How to Calculate Net Revenue Retention

Net revenue retention measures whether your existing customer base grows on its own. Take a cohort of customers as at the start of a period, add their expansion, subtract their contraction and churn, and divide by where they started. Above 100% the base grows without new customers; below, acquisition is filling a leaking bucket.

The calculation is simple. The errors are consistent, and every one of them makes the number look better — which is why a figure quoted without its method is not worth much.

The calculation, precisely

Fix a cohort at the start of the period. Record that cohort's recurring revenue: this is the denominator, and it does not change. At the end of the period, take the recurring revenue of those same customers — including any expansion, net of any contraction, and counting churned customers as zero. Divide.

Including the $40,000 of new business would produce 142%, which describes growth rather than retention and is the most common way the number gets inflated.

The four errors

1. Including new customers

The most common and the most flattering. It converts NRR into a growth metric that reads above 100% for almost any growing company regardless of retention quality, which makes it useless for the decision it exists to inform.

2. Measuring a shifting cohort

Defining the cohort as "customers we have now" rather than "customers we had at the start" silently excludes everyone who churned during the period — precisely the population the metric is measuring. This one is easy to introduce accidentally through a dashboard filter.

3. Mixing recognised revenue with recurring revenue

Professional services, implementation fees and uncommitted usage overages are revenue but not recurring revenue. Including them makes the number move for reasons unrelated to retention, and the movement is usually mistaken for a real signal.

4. Annualising by multiplication

Retention compounds. Multiplying a monthly rate by twelve overstates the annual figure, and the error grows the further the rate departs from 100%.

Method: producing a defensible NRR in an afternoon

1. Choose the period and freeze the cohort. Export every customer with an active subscription as at the first day of the period, with their recurring revenue on that date. Save this file — it is the denominator and it must not be regenerated later.

2. Classify each customer's end state. For the same customer IDs, pull recurring revenue on the last day of the period. Every customer falls into exactly one of: retained flat, expanded, contracted, or churned.

3. Strip non-recurring revenue from both ends. Remove services, one-off fees and usage above committed minimums. If you cannot separate them, that is the first finding and it affects more metrics than this one.

4. Reconcile before dividing. Opening revenue, plus expansion, minus contraction, minus churn, must equal closing revenue exactly. If it does not, a customer is miscategorised — usually one that both expanded on one product and contracted on another.

5. Compute NRR, then GRR on the same cohort by setting expansion to zero, then logo retention by counting customers rather than revenue.

6. Repeat by segment. Run the same calculation separately for self-serve and enterprise, or for whatever segmentation drives different motions.

Reading the result

NRR alone can be flattered by expansion concentrated in a handful of large accounts while many small customers leave. That is why it should never be reported alone.

Logo retention adds the third dimension: strong revenue retention with weak logo retention means small customers are leaving while large ones grow. Whether that matters depends on strategy — if the small segment is where you intend to scale, it is a leading indicator of a motion or product mismatch.

Where this method fails

Three conditions make the standard calculation misleading, and each needs an explicit adjustment rather than a caveat.

• Usage-based pricing: only committed minimums are genuinely recurring. Treating variable consumption as recurring makes NRR swing with customer activity rather than with retention.

• Very small cohorts: with few customers, one departure moves the rate by several points. Report the absolute counts alongside the percentage or the trend will be over-read.

• Multi-product businesses where a customer expands on one product and churns another: the reconciliation in step 4 catches these, but they must be netted at the customer level rather than counted as both an expansion and a churn.

What to do with the number

NRR below 100% is a signal to prioritise the post-sale motion over acquisition, because acquiring into a leaking base compounds the problem rather than solving it. The first diagnostic is splitting voluntary from involuntary churn: failed payments and expired cards are a payments problem with an operational fix, and they are frequently a larger share than expected. Counting them alongside voluntary churn hides a recoverable loss inside a product problem, and the wrong team gets asked to solve it.