Explaining Customer Dissatisfaction Patterns in Express Shipping with Representation Learning

Identifying root causes of customer dissatisfaction within express logistics operational data is, in the era of e-commerce, fundamental to business success. This problem involves identifying the main factors causing negative customer reviews to enable logistics companies to address them. This paper introduces a methodology combining customer reviews with operational shipment data, employing representation learning on the nodes of gradient boosted trees and clustering techniques to segment customers into regions dense of detractors. By means of such regions, our analysis uncovers distinct factors driving customer dissatisfaction, providing actionable insights validated through multiple machine learning models and explainability techniques.