The goals of this research project are: (1) to build computer-based mathematical models that explicitly incorporate the assemble-to-order nature of a system; and given a fixed model, (2) to develop efficient computational procedures for evaluating order-based performance measures under any given inventory-control policy, and (3) to explore the form of the optimal policy in the order-based sense. The research will consist of modeling, mathematical analysis, and computation. The basic methodologies underlying the approaches include the analysis of multivariate stochastic processes, dynamic programming, and the stochastic comparison techniques, especially those based on notions of correlations of random variables. The analyses and results of the investigation will contribute to the understanding of the intricate relationships underlying the system and therefore will provide constructive suggestions on how to manage such systems effectively. This, in turn, will enable manufacturers and distributors to respond more quickly to customer needs at lower cost.