Churn prediction model for Relaybase
A churn model that flags at-risk accounts early enough for customer success to act. A case study by Arjun Patel, AI Engineer — Chatbots & Document AI.
Challenge
Relaybase, a customer messaging platform, only noticed churn when cancellations arrived. Customer success had no way to prioritise hundreds of accounts.
Solution
I built a gradient-boosted churn model from product usage and billing data, explained each prediction with the top drivers, and pushed weekly risk scores into the CRM.