Collaborative and cooperative agreements in the supply chain.

This thesis models several collaborative and cooperative supply chain agreements that occur in practice. Specifically, we model two different implementations of vendor-managed inventory (VMI) agreements, as well as the sharing of demand information between multiple retailers. In each case we find the optimal production and inventory policies for the supply chain agents. Numerical analysis is used to compare the effectiveness of each agreement by comparing it to completely centralized and decentralized settings. The first VMI setting involves a single retailer subject to stationary customer demand and a single manufacturer following a periodic production schedule. The agreement is subject to an initiating contract, which states minimum and maximum inventory levels that the manufacturer is expected to maintain at the retailer or face penalty charges. We investigate the savings due to better coordination of production and delivery facilitated by such an agreement. The optimal replenishment and production strategies for the supplier are found to be up-to policies. A simple application of the newsvendor relation is used to define the retailer's policy. The second VMI setting includes multiple retailers and nonstationary customer demand. The initiating contract evaluates the manufacturer's performance based on the retailers' customer service level and inventory turns. Optimal replenishment and production policies for both a centralized controller case and the VMI case are shown to be up-to policies. We also model demand information sharing between multiple retailers and allow for forecast updating. We show that history dependent up-to policies are optimal in the presence of forecast updating. We propose a specific scheme for forecast updating that utilizes a simple Kalman filter for the single retailer and the multiple retailer cases. A numerical study examining both VMI settings compares the performance of the supply chain operating under VMI with a supply chain operating in centralized and decentralized settings. Our results verify some observations made in industry about VMI and show that VMI can perform significantly better than decentralized settings in many cases, but can perform worse in others. We find that the design of the initializing agreement can have an important effect on the success of such programs.PhDApplied SciencesIndustrial engineeringOperations researchUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/129466/2/3042071.pd