A single blended retention or churn figure averages over customers who joined at wildly different times under different products and onboarding. Cohort analysis un-blends it: group customers by when they started and watch each group over time. It is the difference between "churn is 5%" and "customers who joined after we fixed onboarding retain far better than those before" — an insight you can act on.
How cohorts work
Assign each customer to a cohort by their start period — usually the month they signed up. Then, for each cohort, measure something over the months that follow: how many are still active, or how much revenue they still generate. Laid out as a grid — cohorts down, months-since-start across — the pattern of retention becomes visible in a way an aggregate never shows.
Retention vs revenue cohorts
| Cohort type | Measures | Reveals |
|---|---|---|
| Logo retention | % of customers still active | Whether customers stay at all |
| Revenue retention | % of original revenue retained | Whether accounts expand or contract over time |
Both matter. Logo retention can look poor while revenue retention looks great if the customers who stay expand enough to more than replace those who leave — common in businesses with a self-serve tail and growing enterprise accounts. Reading them together tells you where value actually accrues.
Read the shape of the curve
The retention curve's shape is the real signal. A curve that keeps declining means you never find durable product-market fit — customers keep leaving. A curve that flattens means you have a loyal core that sticks; the height at which it flattens is your true retained base. And comparing curves across cohorts shows whether recent changes actually improved retention — later cohorts sitting above earlier ones is proof your onboarding or product work is landing.
A dip in aggregate churn could be a bad month or a structural problem — the aggregate cannot tell you which. Cohort analysis can: if every recent cohort retains worse, it is structural; if one cohort is anomalous, it is a blip. That distinction changes whether you panic or wait, which is why cohorts are worth the effort.
Cohort analysis groups customers by start period and tracks retention and revenue over their lifetime, revealing what blended metrics hide: whether retention is genuinely improving, whether you have a loyal core, and whether your fixes worked. Read the curve's shape and compare cohorts, and churn stops being a mystery number and becomes a story you can act on.
Independent software engineer in Nairobi specialising in Acumatica customisations, Laravel backends, and tax fiscalisation integrations across East and Southern Africa.