Optimal Policies in a Dual Sourcing Inventory System With Multiple Demand Classes and Supply Disruption Risks

ABSTRACT In this paper, we examine a single‐product dual‐sourcing inventory system with multiple demand classes and supply disruption risks over a finite time horizon under either back‐order or lost‐sales setting. To characterize the structures of optimal ordering and allocation decisions, we build a multiperiod stochastic inventory model and show that the optimal cost functions are multimodular in the system state. The optimal ordering decision is given by a base‐stock policy with two state‐dependent base‐stock levels, which are increasing in the backorder quantities of all demand classes. Specifically, the optimal base‐stock levels are more sensitive to the high‐priority demand class than to the low‐priority demand class. The optimal allocation decision for the highest‐priority demand class is to satisfy its demand requirement as much as possible, and the optimal allocation decisions for the other demand classes are given by a smart dynamic rationing policy. For each demand class, there are two thresholds and one dynamic rationing level. The rationing levels decrease with an increase in backorder quantities. Based on the structure of the optimal policy, we propose new heuristic ordering and allocation policies and numerically verify that they perform well.