The Client
Subsidiary of global telecom operator serving ~288 million customers across 18 markets, operating in highly competitive prepaid mobile environment with complex multi-SIM customer patterns.
The Challenge
Churn models failed to distinguish true churners from rotational customers
Existing churn models overstated churn rates and inflated acquisition campaign performance by failing to distinguish between true churners and rotational customers.
High rotational churn in prepaid markets led to revenue leakage and increased acquisition costs from misidentified customer behaviour.
What We Did
Fingerprint-based predictive methodology
- Developed fingerprint-based methodology combining demographic profiles, call detail records and subscriber data
- Created usage fingerprints with similarity scoring to pair churned and newly activated SIMs
- Applied similarity-based weighted k-Nearest Neighbour algorithm for pattern recognition
- Formulated scenarios considering historical periods and weighted fingerprint components
- Retrained models based on genuine churn events to improve prediction accuracy
The Impact
Optimised retention strategies with reduced campaign spend
95%
Churn prediction accuracy
8%
Retention improvement
20%
Reduction in campaign spend
Achieved 95% accuracy in identifying rotational churners whilst improving retention rates by 8% for at-risk, high-value customers. Reduced marketing campaign spend by 20% through deeper understanding of churn triggers and optimised promotional strategies.
Similar Work
Related Case Studies
Ready to Decode Your Growth Path?
Let’s explore what’s possible for your situation.
Speak with an expert: 0208 004 3015