Key takeaways
- A cohort is a group of customers who started in the same period.
- Groups are compared at the same age, such as six months after they started.
- Cohorts can count customers or the revenue they pay, and the two tell different stories.
- Cutting cohorts by channel, segment or plan shows which customers last.
How a retention cohort table works
A retention cohort table puts one cohort on each row and the months since the cohort started across the columns. Each cell shows how many of that cohort's customers, or how much of their revenue, remains at that age.
Reading down a column compares cohorts at the same age, which is the point of the exercise. Reading along a row shows how one group behaves over its life, including the month where most losses happen.
A cohort analysis example
Here is an illustrative cohort analysis example. A business won 50 customers in January and 60 in April. Six months after starting, 40 of the January customers remain, and 54 of the April customers do.
The April cohort kept nine in ten after six months, against eight in ten for January. Something changed between the two, perhaps a new onboarding programme, a tighter ideal customer profile or a pricing change, and the table tells the team where to look.
Cohorts by customer count or by revenue
Cohorts by customer count show how many accounts stay, while cohorts by revenue show how much money stays. A revenue cohort can end above its starting value when the customers who stay buy more seats or upgrade, which is the shape behind strong net revenue retention.
Looking at both is worthwhile. A revenue cohort that holds up while its customer count falls means a few large accounts are growing while many small ones leave.
SaaS cohort analysis for go-to-market teams
SaaS cohort analysis helps go-to-market teams see which customers turn into lasting revenue, not just which ones sign. Cut the cohorts by acquisition channel, industry, company size, plan or the team that sold them, and compare their retention at the same age.
Marketing learns which channels bring customers who stay. Sales learns which segments of its ideal customer profile fit the product best. Customer success sees whether a change to onboarding has worked months before the yearly renewal figures arrive.
When a cohort analysis cannot be trusted
A cohort analysis cannot be trusted when its definitions shift between cohorts or its groups are too small to mean anything.
- Cohorts so small that one customer leaving swings the result.
- Comparing cohorts at different ages, such as a new one against an old one's full life.
- Changing what counts as the start date partway through.
- Mixing trial users with paying customers in the same cohort.
Where Pebbles Ai fits
Common questions
How big should a cohort be?
Big enough that one customer leaving does not swing the result. If monthly cohorts are too small, group customers by quarter instead.
Is cohort analysis only for retention?
No. The same method tracks spending, expansion, feature adoption or time to first purchase for any group that shares a starting point.
What is the difference between a cohort and a segment?
A cohort shares a starting period, while a segment shares a trait such as industry or size. Combining them, for example the January cohort of mid-sized customers, is often the most useful view.
About this page. Written by the Pebbles Ai team for people who run sales, marketing and growth. Last updated .
