Transshipment policies: A cost benefit model for inventory managers

Inventory managers often support each other in disperse locations by using transshipments to satisfy known demand. Research to date has shown that transshipments reduce the variance of demand during lead time and increase fill rate. The research has often incorporated numerous assumptions to derive analytical solutions to determine the benefits of transshipments. We relaxed many of the simplifying assumptions and evaluated a multi-echelon inventory system with stochastic lead times and stochastic demand. We evaluated five transshipment options for the inventory manager within the system.;We first developed a simulation model of the inventory system with two distribution centers and six retail centers. We defined five transshipment options as the baseline case of no transshipments, the partial pooling policy of holding safety level plus expected demand, the partial pooling policy of holding demand, the partial pooling policy of holding safety level, and the complete pooling policy of shipping to zero inventory level. Then we evaluated the inventory system for each policy to determine the performance characteristics of the system. The performance characteristics were then evaluated in a total cost model to determine the lowest total cost policy in the system. The all pairwise comparison and the multiple comparison with all evaluation techniques were used to determine which policy is statistically significant. The only significant policy was the baseline case of no transshipments.;The results of the total cost model were then evaluated using sensitivity analysis in which we found that the determination of the lowest cost policy is most sensitive to the penalty cost. The assumption of the cost of a stockout drives the selection of a transshipment policy. When the penalty cost is seven times the value of the item being evaluated the complete pooling policy becomes the lowest total cost policy.;The total cost results of the simulation and the total cost model were then used to develop a "meta-model" that can be used to evaluate which transshipment policy to use for a specific part without having to perform a number of simulation runs. The meta-model was then evaluated using an actual item to determine if the model can be used by inventory managers. We found that the meta-model was useful in determining the potential cost of the possible transshipment policies.