Customer Analytics

Predictive Churn Modelling for a B2B SaaS

85%
Model Accuracy
-25%
Churn Reduction

The Problem

The company was experiencing a 6% monthly churn rate but had no way to identify at-risk accounts until they had already requested cancellation.

The Approach

Built an XGBoost classification model using product usage telemetry (logins, core action completion rates) and support ticket sentiment.

The Result

The model predicted churn with 85% accuracy 30 days out. Armed with this list, the Customer Success team reduced actual churn by 25% in the first quarter of deployment.

Let the data tell you
something useful.

Most businesses are sitting on enough data to make significantly better decisions. The question is whether anyone has looked at it properly. Let's find out.