Take every customer who signed up in March. Line them up in a row. Now check what percentage of them are still paying in April, in May, in June, and keep going for as long as you have data.
That’s the mechanic of a cohort retention table.
Many founders have heard the term in a pitch or a board deck and nodded along without ever building one, because the concept sounds more complicated than the spreadsheet actually is.
Here’s why blended retention numbers hide the truth. If you report “85% monthly retention” across your whole customer base, that number mixes together people who signed up two years ago and have been stable for a long time with people who signed up last month and might churn in week three. A company can lose new customers fast while its retention number still looks fine, simply because the old, loyal base is large enough to drag the average up. A cohort table breaks that illusion apart by month, so you can see exactly where the leak is instead of guessing.
Building one starts with a single column: the month a customer first paid. Every customer belongs to the cohort of the month they joined, permanently. Down the rows, you track calendar months. Across the columns, you track how many customers from each starting cohort are still active in each of those later months, usually shown as a percentage of the original group. The January cohort might show 100% in month one, 78% in month two, 65% in month three. The February cohort has its own row, starting fresh at 100% in its own month one. Lay enough of these rows on top of each other and a pattern either shows up or it doesn’t.
What you’re looking for is whether retention curves flatten out or keep sliding. A healthy product usually loses some customers in the first month or two and then the curve levels off, meaning the customers who stick around tend to stay for a long time. A curve that keeps dropping steadily, month after month, without ever flattening, is a warning sign no matter how good your top-of-funnel growth looks. It means you’re constantly refilling a bucket with a hole in the bottom, and every dollar spent on new customer acquisition is fighting churn instead of compounding on top of it.
The other thing a cohort table does well is show whether changes you’ve made are working. If you shipped a new onboarding flow in April, compare the April cohort’s month-two retention against March’s. If you raised prices in May, watch whether the May cohort churns faster in its first ninety days than the cohorts before it. Without cohorts, these effects blur into the same monthly average as everything else. With them, you can isolate a single decision and watch its consequences play out over the following months.
None of this requires complicated software. A basic version can be built in a spreadsheet with customer signup dates and monthly activity status, which is enough to start spotting patterns most founders have never actually seen laid out. The harder part is doing it consistently, updating it every month, and actually looking at it before decisions get made instead of after something has already gone wrong.