Using Social Network Information for Predicting Customer Churn

Subscriber churn is a major problem and expense for telecom operators. To reduce churn, operators focus their energies on identifying subscribers that will churn so they can be targeted for retention campaigns. Typical churn prediction algorithms identify churners based on service usage metrics, network performance indicators, and demographic information. Social and peer influence to churn, however, is usually not considered. In this work, we developed a new churn prediction algorithm that incorporates the influence churners spread to their social peers. Using data from a major operator, we show that social influence improves churn prediction and is among the most important factors.