Académique Documents
Professionnel Documents
Culture Documents
IIKIE
In Tsuen Wan, New Territories, Hong Kong a joint venture was established in 1985 to
coordinate Sport Obermeyers Far East operations. A new factory complex recently
completed at Lo Village, Guangdong, China with new facilities, new workers, and
training still going on. Demand, worker skill levels, and productivity are all hard to
predict.
Prod. Plan. under Uncertainy 1
IIKIE
Lo Village
IIKIE
IIKIE
Obermeyer products were offered in five different genders: men, women, boy, girl,
and preschooler. The company had segmented each gender market according to
price, type of skier, and how fashion-forward the market was. For example, the
company divided its adult male consumer into four types:
Fred: conservative, tends to buy basic styles and colors, likely to wear the
same outfit multiple seasons
Rex: high-tech, wealthy, image-conscious, tends to use the latest technologies
in fabrics, features, and ski equipment
Biege: hard-core mountaineering-type skier, places technical performance
above all else, despises any nonfunctional design elements
Klausie: flamboyant, high-profile skier or snowboarder, wears the latest
styles, often in bright colors such as neon pink or lime green.
Within each gender, numerous styles were offered, each in several colors and a
range of sizes. For each gender, total number of SKUs = (#styles)(#colors)(#sizes).
Prod. Plan. under Uncertainy 4
IIKIE
Womens parkas
Mens parkas
Prod. Plan. under Uncertainy 5
IIKIE
Womens parkas
800
Total Number of SKUs
80-81
82-83
84-85
86-87
88-89
90-91
92-93
Number of Styles
Number of
Styles
20
10
78-79
8
Average Num of
Colors and Sizes 6
per Style
80-81
82-83
84-85
86-87
Avg.Num.of
Colors per Style
88-89
90-91
92-93
4
78-79
80-81
82-83
84-85
86-87
88-89
90-91
92-93
IIKIE
IIKIE
Wally Obermeyer worked part-time for the company and worked as a skipatrol
on Aspen Mountain. Graduated Harvard Business School in 1986 and focussed his
efforts on developing a hydro-electric power generating plant in Colorado. After the
power plant was established he joined Sport Obermeyer full time in 1989 as vice
president.
Wallys management approach relied more heavily on formal data-gathering and
analytical techniques, whereas Klaus took a more intuitive style that was heavily
informed by his executive industry experience.
IIKIE
Order Cycle
products are sold primarily through specialty ski-retail stores, located either in urban
areas or near ski resorts, and direct mail retailers. Most retail sales of skiwear occured
between September and January, with peak sales occuring in December and January.
Design Process (Feb-Jul91)
sense of fashion conceived, sketches drawn, prototypes produced, design finalized
Sample Production (Aug-Dec91)
sample production ordered to subcontractors
sales representatives start to show samples to retail sites
Raw Material Orders, Sourcing, and Production (src:Jul-Nov91 prod:Feb-Jul&Aug92)
sourcing for initial production (typically half of the annual production)
long fabric supply lead times before cutting & sewing can begin
IIKIE
IIKIE
Design Process
Production
Apr
1991
Jun
1991
Aug
1991
Nov
1991
Dec
1991
Feb
1992
Apr
1992
Jun
1992
Aug
1992
Nov
1992
Dec
1992
Feb
1993
IIKIE
AvgError=180
2000
AvgError=760
1500
Forecast
Error 1000
Low
Error
500
0
100
200
300
400
Std.Dev. of Individual Forecasts
Low
High
Aggreement
Aggreement
Season
Sales
4000
3000
2000
1000
IIKIE
Initial Forecast
each data point represents the
forecast and actual season sales
for a particular parka at the style
and color level. For example
parka A had an initial forecast of
2800 and season sales of 1000
units.
A
0
1000
2000
3000
4000
Initial Forecast
Season
Sales
4000
IIKIE
Updated Forecast
after observing the first 20% of
demand, the revised forecast for
parka A was 1100 units.
3000
2000
1000
A
0
1000
2000
3000
4000
Updated Forecast
Season
Sales
4000
IIKIE
Final Forecast
after observing the first 80% of
demand, the final forecast for
parka A was 1020 units.
3000
2000
1000
1000
2000
3000
4000
Final Forecast
IIKIE
Apparel
Manufacturers
Obersport
Sport
Obermeyer
Retailers
production:
capacity of prod. lines
workers
min. order requirements
wages
Hong Kong
faster & more flexible
highly trained
smaller
higher
China
slower & less flexible
inexperienced
larger
lower
IIKIE
A Sample Problem
Wally decides to look at the following sample problem to build intuition about how to
make production decisions. This is smaller version of the companys problem. The
Buying Committees forecasts for the sample of 10 womens parkas are given below.
Since these style represent about 10 percent of Obermeyers total demand, to make
this smaller version representative of the larger problem he assumed he had cutting
and sewing capacity of 3000 units per month (10 percent of the actual capacity) during
the seven month production period. Using these assumptions Wally needed to commit
10,000 units for the first phase of production. The remaining 10,000 units could be
deferred until the after the Las Vegas show. The Chinese factory requires a minimum
of 1200 units for each style ordered whereas the Hong Kong factory requires only 600
units.
Wally studied the buying committees forecasts, wondering how he could estimate the
risk associated with early production of each style. Was there some way he could use
the differences among each members forecast as a measure of demand uncertainty?
IIKIE
Price
Tom
Wally
Avg
StDv
2*StDv
Gail
$110
900
1000
900
1300
800
1200
1017
194
388
Isis
$99
800
700
1000
1600
950
1200
1042
323
646
Entice
$80
1200
1600
1500
1550
950
1350
1358
248
496
Assault
$90
2500
1900
2700
2450
2800
2800
2525
340
680
Teri
$123
800
900
1000
1100
950
1850
1100
381
762
Electra
$173
2500
1900
1900
2800
1800
2000
2150
404
807
Stephanie
$133
600
900
1000
1100
950
2125
1113
524
1048
Seduced
$73
4600
4300
3900
4000
4300
3000
4017
556
1113
Anita
$93
4400
3300
3500
1500
4200
2875
3296
1047
2094
$148
1700
3500
2600
2600
2300
1600
2383
697
1394
20000
20000
20000
20000
20000
20000
20000
Daphne
Totals
Laura Carolyn
Greg Wendy
Laura Kornashiewicz: marketing director; Carolyn Gray: customer service manager; Greg Hunter: production
manager; Wendy Hemphill: production coordinator; Tom Tweed: sales representative; Wally Obermeyer:
vice president
Prod. Plan. under Uncertainy 18
IIKIE
A Sample Problem
An examination of demand from previous years indicated that the forecast accuracy
was the highest for those styles for which the Buying Committee had the highest level
of aggreement. Technically he found that the standard deviation of demand for a style
was approximately twice the standard deviation of the Buying Committees forecasts
for that style.
std.dev=807
mean=2150
Forecast distribution for the Electra parka
Prod. Plan. under Uncertainy 19
IIKIE
Where to Produce
To complete the planning decision, Wally would also need to decide which styles to
make in Hong Kong and which would be better to produce in China. This year,
Obermeyer expected to produce about half of all its products in China. Longer term,
Wally wondered whether producing in China would constraint Obermeyers ability to
manage production and inventory risks. Would Chinas larger minimum order sizes
limit the companys ability to increase the range of products it offered or to manage
inventory risk? Was Obermeyers trend toward increased production in China too risky
given the uncertainty in Chinas trade relationship with the US?
IIKIE
Discussion Questions
1. Using the sample data given, make a recommendation for how many units of each
style Wally should make during the initial phase of production. Assume that all of
the 10 styles in the sample problem are made in Hong Kong and that Wallys initial
production commitment must be at least 10,000 units. Ignore price differences
among styles in your initial analysis.
2. Can you come up with a measure of risk associated with your ordering policy? This
measure should be quantifiable.
3. Repeat your methodology and assume now that all 10 styles are made in China.
What is the difference (if any) between the two initial production commitments?
5. How would you take into account the minimum order requirements?
6. What operational changes would you recommend to Wally to improve
performance?
7. How should Wally think (both short-term and long-term) about sourcing in Hong
Kong versus China? What kind of sourcing policy do you recommend?
IIKIE
Let EPi(Qi) = E[Profiti(Qi)] for convenience. How can we solve the following problem? Is
this problem meaningful?
max
Expected
Profit
Invested
Capital
max
EPi (Qi )
ci Qi
tp
Capital
Profit
EPi (Qi ) tp ci Qi
We can also rephrase the problem as determining the maximum expected profit we
can achieve given a certain budget. Similarly production capacity may impose a similar
constraint:
max Expected Profit
s.t. Invested Capital K
EPi (Qi )
s.t. ci Qi K
max
IIKIE
We can approach this problem using Lagrange multipliers and exploit separability.
max EPi (Qi )
min max EPi (Qi ) Qi C
0 Q
s.t. Qi C
When we approach the inner maximization problem by writing the optimality
conditions we can get for all products i:
d
EPj (Q j ) Q j C 0
dQi j
d
EPi (Qi ) 0
dQi
dEPi (Qi )
Qi* dEPi1 (* )
Prod. Plan. under Uncertainy 23
IIKIE
This reduces the optimization problem into a line search problem. Find * such that
Q*(*) values sum up to C:
Find * such that
Qi* (* ) C
i
Qi* (* )
dEPi
s
(* ) Fi 1 i
s o
1 si *
i
i
i s o
i i
In this line search, value of * may result with a probability value less than 0 (i.e. when
* > si). Such cases must be handled using managerial input deciding whether to not
produce such styles at all or produce at least a minimum amount.
Prod. Plan. under Uncertainy 24
IIKIE
There are interesting extensions of this problem, some of them are described below:
IIKIE
Production Committment
based on
Updated Demand Forecast
Speculative
Production Capacity
Initial Dem.
Forecast
Update Dem.
Forecast
Reactive
Production Capacity
Regular
Selling Season
Time
Line
Prod. Plan. under Uncertainy 26
IIKIE
IIKIE
where ei = premium paid for any units produced in the second order (ci + ei < pi)
Marginal analysis reveals a newsvendor type cost structure:
d
EPQRUi (Qi ) c i (c i ei ) P(Di Qi ) qi P(Di Qi )
dQi
ei P(Di Qi ) (c i qi ) P(Di Qi )
ei (ei oi ) F (Qi )
e
Finally the optimum first order is: Q*i Fi1 i
ei oi
What is the risk of having a Profit less than a target proportional to the avg. demand?
P(PQRUi t i ) ?
Prod. Plan. under Uncertainy 28
IIKIE
Example: Consider the following initial forecast. A 10% premium is charged on the
orders placed with the reactive capacity. Let us find the unlimited reactive capacity
order quantities. We first find the unconstrained optimum orders, corresponding exp.
profits, mismatch costs. The last column gives the loss risk at 12% of the unit cost.
Avg
StDv
Q*
EP
94
1646
827
2204
139925
37843
0.134
152
110
1325
861
1777
87939
41911
0.236
155
103
68
3232
1616
3332
89875
31485
0.156
Kennedy
243
147
106
2730
1919
3740
170771
91309
0.256
Lexington
187
117
70
3677
1177
3970
204128
53262
0.082
McKenzie
177
99
65
2455
1596
3276
129008
62482
0.226
Niki
140
90
57
1741
601
1897
67809
19241
0.103
Prague
125
85
58
3034
1214
3632
113666
54398
0.208
St. Tropez
170
117
86
1250
462
1404
51610
14640
0.129
Ute
100
68
51
2910
960
3288
75753
17367
0.090
28520 1130483
423939
Style
Price
Cost
Aiden
238
130
Aston
250
Ibiza
Totals
Discount
24000
Mm.Cost
Risk
IIKIE
Example: (cont) First order quantities are calculated taking into account a premium
charge of 10%. Exp.Profit for the styles in the first order are calculated via the EPQRU
formula. Mismatch cost equals maximum profit minus EPQRU. The last column gives
the loss risk at 12% margin over average demand:
Style
Q EPQRU
Mm.Cost
R Risk
Aiden
1127
164487
13281
652
0.050
Aston
786
113696
16154
678
0.103
Ibiza
2174
107979
13381
1031
0.038
Kennedy
1518
227143
34937
1519
0.118
Lexington
2684
238087
19303
1124
0.014
McKenzie
1252
170450
21040
1411
0.097
Niki
1265
79689
7361
549
0.020
Prague
2024
145978
22086
1421
0.068
972
59679
6571
355
0.027
2367
85354
7766
714
0.015
16170
1392542
161880
9455
St. Tropez
Ute
Totals
IIKIE
IIKIE
Case Q* > 2m : These items do not pose much risk. We can safely order some amount
with the speculative capacity. We can easily determine an amount such that the
adjustment quantity to be ordered with the reactive capacity after the demand update
is received is likely to be greater than m.
Heuristic #1: We commit to production of as many of the least risky items as possible
in the speculative capacity (depending on how much speculative capacity we have
secured). Once the updated forecast is available we order the most risky items for the
first time using the new information and place orders for the items that were ordered
with the speculative capacity if the adjustment quantities are higher than the
minimum order quantity or close enough so that we order the minimum anyway.
Heuristic #2: We order each item either with the speculative capacity or the reactive
capacity but not both. This is the so called order once constraint. We sort all items in
ascending order of a risk measure then commit to production of a number of items
from the top such that the total quantity is approximately equal to the amount we
would like to in the speculative capacity. After the updated forecast is available we
order the rest of the items. The risk measure we use is the unit mismatch cost:
Unit Mismatch MismatchCo st(Q* )
Cost
Q*
Prod. Plan. under Uncertainy 32
IIKIE
Example: Let us this time find the limited reactive capacity order quantities on the
same initial forecast used in the previous example. First order will consist of about half
of the total order quantity. We report here again the unconstrained optimum orders,
corresponding exp. profits, mismatch costs. The last column gives the loss risk at 12%
of the unit cost.
Avg
StDv
Q*
EP
94
1646
827
2204
139925
37843
0.093
152
110
1325
861
1777
87939
41911
0.179
155
103
68
3232
1616
3332
89875
31485
0.082
Kennedy
243
147
106
2730
1919
3740
170771
91309
0.201
Lexington
187
117
70
3677
1177
3970
204128
53262
0.038
McKenzie
177
99
65
2455
1596
3276
129008
62482
0.180
Niki
140
90
57
1741
601
1897
67809
19241
0.050
Prague
125
85
58
3034
1214
3632
113666
54398
0.137
St. Tropez
170
117
86
1250
462
1404
51610
14640
0.057
Ute
100
68
51
2910
960
3288
75753
17367
0.033
28520 1130483
423939
Style
Price
Cost
Aiden
238
130
Aston
250
Ibiza
Totals
Discount
24000
Mm.Cost
Risk
IIKIE
Example: (cont) We first calculate the unit mismatch costs. Then we add styles to the
initial order one by one in ascending order of the unit mismatch cost until the total
initial order is approximately 12000 units. Exp.Profit for the styles in the first order are
the unconstrained optimum values. For the styles left to be produced with the reactive
capacity, Exp.Profit is the maximum profit and there is no mismatch cost or risk:
Q EPQRL
Style
Mm.Cost/Q*
Mm.Cost
Risk
Aiden
17.2
177768
0.000
Aston
23.6
129850
0.000
Ibiza
9.4
3332
89875
31485
0.082
Kennedy
24.4
262080
0.000
Lexington
13.4
3970
204128
53262
0.038
McKenzie
19.1
191490
0.000
Niki
10.1
1897
67809
19241
0.050
Prague
15.0
168064
0.000
St. Tropez
10.4
1404
51610
14640
0.057
5.3
3288
75753
17367
0.033
13892
1418426
135996
Ute
Totals
IIKIE
Examine the following plots. Product styles are ordered in ascending order of the unit
mismatch cost. The light shaded bars denote the less risky first order and the darker
shaded bars denote the more riskier styles left to be produced with reactive capacity.
We notice that unit mismatch cost roughly increases as Gross Margin (s), or as Critical
Ratio, or as Coefficient of Variation increases:
120
Gross Margin
80
0.8
Critical Ratio
0.70
Coefficient of Variation
0.7
0.5
0.20
Ute
Ibiza
Niki
St. Trp.
Lexing.
Prague
Aiden
McKnz
Aston
Kenn.
0.6
Ute
Ibiza
Niki
St. Trp.
Lexing.
Prague
Aiden
McKnz
Aston
Kenn.
40
Ute
Ibiza
Niki
St. Trp.
Lexing.
Prague
Aiden
McKnz
Aston
Kenn.
0.55
There are 210 = 1024 different ways to place the initial order under the order once
constraint. Exp.Profit value versus total first order size is plotted in the graph on the
next page. The rules of Heuristic#2 describes 11 initial orders out of the 210 possible
orders. These 11 initial orders are all Pareto-optimal and are denoted with red
diamond markers in the plot on the next page.
Prod. Plan. under Uncertainy 35
IIKIE
Expected Profit