Risk Pooling
Chap. 14
• Risk Pooling – “Theory”
• Applications
– Differentiation Postponement/Delay
– Location, …
1. On page 1, it said “We can’t run our business with this level of unproductive
assets.” What are these assets?
2. What is “the I-word” referred to?
3. Is the ink-jet printer a commodity or fashionable product?
4. At the European DC, did HP have too much stock or too little stock?
5. What were the symptoms of the problem in the European DC?
6. When customers buy ink-jet printers, is brand/product loyalty playing an
important factor in their choosing which products to buy?
7. How did the Vancouver Division impress visitors? Was the line suitable for high
volume or low volume production? Why?
8. Were Ink-Jet Printers completely built in the Vancouver factory? Can a printer
that was built for the Germany market be directly sold in Italy market? Why? Be
precise.
9. What were the performance evaluation criteria for European DC? Should the DC
manager be concerned about the inventory level?
10. What were the alternatives for resolving the inventory and service crisis?
HP Case - handed out – preparation questions
Coefficient of Variation
CV = Standard Dev / Average demand
Demand series 1: CV1 = 0.5
Demand series 2: CV2 = 2.0
Which is more volatile?
Risk Pooling
• Which sales are more volatile: the regional
sales or sales at the store level?
• Which demand is more volatile : a family
of products or individual members of the
family?
• Do people wait longer in a multiple-
waiting-lines system than a single-waiting-
line system?
• Implication for forecasting ?
Family vs. Individual Products
Time
Sales
A
B
C
N
Family vs. Individual Products
Model Mean Stdv CV
X 42 32 0.78
XX 420 203 0.46
XY 15,830 5,624 0.36
XH 2,301 1,168 0.51
XC 4,208 2,204 0.52
XY 309 103 0.34
Total 23,109 6,244 0.27
Pooling, Profit & Service Level
(An Example)
• Two products (paints) differ only in colour
• It is fast to mix to the required colour upon
receiving orders
• Assume that the demand for each follows a
distribution given by tossing a dice
An Example
• Alternative I: Make to Stock
• Alternative II: Make to Order (for colouring
only)
• Order-up-to inventory replenishment policy
• One season, c=$2.5, p=$12.5, s=0
For Alternative I: 5 units for each
• For Alternative II: ?
Preliminary Calculations
1 2 3 4 5 6
1 1,1 x
2 x
3 x
4 4,3 4,5 x
5 5,4 5,5 x
6 x x x x x xChance of stockout?
Preliminary Calculations
If D1=4, D2 =5, profit = ?;
If D1=6, D2=4, profit = ?.
1 2 3 4 5 6
1 1,1 x
2 x
3 x
4 4,3 4,5 x
5 5,4 5,5 x
6 x x x x x x
• Holding 10 units of “generic” colour -
pigment, the chance of stocking out in a
period is only
3/36 = 1/12 = 0.083
The “risk” of disservice is lowered.
• Even holding 9 units, a higher level of
“service” will be achieved (as compared with
Alternative 1)
Preliminary Calculations
1 2 3 4 5 6
1 2 3 4 5 6 7
2 3 4 5 6 7 8
3 4 5 6 7 8 9
4 5 6 7 8 9 10
5 6 7 8 9 10 11
6 7 8 9 10 11 12
If D1=4, D2 =5, profit = ; If D1=6, D2=4, profit = 0. If
D1=2, D2=6, profit= . Additional cost?
“Theory”
,1|| general,In )(
,1 If )(
,1 If )(
,0 If )(
2)(
)( , Let
, Stdv
,Mean
, :products Two
2221
21
2121
21
21
21
d
c
b
a
xVar
xmeanxxx
xx
Why?
Implications?
• Variety is the “culprit” of high forecasting
errors, and higher forecasting accuracy can be
achieved if only a few varieties are offered
• Since we can not reduce them, we must find a
way to get around
– By postponing mass customisation
– By redesigning the product
• Universal products and common parts
(modules)
Product Variety Proliferation
• Product proliferation exists in various
forms
– global mkt: “protocols”, languages, phases,
elec.
– local mkt: multiple models differ in features &
capacities
– mkting strategies
• Marketing strategy is the major reason
The world of The Long Tail
Pitfalls of increasing product variety
Risk pooling strategies
• The objective of a risk pooling strategy is to redesign the supply chain,
the production process or the product to either reduce the uncertainty
the firm faces or to hedge uncertainty so that the firm is in a better
position to mitigate the consequence of uncertainty.
• Four versions of risking pooling:
– product pooling -- delayed differentiation/postponement
– location pooling
– lead time pooling
• delayed differentiation (HP case)
• consolidated distribution
– capacity pooling
14-18
Postponement
• Key idea - postpone the commitment of WIP into a particular finished product –SKU
• Delay of product differentiation closer to time of sale.
• Prior to point of postponement, only certain degree of aggregate forecast needed
• Individual forecasts more accurate close to time of sale
Postponement Concepts
• Mainly two forms
– Logistics Postponement: moving customisation point
closer to customers - out of mgr functions
– Form Postponement: delaying differentiation point by
standardisation or process re-sequencing
Logistics Postponement
Manufac- Integra- Customi- Locali- Pack-
turing tion sition tion ing
Supply Chain Process
Distribution Centers
Factory
Logistic Postponement by Process
Resequencing: Paint Retail
Colour pigments,
paint mixing,
packaging
Retail sales
Colour pigments,
white paint
Retail sales,
paint mixing
packaging
Nippon: combined
Before module design of the
metal frame
Integration+ship DC
DC+ pannel assembly
Fib
Fab.
Black
White
Black
WhiteIntegration+ship
Dishwasher
Operations Buffer
Log PP: More Examples
• Rheem Manufacturing Co., kept 120 SKUs (heaters)
at its factory. Some were overstocked while other
fall short - only different in several elements
– Using a 3rd party to hold around 10 basic models
and parts
– Filling orders in hours and saving 15% of
inventory cost
• Even Coffee Rosters use it
• Of course, PC mfgrs apply it
• Some done by customers. More real life examples?
PC: indirect model (traditional)
Suppliers
VARsMfr factory Distribution
Centers
Distributor Customers
PC
CompanyOrders
Product
Configuration:
Disk, memory...
Hybrid Model
Suppliers
VARsFactory: core Distribution
Centers
Distributor Customers
PC
CompanyOrders
Product
Assembly
+Configuration
Other examples of delayed
differentiation
• Private label soup manufacturer:
– Problem: many different private labels (Giant, Kroger, A&P, etc)
– Solution: Hold inventory in cans without labels, add label only
when demand is realized.
• Black and Decker:
– Sell the same drill to different retailers that want different
packaging.
– Store drills and package only when demand is realized.
• Nokia:
– Customers want different color phones.
– Design the product so that color plates can be added quickly and
locally. 14-27
Form Postponementby Common Part
Before After
Sometimes called
standardization
PCA FA&T Customizatio
PCA FA&T
Customization
Operations Buffer
Mono
Color
Mono
Color
Mono/Color Printers
No correlation
Form Postponement by Process Reengineering
PCB Insertion
Series of tests and burn-in
Coupon PCB
Common
tests
Customisation
tests
Benetton: Process Reengineering
Purchase Yarn
Dye Yarn
Finish Yarn
Knit Garment Parts
Join Parts
Old Sequence
Purchase Yarn
Knit Garment Parts
Join Parts
Dye Garment
Finish Garment
New Sequence
This process
is postponed
Process Redesign for Supply Chain:
Postponement at Benetton
dye
knit
Dyeing
operations
postponed
Dye yarn only after the season’s fashion preferences become more
established (knit lead-time much longer than dyeing lead-time).
Example: single product; four colors
knit
dye
Outcome: Reduces demand uncertainty & inventory
Demand correlation
• Correlation refers
to how one random
variable’s outcome
tends to be related
to another random
variable’s
outcome.0
2
4
6
8
10
12
14
16
18
20
0 5 10 15 20
0
2
4
6
8
10
12
14
16
18
20
0 5 10 15 20
0
2
4
6
8
10
12
14
16
18
20
0 5 10 15 20
Random demand for two products (x-axis
is product 1, y-axis is product 2). In
scenario 1 (upper left graph) the
correlation is 0, in scenario 2 (upper right
graph) the correlation is -0.9 and in
scenario 3 (the lower graph) the
correlation is 0.90. In all scenarios
demand is Normally distributed for each
product with mean 10 and standard
deviation 3.14-33
Limitations of product
pooling/universal design
• A universal design may not provide key functionality to consumers with
special needs:
– High end road bikes need to be light, high end mountain bikes need to be
durable. It is hard to make a single bike that performs equally well in both
settings.
• A universal design may be more expensive to produce because additional
functionality may require additional components.
• But a universal design may be less expensive to produce/procure because each
component is needed in a larger volume.
• A universal design may eliminate brand/price segmentation opportunities:
– There may be a need to have different brands (e.g., Lexus vs Toyota) and
different prices to cater to different segments.
14-34
Summary: When is PP
Valuable?
• A lot of varieties
• Demand uncertainties over the variety are high
– Negatively correlated, ?
– Positively correlated, ?
• Differentiation is not too costly to perform locally, or
not time consuming
• The “core components” have high value, but
differentiating parts are of low value
Common Obstacles
• Though often design changes do not cost much, people
resist their implementation
• As production cost may increase, prod. people may oppose
to changes
• They also pose challenges to designers
• Indirect cost savings and intangible benefits