Unmasking Phantom Churn: Redefining Telco Customer Retention Strategies
Industry: Telecommunications
Challenge: Predictive Modelling for Churn Management and Customer Retention
Solution: Advanced Predictive Analytics and Machine Learning
Technology: Python, Oracle
Situation and Challenge
A subsidiary of a global telecom operator, leading in a highly competitive market, sought to enhance the accuracy of their customer churn models. The existing models had become suboptimal, overstating churn rates and inflating perceived acquisition campaign performance. Crucially, they failed to accurately identify ‘rotational churners’ – customers who churn to exploit new subscriber offers before rejoining, particularly prevalent in the prepaid mobile segment. This phenomenon was driving revenue leakage and escalating acquisition costs.
Insight and Action
Our approach encompassed:
- Comprehensive analysis of call detail records, subscriber profiles, and market insights.
- Development of a fingerprint-based methodology to uniquely identify rotational churners.
- Creation of usage fingerprints combining various data variables from demographic profiles, Call Detail Records (CDR), SIM and subscriber IDs.
- Formulation of scenarios considering relevant historical periods, overlapping SIM usage timeframes, and weighted fingerprint components.
- Deployment of a similarity-based weighted k-Nearest Neighbour (k-NN) learning algorithm.
Key findings:
– 20-30% of supposedly churned customers had remained with the company under different contracts or subscriptions.
– Legacy models had been inadvertently trained on both genuine and rotational churn events.
Our intervention delivered substantial benefits:
– Achieved 95% accuracy in identifying rotational churners.
– Enabled 8% improvement in retention rates for at-risk, high-value customers.
– Reduced marketing and retention campaign spend by 20%.
– Provided deeper understanding of churn triggers, enabling optimised promotional strategies.
– Facilitated retraining of models based on genuine churn events, significantly improving overall churn prediction accuracy.
Call us On +44 (0) 208 004 3015
Related Content.
