Use of the Left-Truncated Normal Distribution for Improving Achieved Service Levels

摘要

To determine the safety stock required to achieve a desired service level, demands are commonly assumed to follow a normal probability distribution. However, since demands are necessarily truncated at values below zero, this paper will show that this approach underestimates the actual demand level and results in achieved service levels that are less than those targeted. Use of a left-truncated normal probability distribution - i.e., a normal probability distribution in which values below a truncation point cannot be observed - for demand to determine safety stocks will be shown to improve the achieved service level. This paper describes a methodology for using standardized, left-truncated normal distributions to determine safety stocks, presents tables for use in implementing this methodology, and discusses the improvements possible through this approach.