Supply Chain Management: Risk pooling
Donglei Du
([email protected])
Faculty of Business Administration, University of New Brunswick, NB Canada Fredericton
E3B 9Y2
Donglei Du (UNB) SCM 1 / 24
Table of contents I
1 Introduction
2 The theory behind risk pooling
3 A case study
4 Observations from the case
5 Benefits of Risk Pooling
6 Centralized vs decentralized systems
Donglei Du (UNB) SCM 2 / 24
Section 1
Introduction
Donglei Du (UNB) SCM 3 / 24
Risk Pooling I
Risk Pooling involves using centralized inventory instead of
decentralized inventory to take advantage of the fact that if
demand is higher than average at some retailers, it is likely to be
lower than average at others. This reduction in variability
directly leads to a decrease of the safety stock,
√
ST = zα Lσ, (1)
and eventually leads to reduction in average inventory
√
Lµ + ST = Lµ + zα Lσ.
Donglei Du (UNB) SCM 4 / 24
Risk Pooling II
Thus, if each retailer maintains separate inventory and safety
stock, a higher level of inventory has to be maintained than if
the inventory and safety stock are pooled. Therefore the system
with risk pooling has less overall inventory and is thus cheaper
to operate with the same service level.
Donglei Du (UNB) SCM 5 / 24
Section 2
The theory behind risk pooling
Donglei Du (UNB) SCM 6 / 24
Risk Pooling
Var(X1 + X2 ) = Var(X1 ) + Var(X1 ) + 2Cov(X1 , X2 )
≤ σ 2 (X1 ) + σ 2 (X2 ) + 2σ(X1 )σ(X2 )
≤ (σ(X1 ) + σ(X2 ))2
⇓
σ(X1 + X2 ) ≤ σ(X1 ) + σ(X2 )
The first inequlity above follows from the Cauchy-Schwarz
inequality:
|Cov(X1 , X2 )| ≤ σ(X1 )σ(X2 )
Donglei Du (UNB) SCM 7 / 24
Section 3
A case study
Donglei Du (UNB) SCM 8 / 24
A case study
Let us consider the case.
Decentralized System
1.05Q Warehouse One Market One
Supplier
Warehouse Two Market Two
Lead time =1 SL=0.97, S=60,H=0.27
Centralized System
Market One
Supplier Warehouse
1.10Q
0.97 service level
Market Two
Donglei Du (UNB) SCM 9 / 24
Characteristic of centralized and decentralized
systems I
two products with historical data given in the following table:
Week 1 2 3 4 5 6 7 8
Prod A, 33 45 37 38 55 30 18 58
Market 1
Prod A, 46 35 41 40 26 48 18 55
Market 2
Prod B, 0 2 3 0 0 1 3 0
Market 1
Product B, 2 4 0 0 3 1 0 0
Market 2
maintain 97% service level
$60 order cost
$.27 weekly holding cost
Donglei Du (UNB) SCM 10 / 24
Characteristic of centralized and decentralized
systems II
$1.05 transportation cost per unit in decentralized system, $1.10
in centralized system
1 week lead time
Note that the demand for Product B is fairly small relative to
the demand for product A.
Donglei Du (UNB) SCM 11 / 24
Analysis of the two systems I
The following table gives a summary of the average and
standard deviation of demands for each product in both the
decentralized and centralized systems.
Warehouse Product AVG STD CV s S Avg. %
Inven. Dec.
Market 1 A 39.3 13.2 .34 65 197 91
Market 2 A 38.6 12.0 .31 62 193 88
Market 1 B 1.125 1.36 1.21 4 29 14
Market 2 B 1.25 1.58 1.26 5 29 15
Cent. A 77.9 20.7 .27 118 304 132 36%
Cent B 2.375 1.9 .81 6 39 20 43%
We explain how to get the information in the last four columns
Donglei Du (UNB) SCM 12 / 24
Analysis of the two systems II
1 Column with ”s”: s is the reorder point
q
s = ROP = µD µL + zα µL σD 2 + µ2 σ 2
D L
For example the ROP for product A at market 1 is given by
(note that z0.03 = 1.89
q
s = ROP = µD µL + zα µL σD 2 + µ2 σ 2
D L
p
= 39.3(1) + z0.03 1(13.2)2 + (39.3)2 (0)
= 39.3 + 1.89(13.2) ≈ 65
Donglei Du (UNB) SCM 13 / 24
Analysis of the two systems III
2 Column with ”S”: S is the order-up-to point. Suppose we use
the EOQ order quantity, then
r
∗ 2Ds
S =s+Q=s+Q =s+
h
For example, the S for product A at market 1 is given by
r
2Ds
S =s+S = s+
h
r
2(39.3)(60)
= s+
0.27
≈ 65 + 132 = 197
Donglei Du (UNB) SCM 14 / 24
Analysis of the two systems IV
3 Column with ”Average Inventory”:
Q Q
q
Average Inventory = Safety Stock+ = zα 2 + µ2 σ 2 +
µL σ D D L
2 2
For example, the Average Inventory for product A at market 1 is
given by
2 + µ2 σ 2 + Q
q
Average Inventory = zα µL σ D D L
2
132
= 1.89(13.2) + ≈ 91
2
Donglei Du (UNB) SCM 15 / 24
Analysis of the two systems V
4 Column with decreasing percent: This tells the change of
average inventory between the decentralized and centralized
systems, given by
decentralized average inventories - centralized average inventories
centralized average inventories
For example, the average inventory for Product A is reduced by
(91 + 88) − 132
≈ 36%
132
when we shift from the decentralized system to the centralized
one.
Donglei Du (UNB) SCM 16 / 24
Section 4
Observations from the case
Donglei Du (UNB) SCM 17 / 24
Observations from the case
Note that the average demand faced by the centralized
warehouse is the sum of the average demand faced by each of
the two warehouses in the decentralized system.
However, the variability faced by the centralized warehouse,
measured either by the standard deviation or coefficient of
variation, is much smaller than the combined variabilities faced
by the two warehoused in the decentralized system.
Donglei Du (UNB) SCM 18 / 24
Section 5
Benefits of Risk Pooling
Donglei Du (UNB) SCM 19 / 24
General Observations on Risk Pooling I
Centralizing inventory control reduces both safety stock and
average inventory level for the same service level.
Intuitively, in a centralized distribution system, whenever
demand from one market area is higher than average while
demand in another market area is lower than average,items in
the warehouses that are originally allocated for one market can
be reallocated to the other. This reallocation process is not
possible in a decentralized system where different warehouses
serve different markets.
The benefit of risk-pooling depends on the standard deviation
(SD) or the coefficient of variation (CV) among the different
markets. The higher the SD/CV, the greater that potential
benefit from centralized system becase of equation (1).
Donglei Du (UNB) SCM 20 / 24
General Observations on Risk Pooling II
The benefit of risk-pooling also depends on the demand
correlation among the different markets.
Var(X1 + X2 ) = Var(X1 ) + Var(X1 ) + 2Cov(X1 , X2 )
We say the demands are positively correlated if the demands in
both markers are increasing or decreasing in the same direction;
that is, Cov(X1 , X2 ) ≥ 0.
The benefit decreases as the correlation between demand from
two markets becomes more positive, as the standard deviation
becomes larger.
There are many Types of Risk Pooling, here are some examples
Risk Pooling Across Markets
Risk Pooling Across Products
Risk Pooling Across Time
Donglei Du (UNB) SCM 21 / 24
Section 6
Centralized vs decentralized systems
Donglei Du (UNB) SCM 22 / 24
Centralized vs decentralized systems I
What are the tradeoff that we need to consider in comparing
centralized vs decentralized systems?
When we switch from a decentralized system to a centralized
system:
Safety stock usually decreases, leading to decrease in average
inventory level and hence reduction in inventory holding cost.
The magnitude of the increase depends on the SD and/or CV
and the correlation between the demand from the different
markets.
Service level increases and the magnitude of the increase
depends on the SD/CV and the correlation between the demand
from the different markets.
Overhead costs decreases as there are more economies of scale.
Customer lead time becomes longer.
Donglei Du (UNB) SCM 23 / 24
Centralized vs decentralized systems II
Transportation costs increases on the outbounds and decreases
on the inbounds. The net impact on total transportation cost is
not immediately clear.
Donglei Du (UNB) SCM 24 / 24
