Customer churn prediction for fixed wireless access: The case of a regional internet service provider

Abstract

Customer churn remains a significant area of concern in the telecommunications sector due to its high prevalence. As Fixed Wireless Access (FWA) gains traction, the field is witnessing heightened competition, especially with the integration of 5G technology, which serves dual purposes for cellular communications and FWA. This research delves into a comprehensive analysis of risk factors that impact churn via Random Survival Forests, the Cox regression model, and Kaplan-Meier Survival Plots. Our results indicate that fiber, sales zone (i.e., brand presence), age, speed, and discounts are some of the more important variables with regard to the cumulative hazard ratio. Although discounts were effective in many sales zones, their impact was diminished for more expensive internet services. The insights garnered provide a foundational basis for telecom providers to devise robust strategies to curtail customer churn over time, ensuring sustained growth and customer satisfaction in the burgeoning FWA market.