Cross-vs. In-Region Courier Routing in On-Demand Delivery

Abstract

Problem definition: On-demand delivery platforms have expanded rapidly. A key advantage often attributed to their success is the ability to optimize courier routing by sharing couriers across locations and among multiple vendors (e.g., restaurants and grocery stores). However, in practice, couriers on food delivery platforms tend to work in the same regions, thus limiting the potential benefits of cross-region operations.

Methodology: To examine whether and how sharing couriers across regions benefits a firm, we develop a spatial queuing model where couriers serve as servers, and the delivery time depends on a customer’s location, their chosen vendor, and the dispatch policy. 

Managerial implications: Surprisingly, we find that in many cases, the in-region policy—under which couriers are dedicated to specific vendors—can outperform the cross-region policy, which uses a shared fleet to serve multiple vendors. This result stems from the randomness and potential imbalance in courier allocation that cross-region routing introduces. With an exogenous delivery fee, commonly used in the early stages of market penetration, the cross-region policy is more profitable when the market size is sufficiently large; otherwise, the in-region policy is superior. However, with an endogenous delivery fee—more typical for well-established firms who adjust the delivery fees to influence demand—the in-region policy is even more likely to be optimal. In this case, for the cross-region policy to be preferable, not only a large enough market size is required but also customers need to have a sufficiently high valuation for the service. This is because firms with pricing flexibility may prefer to raise delivery fees and reduce the frequency of distant orders, thereby diminishing the benefits of cross-region routing. Furthermore, we show that the cross-region policy tends to perform better in markets where restaurants offer highly distinctive cuisines, whereas the in-region policy is more advantageous when restaurants have similar offerings. Finally, we extend our base model to consider multiple couriers, optimal location of vendor hubs, and vendors maintaining long backlogs.