Applications of risk pooling for the optimization of spare parts with stochastic demand within large scale networks

Abstract

Amazon is able to deliver millions of packages to customers every day through its Fulfillment Center (FC) network that is powered by miles of material handling equipment (MHE) such as conveyor belts. Unfortunately, this reliance on MHE means that failures could cripple an entire FC. The exceptionally high stock-out cost associated with equipment failure means spare parts must always available when required. This is made difficult as Amazon does not hold any central repository of inventory at present - all inventory is held at a site-level. Unfortunately, FCs have to stock more inventory than required due to unpredictable failures, long lead times from suppliers, and no standard work processes for site-to-site transfers. However, if Amazon is able to pool its spares across multiple FCs, it has an opportunity to reduce the spares kept across the entire FC network, position itself to better respond to catastrophic failures, and consolidate interfaces with suppliers. The goal of this thesis is to identify the inventory model and network design that would maximize parts availability while minimizing cost. Additionally, an implementation roadmap will be developed to outline how such a system (e.g. hub locations, logistic channels etc.) can be developed. This thesis concludes by proposing potential extensions of the work conducted in this thesis to improve the practicality and financial impact of the proposed network and inventory model.