What is Cohort Retention?
Cohort retention groups customers by the month they started paying and follows each group forward, so you can see how long customers stay and how their spend develops, without newer customers masking older ones.
A cohort table is read two ways. Across a row is one signup month aging: month 0 is the month they joined, month 6 is half a year in. Down a column is the same age for successively newer cohorts, which is how you tell whether the product is getting better at keeping people.
Cells hold either customers or revenue. The customer flavor is the share of the cohort still active, and it can only fall. The revenue flavor is the cohort's current MRR against what it committed at signup, and it can rise above 100% when expansion outpaces losses, which makes it net revenue retention traced by cohort.
Cohorts answer the question a blended churn rate cannot: is the leak in the first ninety days or spread evenly across the life of the account? Those two situations look identical in an aggregate churn number and call for completely different fixes.
How to calculate Cohort retention
Customer retention (cohort C, month k) = customers from C still active at month k ÷ size of C x 100
- cohort C
- every customer whose paid life started in the same month
- month k
- months elapsed since the cohort started, k = 0 being the signup month
The revenue flavor
MRR retention (cohort C, month k) = cohort C MRR at month k ÷ cohort C MRR at signup x 100
Cohort retention example
Given
- March cohort: 40 customers, $5,000 of MRR committed at signup
- Six months later: 31 of them still active
- Those 31 now carry $5,400 of MRR between them
Nearly a quarter of the logos are gone, yet the cohort is worth more than it was at signup. The survivors expanded enough to cover the losses, which is the signature of a product that lands small and grows inside accounts.
Why Cohort retention matters
- It shows when customers leave, not just how many. Retention that falls off a cliff in month 2 is an onboarding problem; a steady slope is a value problem.
- It makes product and pricing changes measurable. If the cohorts that signed up after a change hold better at the same age, the change worked.
- Cohort MRR retention is the most credible evidence that expansion is durable, because it follows real groups of customers rather than blending everyone together.
Common Cohort retention mistakes
Reading the newest cohorts as a trend
The bottom rows have only lived a month or two. They will always look better than mature cohorts, because they have not had time to leave yet.
Comparing cells at different ages
Only compare down a column, at equal months since signup. Comparing a mature cohort at month 18 against a young one at month 3 says nothing.
Ignoring cohort size
A cohort of six customers moves 17 percentage points every time one of them leaves. Small cohorts are noise until they are not.
How Kometrics computes Cohort retention
Kometrics builds cohorts from the movement ledger, grouping customers by the month of their first new business movement. Both flavors come from one pass: the customer table counts cohort members with MRR above zero at each month end, and the MRR table divides the cohort's MRR at each month end by the MRR it carried at the signup instant.
Because both come from the same ledger as every other report, a cohort cell and the MRR chart cannot disagree. Column zero is the signup month itself, so a cohort's first cell is where it started, not where it ended its first month.
Related terms
Churn rate is the share of your customers, or of your revenue, that you lost during a period. Which of the two you mean changes the number substantially, so the qualifier matters as much as the figure.
NRRNet revenue retention is what happened to the revenue you already had. Take a cohort of existing customers, add the expansion they bought, subtract what they downgraded or cancelled, and compare against what they were worth at the start. New customers are excluded by definition.
Expansion MRRExpansion MRR is the additional recurring revenue you earn from customers you already have. Upgrades to a higher plan, extra seats, add-on modules and price increases all count; a new customer never does.
LTVCustomer lifetime value estimates how much revenue a single customer produces across their whole relationship with you. In subscription businesses it is derived from average revenue per account and churn rather than observed directly, because most customers have not left yet.
Get this number from your billing data
Kometrics connects to Stripe, Paddle, Creem and Asaas and computes Cohort retention continuously, with every movement behind it. Free under $1,000 MRR.
Start free