Bayesian Design of Inventory Systems to Minimize Expected Operating Costs

Inventory management requires optimal determination of (a) “when to order”, and (b) “how much to order”. One also needs here two answers—what should be stocked initially, and should one adjust (a) and (b) as time progresses? This paper uses the Bayesian approach to answer these. It uses the classical (s, Q) model and a heuristic search in the unstructured decision space. It finds that one can make a decent start by this approach—and stay the course nearly optimally—while upholding a target service level and minimizing cost/unit time. Study finds high value in continually updating (a) and (b).