Risk Pooling Across Customers
Consider this.
In scenario A, our company has 2 warehouses to serve 2 separate markets respectively. We'll call the first warehouse, WH-A and the second warehouse, WH-B.
WH-A solely distributes to the Michigan market while WH-B solely distributes to the Texas market. This is a case of a decentralized distribution system.
In scenario B, our company has 1 warehouse located between Michigan and Texas. This warehouse will serve both the Michigan and Texas markets. This is a case of a centralized distribution system.
The obvious benefits in scenario A (decentralized system) are the speedy delivery times, faster response time, closer and simpler distribution network, and ability to serve market demand. On the other hand, the benefits in scenario B (centralized system) are the lower safety stock inventory and lower overhead costs.
As we increase the number of warehouses, we reduce the outbound (warehouse --> retailers) transportation costs but we also increase the inbound (manufacturing --> warehouses) transportation costs.
// Side-bar here
The Coefficient of Variation (CV) = (Standard Deviation / Average Demand)
CV measures the variability in relation to the average demand. It is a normalized measure relative to the mean. Standard deviation simply gives you an absolute measure of variability across the mean.
//
What does all this have to do with Risk Pooling? Well, we are able to reduce the variability in demand if we can consolidate demand across multiple locations. The rationale here is that we can potentially offset low demand in Michigan by having high demand in Texas. Vice-versa. This allows the company to have a much lower "safety-stock" of inventory.
The higher the CV, the more variability there is in the market demand. In a simplified high level example, the centralized system would be beneficial if the CV is high. If the CV is low, the benefits of a centralized system would not be as obvious.
One should also consider the behavior of the markets. Michigan and Texas are fairly separated and thus, one can "pretend" that the behaviors of both markets are "negatively correlated". This means the company gets high demand in Texas and low demand in Michigan. Vice-versa. With negative correlation, the benefits from risk-pooling are much greater.