EURO 2025 Leeds
Abstract Submission

2934. Optimization of Order Allocation Decisions in Online Marketplaces Considering Vendor Satisfaction

Invited abstract in session TA-47: Retail Inventory Management, stream Retail Operations.

Tuesday, 8:30-10:00
Room: Parkinson B08

Authors (first author is the speaker)

1. Sérgio Castro
INESC TEC
2. Willem van Jaarsveld
Eindhoven University of Technology
3. Gonçalo Figueira
INESC-TEC, Faculty of Engineering of Porto University
4. Bernardo Almada-Lobo
INESC-TEC, Faculty of Engineering of Porto University

Abstract

Online marketplaces connect suppliers with customers by providing a platform for vendors to sell their products worldwide. Order allocation directly impacts fulfillment costs and customer satisfaction, while also determining the volume of business received by suppliers, a critical determinant of supplier retention to the platform. We introduce and study the Multi-Item Order Fulfillment Problem in Online Marketplaces, which explicitly considers the perspective of suppliers via sales targets to be achieved over the selling season.
As solution methods, we adapt various myopic rule-based policies popular in industry, along with two randomized policies with attractive properties proposed by Ma (2022). We moreover propose to apply deep reinforcement learning (DRL) to the problem, leveraging a tailored MDP formulation that enables the application of the Deep Controlled Learning algorithm.
We test these policies on a comprehensive set of problem instances developed based on extensive interactions with an online marketplace. In the absence of supplier sales targets, the policies by Ma perform particularly well when demand over the selling season is relatively high, while DRL is competitive across a wide range of instances, especially when season demand is relatively low. Moreover, when supplier sales targets are considered, the proposed DRL approach achieves balanced outcomes that ensure supplier satisfaction while keeping fulfillment costs low and customer satisfaction high.

Keywords

Status: accepted


Back to the list of papers