Global versus local perspectives on appropriate inventory levels within multiple stage systems

This study examines two perspectives on desirable inventory levels at various stages of a multiple stage system: that of a single decision maker determining ordering schedules for all stages versus separate decision makers seeking to minimize costs at their respective stages. These `global' versus `local' planning paradigms focus on the role of inventory as a means of achieving economies of scale, assuming dynamic, deterministic customer demand, a rolling planning horizon and constant lead-times. The single decision maker is simulated with the use of lot sizing algorithms that require the use of necessary predecessor cost information when scheduling at any stage, while separate decision makers are represented by the sequential application of a single item lot sizing algorithm. In simulation experiments, single decision maker `global' solutions were an average of 22% lower in total cost, yet were associated with an average of 1,637% higher inventory levels for the end item stage, where holding costs are highest, and an average of 37% lower inventory levels for raw material stages. These solutions, counter-intuitive to local planners, averaged within 2% of a lower bound on the optimal solution.