How to Read a Retention Table
A cohort retention table has cohorts (groups of customers acquired in the same month) as rows, and time periods (Month 0, Month 1, Month 2...) as columns. Each cell shows the percentage of that cohort still active or retained at that time period.
Good retention looks like a gradual decline that flattens out — the curve stabilises because loyal, high-value customers remain while low-fit customers exit early. A retention curve that never flattens suggests you lack a retained core audience.
Bad retention is a steep, continuous decline. If Month 6 retention is less than 20% of Month 1 for a B2B SaaS, you have a fundamental product-market fit issue, not a customer success problem.
Identifying Churn Patterns by Cohort
Cohort analysis reveals whether churn is getting better or worse over time. If your 2024 cohorts retain significantly better than 2023 cohorts at the same time period, product improvements are working. If newer cohorts churn faster, you may be acquiring lower-fit customers through new channels — a common failure mode when startups scale paid acquisition aggressively.
Key Pattern: The "Smile" vs "Cliff"
- Smile curve: High D1, modest D7 dip, then stabilises at D30+ — healthy consumer app
- Cliff curve: Strong D1, then freefall — usually means the first experience is impressive but the core loop is broken
- Plateau curve: Moderate D1, holds steady D7–D30 — excellent signal, common in utility / workflow SaaS