Churn prediction & prevention

In fiercely competitive telecom and banking markets, winning a new customer costs five to seven times more than retaining one. Yet most retention campaigns spray broadly: they hand discounts to customers who would have stayed anyway, and miss those genuinely on their way out.

Predicting who will leave is not enough. The real question is who, if contacted, will change their mind — and who, conversely, will be annoyed by the outreach. Without that distinction, retention budget is wasted and churn keeps eroding recurring revenue.

ADST puts in place a two-stage retention system. A first churn-propensity model estimates, for each customer, the probability of leaving within 30 to 90 days based on usage, billing, customer-service interactions and dissatisfaction signals.

Crucially, we add an uplift modeling layer that measures the causal effect of a retention offer: instead of targeting at-risk customers, we target the "persuadables", those whose behaviour actually changes thanks to the intervention. Trained on historical A/B campaigns, this approach avoids wasted discounts and counter-productive effects.

Each at-risk customer is matched to the best next action — pricing offer, proactive call, upsell — and a channel, all orchestrated in your CRM with tracking of the true incrementality of campaigns.