EASILY IMPLEMENTABLE INVENTORY CONTROL POLICIES

This work was initiated and supported by a manufacturer of mail proces sing equipment, which stocks 30,000 distinct parts in a distribution c enter to support field maintenance of their equipment. To find an effe ctive stocking policy for this system we formulate a constrained optim ization model with the objective of minimizing overall inventory inves tment at the distribution center subject to constraints on customer se rvice and order frequency. Because size, integrality, and nonconvexity make this problem intractable to exact analysis, we develop three heu ristic algorithms based on simplified representations of the inventory and service expressions. These lead to what we call easily implementa ble inventory policies, in which the control parameters for a newly in troduced part can be computed in closed form without reoptimizing the rest of the system. Numerical comparisons against a lower bound on the cost function show that even our simplest heuristic works well when a high service level is required. However, we show that a more sophisti cated heuristic is more robustly accurate. We also compare our heurist ics to methods previously in use by the firm whose system motivated th is research and show that they are more efficient in the sense of atta ining the same customer service level with a 20-25% Smaller inventory investment. Finally, we discuss implementation issues related to the s pecific needs of the client firm, such as how to handle parts with no or low recent usage and dynamically changing demand for parts.