Glossary · Metrics · Updated
A cohort groups customers who arrived in the same period and follows them over time. Cohort analysis shows whether the product is improving, by comparing how January's customers behave against June's at the same age.
Blended averages hide what matters. An average retention of 60% can describe a situation improving sharply or deteriorating, depending on whether recent customers stay longer than older ones. Only a cohort view separates the two.
It is also the only way to measure the effect of a change. If you reworked onboarding in April, May's cohort should behave differently from March's at the same age. Without cohorts the effect dissolves into the average and becomes invisible.
For an investor, a cohort table is the most persuasive document in a dataroom, well ahead of the financial plan. It shows a real trajectory rather than a projection, and it is hard to dress up because each row is a fixed population followed over time.
Example: a retention table by cohort
B2B SaaS customer retention measured at three ages, across three monthly cohorts.
| January cohort, month 1 | 100% |
|---|---|
| January cohort, month 3 | 62% |
| January cohort, month 6 | 48% |
| March cohort, month 3 | 68% |
| March cohort, month 6 | 57% |
| June cohort, month 3 | 74% |
| June cohort, month 6 | 65% |
| Gap between January and June at month 6 | 17 points |
Seventeen points of difference at the same age: the product retains far better than six months ago. This is the kind of curve that convinces an investor, because it proves improvement instead of promising it.
The common mistake
Comparing cohorts at different ages. June's cohort will always look better than January's if you read the first at three months and the second at twelve. The comparison only means something at the same number of elapsed months.
Frequently asked questions
What is a cohort?+
A group of customers or users who arrived in the same period, usually the same month, then followed over time as a fixed population. At regular intervals you measure how many are still active or how much revenue they still generate.
What does cohort analysis reveal?+
Whether the product is improving. Comparing different cohorts at the same age isolates the effect of product or commercial changes, whereas a blended average mixes old and new customers and hides the trend entirely.
Related terms
Prepare your raise with the right documents
Onefive brings the dataroom, the investor network and the startup profile into one workspace, built for European teams from pre-seed to Series A.
