Causal insights and predictive models for order cancellations in online food delivery

Abstract

This study addresses the critical issue of order cancellation in China's competitive online takeout service market, with a particular focus on Meituan. It explores the relationship between service response times, specifically waiting times, and the rate of order cancellations, aiming to develop predictive models to identify and intervene in high-risk orders proactively. The research methodology integrates logistic regression, propensity score matching (PSM), and back propagation neural networks to analyze data from a leading Indian takeout service, Swiggy, highlighting the study's international applicability. Initial findings suggest that extended waiting times significantly enhance the probability of order cancellations, impacting customer loyalty and satisfaction negatively. Through empirical analysis, logistic regression identified specific waiting time stages that critically influence cancellation probabilities. PSM was employed to examine the effects of reassignment on order cancellations, showing that strategic order reassignment could mitigate cancellation risks. Lastly, a back propagation model demonstrated high accuracy in predicting cancellations, suggesting potential for early intervention. This study not only provides a robust statistical framework for understanding and reducing order cancellations but also offers actionable insights for improving operational efficiency and customer retention in the fast-paced takeout industry.