Lot-sizing under uncertainty in a rolling schedule environment

SUMMARY This paper presents results from a simulation experiment evaluating 14 different single stage lot-sizing procedures. Uncertainty in the form of forecast errors was used as one factor in the experiment. In order to prevent different service levels from confounding cost comparisons, enough safety stock to achieve a 100% service level in all situations was introduced. The ranking of the rules under uncertainty turned out to be very different from the ranking when no demand uncertainty was present. Statistical differences were found between the six best lot-sizing procedures when there was no uncertainty. When forecast errors were present, however, no differences existed for the six best rules in this situation. Thus, uncertainty not only changed the relationship between the lot-sizing rules, but also the character of this relationship.