Predicting Churn in Mobile Money

East Africa

Client Brief

Churn is hard to define and assess in Mobile Money, as the customer can stop using the service without unsubscribing. Some users can also stop using the service for a while, just not being regular users, and are not churners
Predicting who is about to churn is key for MNOs to retain their clients by conducting targeted marketing actions

Solution

After defining clearly what was considered as churn with the MNO, Masae developed a prediction model, using past transactional data, GSM usage and KYC data (socio-demographic information on customers) to predict the likelihood of each user to churn in several time horizons (in 1, 3, 6 months...)

Deliverable

Source code of the algorithm that allows the MNO to apply the prediction model internally on new data to conduct real-time prediction