Cohort analysis
Splitting customers into groups that started at the same time or in the same way, for example everyone who signed up in January, and tracking each group's behaviour over time. It separates real changes in retention or spending from the noise of a growing or shrinking customer base. Averages across all customers hide this: a product can look stable overall while every new cohort churns faster than the last.
Where it's used
Used in subscription businesses, apps, e-commerce and SaaS to read retention, payback and lifetime value. Good for checking whether a product, onboarding or pricing change actually helped the customers who came after it.
How to calculate it
Cohort retention = customers still active in month N ÷ customers who started in the cohort
Worked example
Illustrative numbersA subscription app changed its onboarding in February. 400 people signed up in January and 148 were still active three months later. 500 signed up in March, after the change, and 230 were still active three months later.
January cohort, month 3
148 active ÷ 400 signed up = 37%
March cohort, month 3
230 active ÷ 500 signed up = 46%
Both cohorts blended
(148 + 230) ÷ (400 + 500) = 42%
What it tells you
The blended 42% looks like a small change. The cohorts show the new onboarding lifted three-month retention by 9 points.