LEARNING FROM MISTAKES - A NOTE ON JUST-IN-TIME SYSTEMS

In this paper we seek to reconcile two competing frameworks for invent ory management, namely the classical ''optimization'' framework and JI T. While the first is predicated on the premise that the optimal level of inventory is obtained by balancing the costs and benefits of inven tory, the latter claims that inventory is a form of waste and thus sho uld be eliminated completely (but gradually). We show that some of the operational characteristics of JIT, as opposed to its rhetoric, can b e obtained within the optimization framework provided that the relevan t factors are taken into account. In particular, we show that in an en vironment which is subject to ''learning from mistakes,'' it is better to carry inventory levels which are lower than what is ''operationall y optimal.'' We also study the (complex) relation between variability reduction, inventory levels, and overall cost. In general, reducing th e variability of a system (in the sense of first or second-order stoch astic dominance) will not necessarily result in a lower level of inven tory, although optimal cost will be reduced. We study conditions on va riability improvement that guarantee that variability reductions will always result in reduced levels of optimal inventory.