How to Calculate NRR by Cohort (and Why Blended NRR Misleads)

Blended net revenue retention gives you one number for the whole company. It is the number the board asks for, and it is close to useless for figuring out what changed. Cohort NRR splits the base into groups that signed at the same time and tracks each group forward, which separates real retention movement from mix shift. This guide covers the calculation, the cohort definition choices, and how to read the output.

What is the cohort NRR formula?

Cohort NRR is the current ARR of a fixed customer group divided by that same group's ARR at cohort start.

``` Cohort NRR (month n) = ARR from cohort customers in month n / ARR from cohort customers in month 0 x 100 ```

Two rules make it work. The cohort membership is frozen at month 0 and never changes, and no customer acquired after month 0 ever enters the cohort. Churned accounts stay in the cohort at zero ARR rather than dropping out, because removing them would turn a retention metric into a survivorship metric.

Expansion is included, which is what makes it net rather than gross. A cohort above 100 percent is expanding faster than it is leaking.

How do you build the cohort table?

Anchor every customer to a signing period, then measure the group's ARR at fixed intervals.

Start with signing quarter as the cohort key. It is the cut that isolates changes in product, pricing, ICP definition, and qualification discipline over time.

The 2025 Q3 cohort is the story. It is the largest cohort by account count and the first one to fall below 100 percent at month 12. Blended company NRR in that period might still read above 105 percent because the 2024 cohorts are carrying it, which means the deterioration is invisible until the older cohorts age out.

Why does blended NRR move when nothing has changed?

Because the mix of cohort ages inside the blend keeps shifting.

Expansion takes time. A customer six months in has rarely added seats or products yet. A customer thirty months in probably has. When a company grows quickly, the share of young accounts in the base rises, and blended NRR falls even if every individual cohort is performing exactly as it did before.

The reverse also happens. A company whose new business slows will show blended NRR rising, because the base ages into its expansion window and no new low-NRR accounts dilute it. Reading that as improved retention is a mistake that gets repeated in board decks every quarter.

Cohort NRR removes the mix effect by comparing month 12 to month 12 across cohorts. That is the comparison that answers whether retention is actually getting better or worse.

Which secondary cohort cuts are worth building?

Segment, entry product, and acquisition channel, in that order. Entry product is the most underused of the four. In multi-product companies, the product a customer lands on frequently determines the expansion ceiling, and a landing product with strong logo retention but no expansion path will drag cohort NRR down for years without ever showing up as churn.

Keep cohorts above roughly twenty accounts. Below that, one large customer moving in either direction swamps the cohort and the curve stops carrying information.

How does cohort NRR change the forecast?

It gives you an expansion forecast grounded in a curve rather than a target.

Once you have three or more cohorts tracked to month 24, you have an empirical expansion curve. New cohorts can be projected against the average curve of prior cohorts, adjusted for whatever you know changed. That produces an expansion number with a defensible derivation, which is materially better than the percentage growth assumption most plans use.

Two cautions. First, a curve built on old cohorts embeds old market conditions. Forecasts fail most often because the business or the market changed and the model is still running on assumptions from before the change, so a pricing shift, a new competitor, or a change in buying behavior invalidates the historical curve. Second, mid-term contraction lands between renewal dates and does not wait for a renewal conversation, so the curve needs monthly resolution rather than annual snapshots to catch it.

For the underlying metric definition, see net revenue retention. For connecting the retention curve to the full revenue plan, see how to forecast revenue and sales forecasting best practices.