Optimized supply routing at Dell under non-stationary demand

Abstract

This thesis describes the design and implementation of an optimization model to manage inventory at Dell's American factories. Specifically, the model is a mixed integer program which makes routing decisions on incoming monitors (a bulky item which incurs great shipping costs) from Asia to Dell's factories in America as well as inventory transfer decisions from factory to factory. The optimization model approaches the inventory allocation problem by minimizing inventory routing costs plus shortage costs across all sites subject to constraints which define the specifics of Dell's supply chain. Shortage costs are assessed using a per part per day back order penalty, however a more precise assessment of shortage costs using actual costs from a combined MIT/Dell study is also presented. The software implementation of the optimization model has been field tested and validated and is now being adopted on a global level for use in balancing supply to all of Dell's factories worldwide. The software design as well as the implementation results are discussed within this thesis. Also, an adaptation of the model to a global scale is presented. This extension of the model, which assumes a "global warehouse" upstream in the supply chain, allocates inventory from the China to regional facilities throughout the world subject to supply chain constraints and the understanding that regional teams will tend to balance out their own region's inventory using intraregional balancing decisions.