Characterizing and improving the service level agreement at Amazon

Abstract

Amazon's Service Level Agreement (SLA) is a promise to its customers that they will receive their orders on time. At the Fulfillment Center (FC) level, the SLA is based on the capability to fulfill open orders scheduled to ship at each departure time. Each center's capability depends on a complex interaction between fluctuating product demand and time-dependent processes. By lowering SLA, Amazon could provide an enhanced the customer experience, especially for same day delivery (SDD). However, providing additional time to the customer also means that the FCs have less time available to fulfill open orders, placing the customer experience of those orders at an increased risk of a missed delivery. This thesis explores cycle time reductions and throughput adjustments required to reduce the SLA at one of Amazon's Fulfillment Centers. First, a method to analyze time-dependent cycle time is used to evaluate the individual truck departure times, revealing that the current process conditions have difficulty meeting current demand. Then, using lean principles, process changes are tested to assess their ability to improve the current processes and allow for an SLA reduction. Although a 1% increase in capacity is possible by improving the processes, system constraints make the changes impractical for full implementation. Consequently, a capacity analysis method reveals that an additional capacity of up to 9.38% is needed to improve the current process conditions and meet current demand. The capacity analysis also reveals that reducing the SLA from its current state requires up to 13.79% more capacity to achieve a 50% reduction in SLA. Through capacity adjustments, the added cost of late orders is mitigated, resulting in a reduced incidence of orders late to schedule and a reduced risk of missed deliveries. The methods utilized in this thesis are applicable to other Amazon FC's, providing a common capability and capacity analysis to aid in fulfillment operations.