Survival Analysis of SaaS Churn
[R][Survival Analysis][Cox PH][Kaplan-Meier]
§1. Abstract
Instead of just guessing *if* a customer will cancel their subscription, I used medical statistics (normally used to predict patient survival rates) to predict *when* they will cancel. I found that getting customers on annual contracts cuts the risk of them leaving by almost half.
§2. Methodology & Implementation
Analyzed 2,847 customer records to understand not just if, but when customers churn. Progressed from Kaplan-Meier survival curves to a semi-parametric Cox proportional hazards model. Identified key drivers of retention (e.g., annual contracts cut churn risk by 47%) and verified assumptions using Schoenfeld residual tests.
§3. Key Metrics
| Model | Cox PH |
|---|---|
| Key Finding | Annual cuts churn 47% |
| N | 2,847 |