This paper explores the performance of a multi-item production-inventory system. We compare two alternative policies, representing different modes of collecting and utilizing information. We derive a closed-form measure of performance for one of them, the familiar first-come-first-served (FCFS) policy, and propose a comparable approximation for the other, the longest-queue (LQ) policy. These results are illustrated and tested through simulations. In this way we address several basic managerial issues: What is the value of centralized information in complex systems? How does the breadth of the product line affect performance?