Factors, Predictability, and Explainability of Mobile Telephony Customer Departure in Telecommunications Companies: A Systematic Review of the Literature

Abstract

The telecommunications sector has experienced exponential growth since the year 2000, reaching 5.31 trillion users by 2022, generating 1.07 trillion in revenue for telecommunications companies. However, this growth leads to intense competition among companies for customer acquisition, often resulting in customer churn or switching between telecommunications providers due to the services or experiences received by users. At the moment, algorithms are being developed for churn prediction to assist in taking actions to prevent customer defection. Nevertheless, there is no comprehensive inventory of churn factors or a list of algorithms that explain the reasons for customer churn, despite the existence of numerous factors influencing user retention or defection. This highlights the importance of our work in allowing us to know in advance which client will be a deserter (churn). The purpose of this research is to identify how factors intervene in the customer churn process in telecommunications companies, as well as to list existing techniques for prediction and finally, to present advances in explainability, using a systematic literature review from 2018 to August 2023 using the Scopus and Wos meta-search engines. Regarding the factors contributing to churn, we identified 19 factors encompassing 87 sub-factors, detailed in 87 out of 112 reviewed articles. In terms of prediction, 26 unique algorithm techniques and 16 combinations were identified, presented in 102 out of 112 reviewed articles. Finally, in the realm of explainability, 4 out of 112 articles were found, detailing 5 techniques that help identify why an algorithm selects a customer as potential defector.

Keywords

• Churn
• telecommunications
• prediction
• machine learning
• explainability
• literature review