Cohort and cohort analysis: what it is and how to read a retention table

A cohort is a group of customers who arrived in the same period: the same week or the same month. Each group is tracked separately: how many of it are left after one month, two, three.

Why it matters

The total number of customers can grow even when every group melts away quickly: new ones simply arrive faster than old ones leave. While the inflow holds, the problem is invisible. Cohorts show it right away.

Example. Of 100 customers who arrived in January, 70 were left a month later, then 55, 50, 48, 47. The losses slow down and stop — this is called a plateau: those who stayed found the product useful. Of 100 who arrived in March, 60 were left, then 40, 30, 22. No plateau, the group keeps melting. So something changed in March: the wrong customers came, the price went up or something broke in the product.

How to build the table

Rows are the month customers arrived. Columns are how many months the customer has been with you: first, second, third. Each cell shows what share of the group still pays. To see whether things are getting better, read a column top to bottom: do newer groups hold up better in month three than older ones?

Common mistakes

• Comparing calendar months rather than months of the customer's life. The January group in April and the April group in April are customers of different "ages".
• Making groups too small. In a group of 10, one customer leaving is 10% at once.
• Mixing paying customers with those only trying. They behave differently, and the mix shows neither.

In Runway 200

Every week of new customers is its own cohort. To leave the first phase, seven in ten have to make it to month two. Without the "Cohort reports" initiative you see this share with noise, and the cause of churn not at all. You can take the initiative already in the first phase. The share is raised by segment fit, product quality, a price below what the segment is willing to pay, and "stickiness": reports, help with the first steps, calls to customers who left. Stickiness acts on all living cohorts at once, while fit and discounts are something each cohort brings with it.

How to calculate churn and why growth hits a ceiling — in Customer churn.

More in the glossary: Burn rate · CAC · LTV · MRR and ARR · Product-market fit · all terms