Real insights and intelligence brought alive by human expertise
Case study : :Churn Risk Prediction – Leading Telecom Operator in Southeast Asia
BACKGROUND
- The client, a major telecom operator with 45 million subscribers, was experiencing monthly churn rates of 2.8%, significantly above industry benchmarks.
- Existing retention programs were reactive and untargeted, resulting in high costs and low effectiveness. The client needed to identify at-risk customers before they churned and understand the underlying drivers.
METHODOLOGY
- Integrated analysis combining internal CRM and usage data with primary research (n=3,500 quantitative survey + 24 in-depth interviews with churned customers).
- Machine learning models (Random Forest, XGBoost) trained on 18 months of historical data to predict churn probability. Feature importance analysis identified key behavioral and attitudinal drivers.
- Desk research on competitor offers and market dynamics provided context for churn triggers. Model validated over 3-month pilot period.
OUTCOME
- Targeted retention campaigns reduced churn by 23% among high-value segments. Estimated annual savings of $4.2 million in customer acquisition costs. The model is now deployed for real-time scoring of the entire subscriber base.
- Client was able to identify key areas to focus on from NPD and GTM messaging perspectives and take actions.