Incentives and Flexibility in a Decentralized Multi-Product

Assemble-to-Order System

 

Fernando Bernstein

Duke University

Durham, NC

 

In this paper, we explore the impact of decentralized decision making on the behavior of assemble-to-order systems. Specifically, we consider a system where three components (two product-specific and one common) are used to produce two end products to satisfy stochastic customer demands.  Component capacity decisions must be made before a single selling season, while end-product assembly decisions are made after observing demands. We study the system under both centralized and decentralized decision making.  In the latter case, an assembler chooses wholesale prices to pay three independent suppliers and also makes end-product production decisions, while the suppliers choose component capacities. In the decentralized system we prove that, for any wholesale prices, there exists a unique Pareto-optimal equilibrium in the suppliers' capacity game. We show that the assembler's optimal wholesale prices lie in one of two regions – one leads to capacity imbalance and one does not. Similar to other decentralized supply chain settings, we find that decentralization leads to understocking in terms of component capacities. However, we also identify new types of inefficiencies related to the multi-component, multi-product setting studied here. We find that capacity imbalance occurs less frequently in the decentralized system, and that its presence in that system depends on the distributions of end-product demands (unlike the centralized system). In addition, we demonstrate that in some situations the wholesale prices in the decentralized system can alter the assembler's profit margins so that end-product production priorities are reversed from those in the centralized system. Finally, by comparing the decentralized system to one where the common component is replaced by two dedicated components, we find that the apparent flexibility provided by a common component may actually hurt the assembler's performance in a decentralized system.

 

 

Back to Presenters’ Page