Fill a bucket with the tap wide open. If the bottom has a hole, it doesn't matter how expensive the tap is or how hard you open it: the water never reaches the rim. A subscription business is exactly that bucket. Acquisition is the tap, cancellations are the hole, and for months almost everyone stares only at the tap. Net revenue retention is the metric that finally makes you crouch down and look at the bottom.
It answers one question: if you stopped acquiring a single new customer tomorrow, would your revenue grow, hold, or shrink? Below 100%, the base leaks. Above 100%, it compounds on its own. That is why NRR is the most valuable engine in software, and the one a marketing-led go-to-market usually forgets by looking only at the mouth of the funnel.
What is net revenue retention?
Net revenue retention measures how much recurring revenue you keep from your existing customers over a period, including what you gain from expansion and lose to contraction and churn, but excluding any new logos.
Take a cohort of customers and the recurring revenue they generated at the start of the window. A year later, some upgraded (expansion), some downgraded (contraction), and some left (churn). NRR is what's left of that same cohort, expressed as a percentage of where it started. New customers acquired during the window don't count. That exclusion is the whole point: NRR isolates the health of the relationships you already have from the noise of your sales machine.
The NRR formula
$$ \text{NRR} = \frac{\text{Starting MRR} + \text{Expansion} - \text{Contraction} - \text{Churn}}{\text{Starting MRR}} \times 100 $$
Use the same period and the same cohort in every term. If you start with $100,000 in MRR, add $20,000 in expansion, and lose $12,000 to churn with no downgrades, your NRR is (100,000 + 20,000 − 0 − 12,000) / 100,000 × 100 = 108%. The base grew 8% without a single new customer.
Plug your own numbers into the calculator and it will tell you which side of 100% you're on, and what that means.
{/* calculator:net-revenue-retention — rendered from frontmatter.calculator (inputs: mrr_start, expansion, contraction, churn) */}
NRR vs GRR vs NDR: don't confuse the three
Three acronyms circle the same idea, and mixing them up is how boardrooms end up arguing about numbers that were never comparable.
- NRR (Net Revenue Retention) includes expansion. It can exceed 100%, because a customer who doubles their seats more than offsets one who left. This is the growth metric.
- GRR (Gross Revenue Retention) strips expansion out. It counts only what you lost to churn and contraction, so it is capped at 100% and never higher. GRR tells you how leaky the bucket is on its own, before any upsell papers over the holes. A company can post a healthy 115% NRR while its GRR quietly sits at 82%, meaning it's expanding fast enough to hide a serious retention problem. Always read them together.
- NDR (Net Dollar Retention) is not a third metric. It's the same calculation as NRR, named in dollars instead of "revenue". Some investors prefer "net dollar retention" because it makes the currency explicit; the formula is identical. If someone reports NDR and NRR as different numbers, ask which cohort and which period they used, because the metric is the same.
The practical rule: quote GRR and NRR side by side. GRR shows the size of the hole; NRR shows whether expansion is filling it faster than it drains.
The period error nobody catches
Here's the mistake that quietly ruins the metric: an NRR figure with no stated period means nothing. A monthly NRR of 100% and an annual NRR of 100% describe completely different businesses. The monthly one, compounded over a year, is a slow leak; the annual one is flat.
Always say the window out loud. Write "12-month NRR" or trailing-twelve-month NRR, not just "NRR". When a founder tells you their retention is "around 105", the first question isn't whether that's good, it's "over what period?" Half the retention debates in SaaS evaporate once both sides agree on the window.
Below 100% is a leaky bucket: retain before you acquire
When NRR sits under 100%, your existing base shrinks every period. Pour more acquisition into it and you're pouring water into a bucket with a hole: you can run the tap harder, but the level barely moves, and you're paying CAC to stand still.
This reorders your whole go-to-market. If your NRR is 85%, the highest-leverage work isn't more leads, it's the right side of the funnel: a clean sales-to-onboarding handover, a fast time-to-first-value so accounts reach their "aha" before they drift, a health score that flags at-risk accounts before the renewal instead of the day they give notice. Expansion is designed, not hoped for. Snowflake, Rippling and Datadog don't post 120%+ NRR by luck; they instrument land-and-expand so the same account grows in seats, teams and plan without paying CAC again.
There's also a leak nobody decides: failed payments. According to public data from ProfitWell and Paddle, involuntary churn (expired cards, declined charges) accounts for a quarter to 40% of all cancellations in card-billed SaaS. HubSpot found a chunk of its churn wasn't people leaving at all, but users whose browser extension had stopped loading or who'd accidentally logged out. Fixing that plumbing lifted retention more than any campaign. Split voluntary from involuntary churn in your dashboard; they're two different diseases with the same symptom.
For the unit-economics side of this same picture, see the LTV/CAC ratio, and for whether your GTM spend is actually converting to growth, the SaaS magic number.
NRR benchmarks by motion: the anti-magic-number
Ask "what's a good NRR?" and you'll get a number back. Ignore it. There is no magic NRR, and treating one as universal is how good businesses get judged as broken.
Two giants can sit at opposite ends of the NRR range and both be perfectly healthy. The difference isn't that one is thriving and the other is broken; it's how much of the value each charges on day one. A company that sells everything upfront (full-price seats, the whole platform at signature) has little room left to expand, so an NRR that only just clears its starting base can be perfectly healthy. A company that lands small and grows with usage (consumption pricing, seat-by-seat adoption, a low entry plan) leaves headroom on the table by design, so it should run well into expansion territory, or something is wrong.
So the honest benchmark is contextual: contextual — no magic number. The right way to use it isn't as a target to hit but as a place to look. If a usage-priced product posts 105% NRR, the flat expansion is the story, not the passing grade. If a fully-loaded enterprise deal posts 102%, that may be excellent. Read NRR against your pricing and motion, never against a leaderboard.
The book puts real numbers on this with Flowdesk, a €12k-ACV SaaS. On the same €100k base, worked one way it bled to 72% NRR; with the right side built (handover, time-to-first-value, QBR, expansion triggers) the identical customers grew to 108%. Same base, opposite engine, purely down to how retention and expansion were designed.
How to read your own NRR
- State the period. 12-month NRR or nothing.
- Pair it with GRR. High NRR plus low GRR means expansion is masking a churn problem you'll pay for later.
- Judge it against your pricing, not a benchmark. Upfront-charging models sit lower and are fine there; land-and-expand models should sit higher.
- Split involuntary churn out. If a fifth of your losses are failed cards, that's plumbing, not a value problem.
- Act on the diagnosis. Under 100% means fix the hole before opening the tap wider. This whole logic sits inside a broader go-to-market strategy, where retention is the right half of the funnel.
Frequently asked questions
Is net revenue retention the same as net dollar retention?
Yes. NRR and NDR are two names for the same calculation. "Net dollar retention" just makes the currency explicit. If two reports show different figures, the difference is the cohort or the period, not the metric.
What's the difference between NRR and GRR?
GRR (gross revenue retention) excludes expansion and is capped at 100%; it measures pure leakage from churn and contraction. NRR includes expansion and can exceed 100%. Read them together: GRR shows the hole, NRR shows whether upsell is filling it.
What is a good NRR for SaaS?
There's no universal number. It depends on how much value you charge upfront. Businesses that expand with usage tend to run higher; businesses that sell the full package at signature run lower and can still be healthy. Benchmark against your own pricing motion, not a headline figure.
Can NRR be over 100% while the company is still in trouble?
Yes. A high NRR can hide a weak GRR: expansion from a few accounts masks heavy churn underneath. It can also hide low new-logo acquisition. NRR is powerful precisely because it excludes new customers, but that's also why it should never be read alone.
Written by Mario Hernández, GTM consultant and founder of creactia, where he helps B2B software teams build the right side of the funnel, not just the tap. He is the author of «Go-To-Market: from zero to specialist», the field manual this guide draws its cases and diagnostics from.
If these numbers look wrong for your startup, that diagnosis is exactly what our GTM audit does — we read your retention against your motion and tell you where the bucket leaks.