Two ways to read what last year's customers are worth this year: gross revenue retention ignores expansion, net revenue retention counts it.
In practice
These two rates are almost never written one without the other, and that's deliberate: NRR is the figure that goes in a deck, GRR is the one that says what the product actually keeps. GRR counts only losses — cancellations and downgrades — so it can never exceed 100%; NRR adds expansion on top and can. The gap between them is exactly what expansion buys. An NRR of 105% next to a GRR of 90% describes a leaking base where a handful of large accounts pays for the ones walking out: sustainable for a few quarters, never for a few years, and usually discovered the moment the biggest account stops growing. Reading both together is the only way to see that coming.
The formula
NRR = (starting MRR + expansion − contraction − churn) ÷ starting MRR, and GRR = the same line without the expansion term
- Starting MRR
What the customers you already had at the start of the period were paying. New customers won during the period are excluded from both formulas — including them is the single most common way an NRR gets quoted at 140%.
- Expansion
Upgrades, extra seats, usage above a plan, modules bought — money from customers who were already there. It appears in NRR and never in GRR, which is the entire difference between the two.
- Contraction and churn
Contraction is a customer who stays and pays less; churn is a customer who leaves. Both are losses, they have different cures, and lumping them together hides which one you have.
GRR can never exceed 100% — it only subtracts. NRR can, and that is why it is the number quoted in fundraising decks. Reading them as a pair is the point: GRR is the state of the bucket, NRR is the bucket plus the tap.
Worked example
One month, read twice
- Start of month: 400 customers, €20,000 MRR. During it: 12 cancel (−€600), 8 downgrade (−€400), 15 upgrade (+€900).
- GRR = (20,000 − 400 − 600) ÷ 20,000 = 95%. Five points of the base leaked, and nothing in this figure can ever offset that.
- NRR = (20,000 + 900 − 400 − 600) ÷ 20,000 = 99.5%. Expansion closed four and a half of those five points.
- An expansion playbook — the upgrade offered the moment an account hits its seat limit, instead of waiting for it to find the pricing page — takes that month's expansion from €900 to €1,080. NRR becomes (20,000 + 1,080 − 400 − 600) ÷ 20,000 = 100.4% — above the line at last. GRR is still 95%: not one customer was saved.
That last line is why the two are quoted together. An NRR just over 100% looks like a business whose base grows on its own; the GRR next to it says the base is still leaking 5% a month and a handful of growing accounts is paying for the ones walking out. Both statements are true, and only one of them is in the deck.
Orders of magnitude
- The figures most often quoted for B2B SaaS: NRR of 110-130% is considered strong, and above 120% is the enterprise territory where seat growth inside big accounts does the work. Under 100% is normal and perfectly healthy in self-serve and SMB, where there is structurally less room to expand.
- GRR is commonly quoted at 85-95% annually in B2B, and the number matters more than NRR for judging the product itself: nothing a sales team does can raise GRR, only the product and the service can.
- A pricing model with nowhere to grow — one flat plan, everything unlimited — caps NRR at GRR by construction. That is a pricing decision showing up as a retention number, and it is usually taken by accident.
How to improve it
- Read GRR first, always. Expansion can outrun a leak for a few quarters and never for a few years, and a team that only watches NRR finds out at the moment the biggest account stops growing.
- Segment both by cohort size. One account doubling can carry a whole month's NRR, which means the headline number describes that one account rather than your customer base.
- Split contraction from churn in the reporting. They are the same subtraction and opposite problems: contraction usually means the pricing tiers are wrong, churn usually means the product or the onboarding is.
- Measure both on the same cohort definition and the same period length, month after month. An NRR computed annually and a GRR computed monthly are not a pair, they are two unrelated numbers printed next to each other.
In the Tour
Do you have an expansion playbook (upsell/cross-sell)?
- Yes, active and used20 pts
- Some ideas, not systematic7 pts
- Nothing0 pts
The Tour asks whether you have an expansion playbook — 20 points for a systematic one, 7 for "some ideas, nothing systematic", 0 for none. NRR is the number that answers it from the other end: a team with no playbook has an NRR pinned to its GRR, and usually does not know either figure. The 7-point answer is the interesting one, because ideas never show up on the expansion line — which is exactly what the gap between these two rates measures.
Questions people ask
Is NRR the same as net dollar retention?
Yes. Net revenue retention, net dollar retention and NDR are the same calculation under three names, and you will meet all three in the same funding round. Net revenue retention is the version that survives a change of currency.
Can NRR be above 100% while the business shrinks?
Yes, and it happens more often than the metric's reputation suggests. Lose half your customers, have the survivors double their spend, and NRR reads perfectly healthy on a base that has halved in headcount. That is why logo churn belongs next to it, and why GRR is the honest half of the pair.
Should new customers be included?
No, and this is the mistake to watch for when reading someone else's number. Both rates measure what an existing base does over time; adding new customers turns them into a growth rate wearing a retention label, which can read well above 100% for a business losing every cohort it wins.
Monthly or annual?
Annual is the convention for comparison, monthly is what you steer with. Careful with the conversion: a monthly NRR of 99.5% is not an annual 99.5% — compounded over twelve months it is about 94%. Rates compound, they do not add.