How to Read a Cohort Retention Table (and What It Tells You That Churn Doesn't)

A monthly churn rate tells you how many customers left. It doesn't tell you which customers, or when in their lifetime they left, and those are usually the questions that lead to a fix. A cohort retention table answers both by grouping customers by the month they signed up and following each group over time.

How the table is built

Each row is a cohort: everyone who started in a given month. Each column is months since sign-up. Each cell is the share of that cohort still active at that point. Reading across a row shows one cohort's life; reading down a column compares cohorts at the same age.

Three things to look for

1. Where the curve flattens. Most products lose customers quickly at first, then settle. The month where the drop slows tells you how long it takes customers to find lasting value, and the level it settles at is your long-term retention.
2. Whether newer cohorts sit higher. If February's month-1 retention (88%) beats January's (85%), something you changed is working, perhaps onboarding or targeting. A single churn figure would hide that improvement inside the average.
3. Sudden drops in a column. If every cohort loses more customers at month 3, look at what happens then: a trial ending, an annual discount expiring, a feature limit being hit.

Weighted, not simple, averages

When you summarise a column, weight by cohort size. A simple average of 85% and 88% is 86.5%, but because February was the bigger cohort, the true month-1 retention across both is 86.7%. With cohorts of very different sizes, the gap between the two methods can be large.

Customers or revenue

The same table can track revenue instead of customer counts. Revenue retention can rise above 100% for a cohort if the customers who stay upgrade by more than the ones who leave paid. Seeing that in a cohort table is one of the clearest signs of a healthy product.

Turning the table into action

• If most losses happen in month 1, work on onboarding: the first session, the first result, the first week of emails.
• If losses spike when a trial or discount ends, test the offer and the reminder emails before the end date.
• If older cohorts decline slowly but steadily, look at what long-term customers use and what they ask support about.
• If one acquisition channel's cohorts retain much worse, move budget away from it even if its sign-ups are cheap.

Re-build the table every month. Its value comes from watching new rows appear and comparing them with the rows above.

Tip: Ignore the newest cohorts' latest cells when judging trends. They're based on only a few weeks of data and will move.