI smoothing constant is good?
Accuracy measurement is a vital aspect oI
Iorecasting
Impossible to correctly predict Iuture values
Important to include an indication oI how big the
Iorecast deviate Irom the actual values
2
Forecast ccuracy
W Error - difference between actual value and
predicted value
W Mean absolute deviation (MD)
verage absolute error (weights all errors evenly)
W Mean squared error (MSE)
verage of squared error (weights errors according
to their squared values)
W %racking signal
Ratio of cumulative error and MD

MD & MSE
error. Iorecast oI variance Ior the estimator unbiased the is MSE : says Statistics
MSE s deviation standard error) (Iorecast oI Estimate

) (
1
) (
' '

1
1
2
1
2
1
= =

<

=
=

=
= =
=
MAD
F A
Signal Tracking
n
F A
n
F A
MSE
n
F A
MAD
Forecast Actual error Forecast
n
t
t t
n
t
t t
n
t
t t
n
t
t t

&se for MD & MSE

Compare the accuracy oI alternative
forecasting methods using MAD and MSE.
parameter (such as alpha) values used in Iorecasting
by using MAD and MSE
Determine which method yields the lowest MAD
or MSE Ior a given set oI data.
5
Controlling the quality of forecast
Necessary to monitor Iorecast to ensure that the
Iorecast is perIorming adequately
This is accomplished by comparing Iorecast errors to
predetermined values
Errors that Iall within the limits are considered
acceptable
Errors outside either limit indicates that corrective
action is needed.
Tracking signal values are compared to predetermined limits (,-) based
on iudgment and experience
Upper and lower limits Ior individual Iorecast errors are calculated using
control chart techniques. We will learn about control charts in quality
chapters.

Choosing a forecasting technique

No single technique works best in every situation
No single technique works best in every situation
The Iorecast horizon
Forecasting Irequency
Forecasting is not Iree
Consider cost and accuracy
Weigh cost-accuracy trade-oIIs careIully
Forecast detail, part / product level?
Availability oI
historical data
computers
able users / decision makers

Choosing a forecasting technique (cont.)

Moving Averages and Exponential
Smoothing are short range techniques. They
produce Iorecast Ior the next period
Trend equations are used Ior much longer
time horizons.
More than one Iorecasting techniques might
be used to increase conIidence.

Summary
We studied the steps oI Iorecasting
We examined three Iorecasting techniques:
Judgmental
Associative
Time Series
We learned about seasonality, trend, cyclical data
Discussed monitoring Iorecast accuracy