| Cohort | Customers | Month 0 | Month 1 | Month 2 | Month 3 |
|---|
Read the heatmap correctly
Each cell divides returning customers by the original cohort size. The overall rate is weighted by customer counts, not an average of percentages. A customer can return in month 3 without returning in month 2. Channel differences are descriptive and do not establish that a channel caused retention.
Inspect the SQL
Student challenge: compare weighted retention with an unweighted cohort average. Explain when they differ and which question each answers.

Analytics, Inference