doi: 10.18178/joebm.2026.14.4.962
Predicting Telecom Customer Churn and Translating Risk into Retention Action: Evidence from the Orange Telecom Dataset
Jiayi Wan
- School of Mathematics, The University of Edinburgh, Edinburgh, UK
- Manuscript received August 2, 2026
- accepted October 9, 2026
- published October 27, 2026
Abstract
Telecom churn models can identify customers who are likely to leave, but their predictions do not always indicate which customers should receive retention efforts first. This study investigates whether churn predictions can be used to distinguish customers with different retention priorities. Using the Orange Telecom Customer Churn Dataset, logistic regression and random forest models with balanced class weights were compared on the test set based on accuracy, precision, recall, F1-Score and ROC-AUC. Among the examined variables, higher customer service calls, daytime usage and international-plan subscription were positively associated with churn, whereas voice-mail-plan subscription was negatively associated. Random forest achieved higher precision, F1-Score and ROC-AUC than logistic regression while maintaining similar recall, mainly due to fewer false-positive predictions. The predicted probabilities separated customers into four risk tiers, with observed churn increasing from 2.4% and 3.2% in the Low- and Medium-risk tiers to 37.8% and 98.5% in the High- and Critical-risk tiers. High-risk customers had more customer service calls on average, whereas Critical-risk customers had a higher proportion of international-plan subscribers. These findings suggest that combining churn probabilities with customer characteristics helps translate predicted churn risk into more targeted retention actions.
Keywords
- customer churn
- telecom industry
- machine learning
- customer retention
- random forest
How to Cite
Jiayi Wan, "Predicting Telecom Customer Churn and Translating Risk into Retention Action: Evidence from the Orange Telecom Dataset," Journal of Economics, Business and Management, vol. 14, no. 4, pp. 312-316, 2026. https://doi.org/10.18178/joebm.2026.14.4.962
Copyright & License
Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).