Evaluating on-demand warehousing via dynamic facility location models

Abstract

On-demand warehousing platforms match companies with underutilized warehouse and distribution capabilities with customers who need extra space or distribution services. These new business models have unique advantages, in terms of reduced capacity and commitment granularity, but also have different cost structures compared with traditional ways of obtaining distribution capabilities. This research is the first quantitative analysis to consider distribution network strategies given the advent of on-demand warehousing. Our multi-period facility location model – a mixed-integer linear program – simultaneously determines location-allocation decisions of three distribution center types (self-distribution, 3PL/lease, on-demand). A simulation model operationally evaluates the impact of the planned distribution strategy when various uncertainties can occur. Computational experiments for a company receiving products produced internationally to fulfil a set of regional customer demands illustrate that the power of on-demand warehousing is in creating hybrid network designs that more efficiently use self-distribution facilities through improved capacity utilization. However, the business case for on-demand warehousing is shown to be influenced by several factors, namely on-demand capacity availability, responsiveness requirements, and demand patterns. This work supports a firm’s use of on-demand warehousing if it has tight response requirements, for example for same-day delivery; however, if a firm has relaxed response requirements, then on-demand warehousing is only recommended if capacity availability of planned on-demand services is high. We also analyze capacity flexibility options leased by third-party logistics companies for a premium price and draw attention to the importance of them offering more granular solutions to stay competitive in the market.

Keywords:

• Warehouse
• distribution
• on-demand
• supply chain network design
• facility location models

Data availability statement

The reproducibility report is available in the supplemental material; Data and computer files that support the findings are available at https://jenpazour.wordpress.com/research-2/

Additional information

Funding

This work was partially funded by the National Science Foundation #1751801.

Notes on contributors

Kaan Unnu

Kaan Unnu is a Data Scientist of the Development Supply Chain group at Amgen. He received his PhD degree in decision sciences and engineering systems from Rensselaer Polytechnic Institute in 2020.

Jennifer Pazour

Jennifer Pazour is an associate professor and the undergraduate coordinator of industrial and systems engineering at Rensselaer Polytechnic Institute. Her research and teaching develop and use mathematical models to guide decision making for logistics and supply chain challenges.