Using forecast updates and risk-sharing agreements in a three-echelon supply chain

This dissertation explores coordination schemes that can be used to align the incentives of an original equipment manufacturer (OEM) and its contract, manufacturer (CM). The coordination is mediated by demand forecast sharing and component overage risk sharing. Three major problems are studied: (1) two-component newsvendor model in which the CM needs to make ordering decisions on two complementary components at two different stages with new demand information being revealed in between; (2) an OEM-CM joint planning problem in which the CM makes its own purchasing decisions while the OEM can influence the CM's decisions through information-sharing or risk-sharing mechanisms; and (3) a two-mode problem in which a company can split its freight between a slow mode and a fast mode and needs to optimize its shipping decisions using demand forecast updates.; We use a set of two related uniform distributions (U-U) to develop a new model for demand forecast updating. The U-U formulation models demand in the second stage as a uniform random variable whose mean is unknown in the first stage, although its width is known in both stages. To capture risk sharing between the OEM and the CM, we propose a contract under which the OEM compensates the CM for unused components.; Using the U-U model, we first study the CM's planning problem in isolation and obtain an in-depth understanding of the CM's ordering behaviors. The CM tends to increase the order quantity of the component with a longer lead time if better information becomes available at the second stage. Then we investigate the OEM-CM joint planning problem based on different combinations of information sharing and risk sharing. We find that information sharing is not necessarily a substitute for risk sharing. Through information sharing alone, both the OEM and the CM are better off; however, information sharing, when combined with risk sharing under certain agreements, could hurt the OEM's performance. Finally, we apply the U-U model to the two-mode problem. Computational results suggest that better information helps to improve both service and cost performance.