Modeling UX Frictions and Churn Intent in Digital Food‐Delivery Ecosystems

ABSTRACT The rapid growth of mobile e‐commerce has intensified competition for user retention, making customer churn a critical concern for digital platforms. User reviews contain rich signals about experience failures, yet existing approaches struggle to link specific UX frictions to churn outcomes. This research introduces a data‐driven framework that combines text classification, topic modeling and regression analysis to quantify how distinct friction themes influence churn intent. Churn intent is operationalized as explicit abandonment cues extracted from reviews, identified in 1.3% of all feedback and 13.4% of negative reviews. Of the 19 625 reviews retained after topic assignment, 494 carry explicit churn signals and serve as the dependent variable in all logistic models. Using BERTopic, 20 latent UX friction themes are derived, and logistic regression estimates their marginal effects on churn probability. The findings offer managers a prioritization model for addressing high‐impact UX issues. By treating churn intent as an emergent outcome of interacting experiential and operational subsystems, the study contributes a socio‐technical perspective on user attrition in digital service ecosystems.