Questioning the relative virtues of dynamic lot sizing rules

This paper begins by relating a brief history of lot sizing to a review of current educational literature, identifying the patterns and inconsistencies concerning which rules are used for demonstration. To independently establish the relative virtues of this broad range of techniques, nine such rules are selected and subjected to 3060 rolling horizon simulation tests each. The results of this comprehensive deterministic study are succinctly summarized, demonstrating not only that various published lot sizing rules vary in terms of cost performance, they also possess distinct strengths and weaknesses with respect to sparse demand patterns, short versus longer planning horizons, and degree of nervousness. In the case of nervousness, some rules are found vulnerable to an intriguing harmonic effect created by the apparent interaction between order cycle length and planning horizon. The study also found that one of the most impressive lot sizing techniques with respect to all these factors is a simple rule published in 1981 but remarkably overlooked in lot sizing literature since.