Effective Dual Sourcing with a Single Index Policy

We consider a single stage periodic-review inventory system with backorders and dual sourcing: Replenishment can occur through a regular channel and/or a more expensive expedited channel with a smaller lead time. Globally optimal policies for these problems are in general highly complex; we therefore introduce a simple class of heuristic base stock policies. We call our class of policies the single index policies, as they require maintaining only one system statistic. We derive simple expressions for optimal single index parameters, enabling us to analytically find the optimal single index policy (by solving a newsvendor problem) in less than a second. Despite its simplicity, we establish computationally that the single index policy can be highly effective, especially on large problems. In this case it is comparable in performance to the more complex dual index heuristic, which has been shown in the literature to perform within a few percent of the optimal solution. Moreover, we demonstrate when dual sourcing can achieve significant savings compared to single sourcing, and establish that for our model single sourcing with the expedited supplier is never optimal.