Financial Project

SaaS Churn Prediction Model

Machine learning model to identify at-risk customers before they cancel.

The Concept (In Plain English)

For a subscription business, losing a customer (churn) is the silent killer. By the time they cancel, it's too late. I built a 'warning system' that looks at user behavior—like logging in less often or submitting support tickets—to predict *who* is likely to quit next month. This gave the success team a 'hit list' of at-risk accounts to save, directly protecting the company's revenue.

The Build (Technical Deep Dive)

Developed a logistic regression model in Python (scikit-learn) to predict customer churn probability. Feature engineering included usage frequency, support ticket sentiment analysis (NLP), and contract length. The model achieved an AUC-ROC of 0.85. The results were piped directly into the CRM (Salesforce) via API, flagging high-risk accounts for immediate CSM intervention.

Key Metrics

AUC-ROC0.85
Revenue Saved$200k/yr
ModelLogistic Regression

Tech Stack

Machine LearningPythonPredictive ModelingSaaS