Estimating a customer churn model in the ADSL industry in Portugal: The use of a Semi-Markov model

Abstract

Customer churn has been stated as one of the main reasons of profitability losses in the telecommunications industry. As such, it seems critical to have an a priori knowledge about the risk of a given customer to churn at any moment, in order to take preventive measures to avoid the defection of potentially profitable customers. This study intends to develop a duration model of the residential customer churn in this industry in Portugal. We found empirical evidence that the variables that influence customer churn are the total number of overdue bills since ever, average monthly spending, average value of additional internet traffic, payment method, equipment renting, and the subscription of a flat plan. We also found that the probability of a customer to churn is neither constant over time nor across customers.