Net Revenue Retention: Formula and Benchmarks

Take the customers who were paying at the start of the period, and compare what they pay now:

NRR = (revenue from the starting cohort today)
    ÷ (revenue from the same cohort at period start)

GRR = (starting cohort revenue, expansions excluded)
    ÷ (starting cohort revenue at period start)

Both are normally quoted over twelve months and expressed as a percentage. The critical word is cohort: new customers signed during the period are excluded from both the numerator and the denominator. If they sneak in, you are measuring growth, not retention, and the number becomes meaningless.

Because expansions are included, NRR can exceed 100%. Because GRR excludes them, it cannot.

They answer different questions, and a business can look healthy on one and sick on the other.

Report both. NRR is the growth story; GRR is the leak rate. A company at 115% NRR and 82% GRR is growing its base while quietly losing nearly a fifth of it every year — that is a business whose growth depends on continuing to upsell, and that dependency is invisible if you only quote NRR.

Published benchmarks cluster by segment rather than around a single number:

• Enterprise B2B SaaS — commonly cited best-in-class is above 120%, with 110% treated as healthy. Large accounts have more seats and more modules to expand into.
• Mid-market — roughly 100% to 110% is typical, with anything above 105% considered strong.
• SMB and self-serve — 85% to 95% is normal. Small accounts leave or downgrade faster, and no amount of expansion fully offsets it.

These are orientation bands, not targets. A business at 92% NRR in a segment where the median is 88% is doing better than one at 105% in a segment where 115% is common. Compare within your own motion.

One caveat worth stating plainly: benchmark figures circulated online are not a single audited dataset. They come from different survey populations, different cohort definitions and different periods. Use them to check an order of magnitude, not to settle a board argument.

1. New customers counted as expansion. If a customer signed in March and bought more in June, that is a new logo expanding — not retention of an existing cohort. Exclude them from the starting cohort.

2. Price rises and plan migrations. A 10% list price increase applied to the whole base shows up as NRR expansion without a single customer getting more value. It is real revenue, but it is not retention, and it does not recur next year. Many teams report it separately as price-driven expansion, and you should too.

3. Non-recurring revenue in the base. Professional services, onboarding fees, overage on usage and one-off add-ons all inflate the starting base and then disappear. NRR should be computed on recurring revenue only — the same definition as MRR.

Each of these can move the number by five to fifteen points. Together they can turn a 95% business into a 110% slide.

You need three columns per customer per month: customer id, plan, and MRR. Then, for a twelve-month window starting at month t0:

1. Take the set of customers with MRR > 0 in month t0.
2. Sum their MRR in t0 — that is the denominator.
3. Sum the MRR of that same set in t0 + 12 — customers who left contribute zero. That is the NRR numerator.
4. For GRR, cap each customer’s month-13 MRR at their month-1 MRR before summing, so expansions contribute nothing and contractions still count.

Do this for twelve monthly cohorts and average them, rather than computing one window once. A single window picks up whatever renewal cycle happened to fall inside it.

If you cannot reproduce your reported NRR from raw billing rows, you do not have an NRR — you have a number someone assembled by hand.