Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Amazon Locker Capacity Management

Samyukta Sethuraman, Ankur Bansal, Setareh Mardan, Mauricio G. C. Resende, Timothy L. Jacobs

TL;DR

The paper tackles capacity management for Amazon Locker under the challenge of unknown package dwell times. It combines ML-based demand forecasting (random forests) with dwell-time probability estimation (random forests with isotonic calibration) to feed a linear program that optimally reserves capacity for different ship options and maximizes throughput over a 7-day horizon. The approach yields substantial gains, including a reported 9% year-over-year improvement during 2018 holidays and, in a two-week test, an average throughput increase of about $6\%$ across lockers, with up to $23\%$ in some cases, while reducing unjustified rejections. The work demonstrates a scalable, data-driven yield-management solution for parcel lockers with broad applicability to other capacity-constrained, time-sensitive delivery systems.

Abstract

Amazon Locker is a self-service delivery or pickup location where customers can pick up packages and drop off returns. A basic first-come-first-served policy for accepting package delivery requests to lockers results in lockers becoming full with standard shipping speed (3-5 day shipping) packages, and leaving no space left for expedited packages which are mostly Next-Day or Two-Day shipping. This paper proposes a solution to the problem of determining how much locker capacity to reserve for different ship-option packages. Yield management is a much researched field with popular applications in the airline, car rental, and hotel industries. However, Amazon Locker poses a unique challenge in this field since the number of days a package will wait in a locker (package dwell time) is, in general, unknown. The proposed solution combines machine learning techniques to predict locker demand and package dwell time, and linear programming to maximize throughput in lockers. The decision variables from this optimization provide optimal capacity reservation values for different ship options. This resulted in a year-over-year increase of 9% in Locker throughput worldwide during holiday season of 2018, impacting millions of customers.

Amazon Locker Capacity Management

TL;DR

The paper tackles capacity management for Amazon Locker under the challenge of unknown package dwell times. It combines ML-based demand forecasting (random forests) with dwell-time probability estimation (random forests with isotonic calibration) to feed a linear program that optimally reserves capacity for different ship options and maximizes throughput over a 7-day horizon. The approach yields substantial gains, including a reported 9% year-over-year improvement during 2018 holidays and, in a two-week test, an average throughput increase of about across lockers, with up to in some cases, while reducing unjustified rejections. The work demonstrates a scalable, data-driven yield-management solution for parcel lockers with broad applicability to other capacity-constrained, time-sensitive delivery systems.

Abstract

Amazon Locker is a self-service delivery or pickup location where customers can pick up packages and drop off returns. A basic first-come-first-served policy for accepting package delivery requests to lockers results in lockers becoming full with standard shipping speed (3-5 day shipping) packages, and leaving no space left for expedited packages which are mostly Next-Day or Two-Day shipping. This paper proposes a solution to the problem of determining how much locker capacity to reserve for different ship-option packages. Yield management is a much researched field with popular applications in the airline, car rental, and hotel industries. However, Amazon Locker poses a unique challenge in this field since the number of days a package will wait in a locker (package dwell time) is, in general, unknown. The proposed solution combines machine learning techniques to predict locker demand and package dwell time, and linear programming to maximize throughput in lockers. The decision variables from this optimization provide optimal capacity reservation values for different ship options. This resulted in a year-over-year increase of 9% in Locker throughput worldwide during holiday season of 2018, impacting millions of customers.
Paper Structure (11 sections, 3 equations, 9 figures)

This paper contains 11 sections, 3 equations, 9 figures.

Table of Contents

  1. Introduction
  2. Legacy Practice
  3. Locker Capacity Management Model
  4. Locker Demand Forecast
  5. Package Dwell Time Probability Estimation
  6. Ship Option Capacity Reservation Optimization
  7. Model Evaluation by Simulation System
  8. Experimental Results
  9. Simulations
  10. Lockers in Production
  11. Conclusion

Figures (9)

  • Figure 2: Three distinct Locker Capacity Management disciplines. On top excess capacity is reserved for expedited shipments and locker is underutilized. In the middle, the first-come-first-served (FCFS) rule is implemented and customer ordering expedited cannot ship to locker. An optimal capacity reservation scheme is shown in the bottom.
  • Figure 3: Ratio of number of Standard to Two-Day packages delivered to home and locker Ruby over a ten-week period in 2018.
  • Figure 4: Locker Capacity Management Model overview
  • Figure 5: Baseline (proportion rule) and random forest forecasts versus actual number of packages delivered to Amazon Locker Ruby (top) and Lambda (bottom)
  • Figure 6: Simulation system accuracy metrics
  • ...and 4 more figures