Survival Analysis of Customer Retention using Postgres, SQL and R
summary
Shows a working pipeline that computes the Kaplan-Meier estimator from event rows using Postgres SQL and R via the plr extension, handles censored observations correctly, and exposes results to dashboards. Finds a stark segmentation example — AdWords vs Facebook retention — where AdWords cohort stays while Facebook cohorts churn. Explains that an optimized SQL version outperforms the initial R approach, and warns about installing R packages on DB servers, memory use, and crash surface.
AI-written summary · about
chapters (13)
- 0:00Start
- 1:07What Is Retention Analysis
- 4:11Cohort Analysis
- 12:45Filtered Aggregates
- 14:01Compute the Relative Retention
- 29:32Periscope
- 30:09Custom Dimensions
- 37:55Stability Guarantees
- 39:35General Tips
- 40:24Working with Sparse Data Sets
- 41:02Date Range Types
- 42:05Constraints and Exclusion Constraints
- 44:57Kaplan-Meier Curves
related talks
35mUsing Survival Analysis to understand customer retention
22mSurvival analysis for conversion rates
29mStay or Go? Predicting Customer Churn through Graph Relationships
25mSql Server and R for Real-time Predictions
35mBreakout Session
27mLife, Death, and Shopping
49mThe Analytics of Growth Hacking
42mMeasuring the Efficiency of Paid Customer Acquisition Campaigns