Most NRR numbers are wrong in the same way — they let new customers into a calculation that is supposed to exclude them. Here is the arithmetic, with a worked example.
Net revenue retention is the metric investors ask about first and the one most often calculated incorrectly. The error is almost always the same, and it flatters the number.
One question: are the customers we already had getting bigger or smaller on their own?
That is why new customers must be excluded. A strong sales quarter can hide a leaking bucket for a long time. Taking new business off the table is the entire purpose of the metric.
Freeze the customer list on day one. Every account paying you anything at the start of the period joins the cohort. Nobody joins later, however large they grow.
Add up what that cohort was paying. That is your denominator.
Track only what happened to those accounts during the period — expansion, renewal uplift, contraction, churn.
Divide the end by the beginning.
Five customers were paying you on 1 February, totalling €305,000. During the year:
A sixth customer signed in June for 60,000. They are not in the calculation.
Below one hundred percent, which tells you the existing base shrank. Note that total ARR still went up that year, because of the new customer. Both facts are true, and reporting only the first one hides the problem.
NRR is simply what the cohort pays at the end divided by what it paid at the start. Compute it both ways — by summing the movements, and by comparing the two balances. If they disagree, an event has been categorised wrongly or a new customer has leaked in.
Gross retention is the same calculation with the additions removed:
It can never exceed one hundred percent. If yours does, a positive movement has found its way into the formula.
The gap between the two — here, about fifteen points — is your expansion engine. NRR alone cannot tell you whether you are growing because customers love you or shrinking slowly while two big upsells paper over it.
In the example, four of five customers stayed: eighty percent logo retention against sixty-nine percent gross revenue retention.
That gap is the alarm. You kept most of your customers but lost a disproportionate share of the money, which means the account that left was a large one. If the numbers were reversed — many small losses, little revenue — that is a far less urgent problem. The two are indistinguishable if you only track euros.
New business leaking in. The most common error. Structurally impossible if you build the cohort correctly, because a new logo had a zero balance before the period.
Currency movement. If a customer pays in sterling and the rate moves, your reported ARR changes while nothing commercial happened. Calculate in constant currency, or exchange-rate noise appears as expansion.
Account versus customer. If one company has three subsidiaries as separate accounts, roll them up first. Otherwise a customer consolidating contracts looks like two churns and an expansion.
One-off revenue. Onboarding fees and services are not recurring. If they sit in the base, NRR swings for reasons that have nothing to do with retention.
The cohort is the problem. Reporting tools can sum values by category; they cannot easily freeze a list of accounts at a point in time and then measure only those accounts across a later window. That is a self-join across two periods.
It becomes straightforward when every ARR change is recorded as a dated, permanent event — because then the cohort is one query and the movements are another. That is the approach Revligent takes, and it is why the number can always show its working.
Thirty minutes in your Salesforce sandbox. We install it, configure your pricing, and quote a real opportunity end to end.