Two companies have the same number of customers and the same revenue — but one is growing healthily and the other is hollowing out underneath. A total does not tell them apart; cohort analysis does. It is the technique that reveals whether you are really retaining customers or just replacing the ones you lose.
What a cohort is
A cohort is a group of customers who share a starting point — usually the month they began. "Customers who joined in January" are a cohort; "February's", another. The idea is to follow each group over time, instead of mixing everyone into a single total.
Why the total misleads
A monthly total can stay stable while, underneath, you lose old customers and gain new ones at the same rate. It looks healthy, but it is a bathtub draining with the tap open. Only by splitting into cohorts do you see the truth: how long each group actually stays.
How to read a cohort table
In a cohort table, each row is a group (entry month) and each column is the time since entry (month 1, 2, 3...). The cells show how many are still active. Reading horizontally, you see each group's retention curve; vertically, you compare groups against each other.
What cohorts reveal
- Real retention: how many customers stay after 1, 3, 6 months.
- Whether you are improving: do more recent cohorts retain better than old ones?
- Where you lose customers: at which point in the relationship the curve drops sharply.
From diagnosis to action
If cohorts show most abandon by the second month, you know exactly where to act: the initial experience. If new cohorts retain better, something you changed is working. Cohort analysis turns "we think we retain well" into evidence about where and when to act.
In practice
Group your customers by the month they started and follow each group's retention over time. You will see what a total hides. It is one of the most revealing — and most ignored — analyses in any relationship-based business. Is your retention improving cohort by cohort, or do you just look stable on the outside?
