Risk Pooling

Customer management
Supplier management
Inventory management
ROI

Location management
𝑹𝑬𝑽𝑬𝑵𝑼𝑬 − 𝑪𝑶𝑺𝑻𝑺
𝑶𝒑𝒕𝒊𝒎𝒂𝒍 𝑨𝑺𝑺𝑬𝑻𝑺
Transport management

Value chain integration

Inv. Mgt.- Inventory Positioning
Introduction
1. Definition
• Ordering policies Learning outcomes
2. Types
• EOQ, Economic order • Explain the risk pooling
3. Relevance:
quantity effect and its benefits in a
1. Financial • Continuous review: Min- supply chain
2. Operational Max policy
• Interpret the relationship
ABC classification • Safety stock
1. Description • Average Inventory between demand variability
2. ABC classification procedure • Periodic review: Base-Stock and risk pooling
3. Classification criteria policy • Decide when centralized
4. Case study for a large distributor • Safety Stock stocking or distributed
Inventory Policy Factors • Average Inventory stocking is more appropriate
• Inventory ordering costs, and inventory

•
holding costs • Inventory Positioning in the supply chain due to
Service level requirements
• Risk Pooling the risk pooling effect
• Forecast error
• Length of planning horizon • Centralized versus • Calculate the inventory costs
• Replenishment lead-time decentralized inventories associated with centralized
•
•
Product variety
Demand Variability and Forecasting error
• Strategic Safety Stock the and decentralized inventory
• Safety stocks supply chain: in the supply chain
• Push-pull boundary
CASE STUDY: Risk Pooling
Risk Pooling refers to the reduction in demand variability resulting from
aggregating demand streams from different locations.
To illustrate the concept, consider the case of ElecTex. The company has a
manufacturing plant in Temple, Texas. ElecTex sells around 900 different
products, through regional distribution warehouses in Dallas, Houston and San
Antonio; the warehouses deliver orders to 8,500 customers (retailers). The
company strategy is to provide with 98% fill-rate customer service level. The
warehouses use a max-min inventory review policy; and uses EOQ as the order
quantity. The plant’s delivery lead time to the warehouses is 2-weeks; and the
delivery from the warehouse to the retailers follows a same-day-delivery policy.
The company is evaluating the positioning of its inventory; specifically, they
would like to close the three regional warehouses and consolidate the inventory
into a single warehouse, called Centralized Distribution Warehouse, CDW, that
would serve all customers with the same customer service level. The location of
the new CDW would be in Cameron-Texas close to the Manufacturing facility.
Management thinks this might result in reduced costs due to better forecasts
and reduced safety inventory levels for the same fill-rate, due to the risk pooling
effect, in addition to a reduction in staff and operators that would result from
reducing the number of warehouses, and automation in the new CDW.
Possible disadvantages from the CDW configuration are additional
transportation costs and delivery times from the warehouse to the customers,
and additional distance between sales workforce and customers.
Notice that, under the proposed idea, the distance from the manufacturer to
the warehouse is reduce; however, it is unclear whether this reduction and the
inventory reductions will exceed the additional transportation cost between the
warehouse and the customers.
 Risk Pooling Case Study - Analysis
Let’s first study the impact that will result from risk pooling reducing inventory. If the results are promising, the study would need to also
consider transportation costs.
All 900 products must be included in the study; however, here the analysis will be performed on two products V and W for clarity and brevity.
The data and calculations are available in the provided spreadsheet “Risk Pooling – Case [Link]”.
The tables below show historical weekly demand for the two products, as recorded for each of the current warehouses.
Product V SUMMARY
Week 1 2 3 4 5 6 7 8 9 10 11 12 13 average std cv
Houston 44 40 56 26 89 38 48 38 59 40 34 55 45 47.08 15.64 0.33
Dallas 30 18 54 26 21 35 29 21 18 34 38 25 33 29.38 9.90 0.34
San Antonio 31 30 26 36 36 34 18 17 33 35 37 45 35 31.77 7.69 0.24
TOTAL 105 88 136 88 146 107 95 76 110 109 109 125 113 108.23 19.47 0.18

Product W SUMMARY
Week 1 2 3 4 5 6 7 8 9 10 11 12 13 average std cv
Houston ` 13 9 9 11 9 0 0 2 0 6 8 4 5.54 4.61 0.83
Dallas 4 0 0 1 3 2 3 0 5 1 0 9 4 2.46 2.63 1.07
San Antonio 0 3 0 0 0 0 2 0 4 3 0 0 6 1.38 2.02 1.46
TOTAL 5 16 9 10 14 11 5 0 11 4 6 17 14 9.38 5.14 0.55

Notice that, the total demand in each week would represent the aggregate demand that the centralized warehouse would have received if
implemented. We will use the aggregate weekly demands to evaluate the performance of the CDW.
One can immediately observe that product V exhibit more stable, and higher demand than product W.
As we have learned, we can quantify these observations by characterizing the demand in terms of average, standard deviation and coefficient
of variability. [ANIMATION] These calculated statistics are displayed in green.
Observing the CV values, confirms our observations. Product V has CV values between .24 to .34, corresponding to a low demand variability
product; on the other hand product W’s Cv is between 0.83 to 1.46 suggesting, currently the warehouses might be experiencing large forecast
errors and higher safety stocks for product W.
 Product V
160

140
SUMMARY – Product V
120
average std cv
100
47.08 15.64 0.33
80
29.38 9.90 0.34
60
31.77 7.69 0.24
40
108.23 19.47 0.18
20

0
SUMMARY – Product W 1 2 3 4 5 6 7 8 9 10 11 12 13

average std cv Houston Dallas San Antonio TOTAL

5.54 4.61 0.83

Product W
2.46 2.63 1.07
20
1.38 2.02 1.46 10

0
9.38 5.14 0.55 1 2 3 4 5 6 7 8 9 10 11 12 13

Houston Dallas San Antonio TOTAL

Risk pooling causes the centralized warehouse to experience a demand that is the sum of the
individual warehouses, but with a much lower variability. This will result in lower safety stocks for
the same customer service level!
Risk Pooling Case Study – Analysis (contd.)
Observing the CV values, confirms our observations. Product V has CV values between .24 to .34,
corresponding to a low demand variability product; on the other hand product W’s Cv is between
0.83 to 1.46 suggesting, currently the warehouses might be having large forecast errors and higher
safety stocks for product W.
The Risk Pooling effect is evident when observing the CV for the aggregate demand. Product V
aggregate demand is 0.18 that is noticeable lower than the .24 to .34 CV observed by each
warehouse. The effect is even more pronounced for Product W, where the aggregate demand CV is
0.55 that is much less than the 0.83 to 1.46 CV experienced by the individual warehouses. In fact,
product W’s demand is highly variable for the individual warehouses, but is only borderline low and
moderate variability for the proposed centralized warehouse!
How is this possible, if the same data is used to calculate these values? By carefully observing what
happens with the random ups-and-downs of the individual demands from week to week across
each warehouse, when the demand of one warehouse increases, the other can decrease, and vice-
versa. As a result, when added the sum of the individual demands resembles the value of the
aggregate demand average. This effect tends to be more noticeable when the variability of the
individual demands is higher. However, if the different demand streams tend to increase or
decrease simultaneously, then the variability cancellation effect will not work. More formally, the
risk pooling effect diminishes as the correlation between the different demands increases. The plot
below shows a line plot of the weekly demands for product W.
 Risk Pooling Case Study – Analysis (contd.)
The total inventory for the current and centralized strategies can be estimated by comparing the corresponding average inventories. The
average inventory, I-bar, is calculated as Q/2 + Safety Stock.
𝑄 2𝐴𝐷
Since ElecTex uses min-max inventory policy, then 𝐼 ҧ = 2
+ 𝑘 𝑠𝑑 𝐿. Also in ElecTex the order quantity used is Q = EOQ = ℎ
. In this case
both products have the same ordering and holding costs.
[Animation] The average inventory calculations are shown in the following table.

[Animation] The average inventory for the current system and the
proposed centralized system:
INVENTORY POLICY CALCULATIONS
Location Product DDLT SS Q r Ave I Centralized versus Current (descentralized) Warehouse (units)
Houston V 94.2 45.4 127 139.6 108.9 Decentralized Centralized %Change
Houston W 11.1 13.4 44 24.5 35.4
Product V (units) 263.0 153.1 -0.4
Dallas V 58.8 28.8 101 87.5 79.3
Product W (units) 74.4 43.4 -0.4
Dallas W 4.9 7.6 29 12.6 22.1
TOTALS 337.5 196.5 -0.4
San Antonio V 63.5 22.3 105 85.9 74.8
San Antonio W 2.8 5.9 22 8.6 16.9
Holding cost ($/wk) 108.0 62.9 -0.4
Cameron (Central) V 216.5 56.6 193 273.0 153.1
Cameron (Central) W 18.8 14.9 57 33.7 43.4
[Animation] It can be observed that the reduction in inventory by
implementing a centralized distribution warehouse is about 50% of the
current inventory.
We will study how to analyze transportation costs later.
Risk Pooling’s Square-Root Rule
There is a useful relationship between the individual demands and the aggregate
demand parameters. Let 𝜎1 , 𝜎2 , … 𝜎𝑛 be the standard deviation of 𝑛 independent
demand streams; the standard deviation of the aggregate sum of these
demands,𝜎𝐴 , is the square-root of the sum of squares of the individual standard
deviations:
𝜎𝐴 = 𝜎12 + 𝜎22 + ⋯ + 𝜎𝑛2
This is known as the “Square-Root Rule”.
Let 𝜇1 , 𝜇2 , … 𝜇𝑛 , be the mean values of the corresponding demands, then the
aggregate mean demand is sim ply the sum of the individual means:
𝜇𝐴 = 𝜇1 + 𝜇2 + ⋯ + 𝜇𝑛
In practice, 𝜎 and 𝜇 values are estimated by the sample average and standard
deviation, 𝑥ҧ and 𝑠, respectively.
 Example: Apply the Square root rule to the ElecTex demand in the Risk
Pooling case study data.
Applying the square-root formula to the individual standard deviations for Product V we
obtain:
2 2 2
𝑠𝐶𝐷𝑊,𝑉 = 𝑠𝐻𝑜𝑢,𝑉 + 𝑠𝐷𝑎𝑙,𝑉 + 𝑠𝑆𝐴,𝑉

= 15.642 + 9.902 + 7.692 = 20.05 ≈ 19.47 units

Applying the square-root formula to the individual standard deviations for Product W we
obtain:
2 2 2
𝑠𝐶𝐷𝑊,𝑊 = 𝑠𝐻𝑜𝑢,𝑊 + 𝑠𝐷𝑎𝑙,𝑊 + 𝑠𝑆𝐴,𝑊

= 4.612 + 2.632 + 2.022 = 5.68 ≈ 5.14 units

In both cases the result obtained is similar to the one calculated from data; the differences
are caused by random sampling error.
Characteristics of Risk Pooling
From the case study we have learned the following characteristic of risk
pooling:
1. Centralizing inventory reduces both safety stock and average inventory in
the supply chain.
2. The higher the coefficient of variation, the greater the benefit obtained
from centralized systems. This is because the average inventory depends
on the order quantity and the safety stock. Most of the reduction in
inventory comes from the safety stock, therefore, the larger the CV the
larger the reduction in inventory.
3. The effect of risk pooling is diminished if the individual demand streams
are positively correlated because, if they are, the demands will tend to
go up and down simultaneously rather than in opposite directions.
Positive correlation attenuates the variance reduction effect.
Centralized versus decentralized Inventory
system trade-offs
Attribute Trade-off
Safety stock Safety stocks decrease with centralized inventory systems.
Service level For the same level of inventory, centralized systems offer a better service level.
Overhead costs Overhead costs tend to be much higher in decentralized system because of
reduced economies of scale.
Customer lead time With decentralized inventories, the product can be stocked closer to the customer
resulting in faster response times to the customer.
Transportation costs It depends on the geographic location of warehouses, manufacturers and markets.
Inbound transportation costs tend to decrease, while outbound transportation
costs tend to increase; which one is greater depend on the specific situation.
Assignment: Risk Pooling Case Study
Use the Case Study data provided in a spreadsheet file to answer the following questions
1) Verify the calculations for Products V and W in Table INVENTORY POLICY CALCULATIONS by
solving the two rows for Dallas. Your answer must include your solution procedure.
2) Define risk pooling and explain why it works.
3) Explain the relationship between risk pooling and centralized inventory systems.
4) Why does risk pooling work better when there is high variability?