Churn Prediction & Retention Engine

Challenge

A fintech platform was losing mid-market accounts with little warning. Retention plays were reactive and inconsistent across account managers.

Approach

We engineered churn features from product usage, billing, and support signals, trained a gradient boosting model, and connected high-risk scores to Salesforce playbooks and automated outreach.

Outcomes

• At-risk accounts surfaced weeks before typical churn, not on cancel day
• Retention playbooks standardized across the account-management team
• Model outputs logged for review, not a black-box score
• Outreach triggered from CRM, so ops does not live in a spreadsheet

Stack