Service-oriented performance of inventory models with partial information on unimodal demand lead-time distributions

Abstract

Facing uncertainty in demand, companies try to avoid stock-outs by holding safety inventories, depending on a pre-set customer service level. The knowledge of the demand distribution during lead-time serves to determine the safety inventory level. Many times the distribution is not fully known, except maybe for its range, mean or variance. However literature shows that the performance of holding safety stock strongly depends on the characteristics of the distribution. One option is to protect against the worst case distribution given some information like range or moments. But this worst case is a two-point distribution, bringing unbelief to managers that such an occurrence would ever appear. Mostly they share the opinion that the demand distribution is unimodal. This research develops a technique to derive the safety stock for unimodal demand distributions of which the mode either is known or can be estimated. In this way, the managers obtain solutions to the decision problem including a higher belief that the related type of distribution might appear in practice.