Synthetic profiles
Generate 4 000 fictional customers with tenure, charges, contract type, payment method, internet service, add-ons, and tickets.
Live · Telecommunications
Churn prediction
Configure a synthetic customer profile. A random-forest classifier trained on 4 000 fictional customers returns a churn probability with the feature contributions ranked.
Try a preset
Churn probability
Submit a profile to predict.
How it works
Label generation
Churn risk is a function of tenure, contract type, payment method, ticket count — with a stochastic flip to label each row.
Random forest
160 trees, max depth 8, sklearn defaults. Robust, no GPU required, trains in under a second on this CPU.
Predict
For a new customer profile, return P(churn) and the global feature importances combined with the customer's feature deviation.
Verdicts
< 0.35 low-risk · 0.35–0.6 watch · > 0.6 high-churn-risk. Treatment teams act on the top tier first.
Production swap
Real systems use gradient boosting (LightGBM / XGBoost) with calibrated probabilities + uplift modelling to decide whether a save-offer is worth sending.
Want this on your real customer base?
We build production churn systems with your CRM, billing, network usage, and care-ticket data — calibrated to feed save-offer treatment plans and measured against a holdout cohort.
Ready to start
Turn one AI use case into measurable production value.
Book a 30-minute consultation. We will walk through the use case, sketch the value case, and tell you honestly whether we can help.