A theoretical validation of the DDMRP reorder policy

A recent heuristic called Demand Driven MRP, widely implemented using modern ERP systems, proposes reorder policy based on buffers. Buffers are amounts of inventory positioned and set to control the net flow position, responding to stochastic demand and lead time. Our primary goal is to propose a theoretical foundation for such a heuristic approach. To this aim, we develop an optimization model inspired by the main principles behind the heuristic algorithm. Specifically, optimal policies are of the type (s(t), S(t)) with time-varying thresholds that react to short-run real orders. We introduce constraints related to the service levels, that are written as tail risk measures to ensure fulfillment of realized demand with a predetermined probability. Interestingly, it turns out that such constraints allow to analytically justify an empirical rule that the DDMRP employs to set the risk parameters used in the heuristic. Finally, we use our model as a benchmark to theoretically validate and contextualize the aforementioned heuristic.