Dormancy Prediction Model in a Prepaid Predominant Mobile Market : A Customer Value Management Approach

Abstract

Previous studies have predicted customer churn in the mobile indutry especially the postpaid customer segment of the market. However, only few studies have been published on the prepaid segment that could be used and operationalised within the marketing team that are responsible for the management of incident of prepaid churn. This is the first identifiable literature where customer dormancy is predicted along the customer value segmentation. In this article, we use a popular data mining technique to predict when a customer will go dormant or stop performing revenue generating events in a prepaid predominant market. Our study is unique as we considered ~1,451 attributes derived from CDR and SIM registration database (previous studies only considered maximum of ~1,381 potential variables). We built 3 different models for Very High, High and Low value segments. We applied our models on the prepaid base and the output was later compared with the actual dormant customers. Very High segment has the highest accuracy and lift while Low segment has the least at the same threshold. We show that once the problem of prepaid churn is well defined, it can be predicted. We recommend a value segmentation dormancy prediction with decision tree for prepaid segment with a certain threshold. Our study shows that this approach can be easily adopted and operationalised by the campaign management team responsible for the management of prepaid churn in a mobile industry.