Analyzing the Impact of Distribution Center Location on E-Commerce Network Design

Open Access
- Author:
- Jackson, Luke
- Graduate Program:
- Data Analytics (MS)
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- November 08, 2022
- Committee Members:
- Colin Neill, Program Head/Chair
Guanghua Qiu, Thesis Advisor/Co-Advisor
Raghu Sangwan, Committee Member
Youakim Badr, Committee Member - Keywords:
- Supply Chain
Data Analysis
Clustering Model
Geospatial Analysis
E-commerce - Abstract:
- Strong E-commerce demand over the past decade has resulted in market constraints on available land and labor that limit the ideal location selection of new E-commerce facilities. This research investigates the impact that Middle-Mile facility locations have on the network design process. Our research showed that Middle-Mile facility location has an outsized impact on the overall delivery speed and transportation cost of the network. A strong correlation was found between the location of Middle-Mile facilities and the population density of their surrounding Urban Areas. We measured the impact Middle-Mile (Distribution) facility locations have on the upstream First-Mile (Fulfillment) and downstream Last-Mile (Delivery) segment distances. Using these relative distances, we found an optimized design that shifted Middle-Mile facility locations 10 miles closer to their nearest population centers. To test this design, we built a network simulation model to quantify network impact. We constructed network edges and vertices by calculating the minimum end-to-end distances from 739 E-commerce facilities to 481 Urban Areas. Population density was used to represent customer demand and weighted with network distance to give an objective measure of performance. However, the results show that the 10-mile shift approach increased distances between First-Mile and Middle-Mile more than it reduced the distances between Middle-Mile and Last-Mile segments. This resulted in a +3.5% increase in total end-to-end network distance which is cost unfavorable. Despite the poor performance of the population density approach, this research found that the simulation model provides real-world value in what-if scenario testing for network designers and decision makers. Using the simulation model, network designers can quantify the impact adding or removing facilities has on the overall network. We provide a blueprint for future work to further optimize network design using the simulation model, methodologies, and tooling developed in this research.