Net revenue retention (NRR, also called net dollar retention) answers one question: if you signed zero new customers this year, would your revenue grow or shrink? It is the cleanest read on whether the business you have already built compounds on its own, or quietly leaks. It is also the most casually mis-stated metric in reporting: the formula looks simple, so two people assume they are calculating the same thing when they are not.
It applies to any company with recurring revenue, not just SaaS. Subscription products, managed services, agencies on retainer, maintenance contracts, and equipment leasing all have an installed base that can expand, shrink, or leave. If your revenue renews, NRR applies to you.
The formula
NRR measures one cohort of customers across one window:
NRR = (starting ARR + expansion - contraction - churn) / starting ARR
- Starting ARR is recurring revenue from a fixed set of customers at the start of the window.
- Expansion is additional revenue from those same customers: upsells, cross-sells, seat growth, usage growth, price increases.
- Contraction (downsell) is revenue lost from customers who stayed but shrank: fewer seats, a downgrade, a renegotiated rate.
- Churn is revenue from customers in that cohort who left entirely.
KeyBanc Capital Markets and Sapphire Ventures define it identically in their 2024 SaaS Survey: net dollar retention = (Beginning ARR + Expansion ARR - Churned ARR - Downsell ARR) / Beginning ARR, with gross dollar retention the same minus the expansion term. Note the absence: there is no term for new customers.
Why excluding new customers is the whole point
The cohort is fixed at the start of the window. Every customer you sign after that date is invisible to the calculation, no matter how much they pay you. NRR exists to isolate one variable: what happens to revenue you have already won. New sales are the output of your sales engine; retention and expansion are the output of your product and delivery. Mix them and the number tells you nothing reliable about either: one strong quarter of new sales can mask a base falling apart underneath it.
A worked calculation
Take a company with a recurring-revenue base on January 1. Fix the cohort there and follow those customers for twelve months.
| Component | Amount | Running ARR |
|---|---|---|
| Starting ARR (40 customers, Jan 1) | $2,000,000 | $2,000,000 |
| Plus expansion (upsells, seats, price increases) | +$260,000 | $2,260,000 |
| Less contraction (downgrades, seat cuts) | -$90,000 | $2,170,000 |
| Less churn (customers who left entirely) | -$150,000 | $2,020,000 |
| Ending ARR from that same cohort | $2,020,000 | $2,020,000 |
NRR = $2,020,000 / $2,000,000 = 101%. The base you started with is worth slightly more than it was, despite losing customers along the way.
Gross retention strips out expansion and asks only what you kept: ($2,000,000 - $90,000 - $150,000) / $2,000,000 = $1,760,000 / $2,000,000 = 88%.
Now the part that matters. Suppose this company also signed $500,000 of new business during the year. Total ending ARR is $2,520,000, and dividing that by starting ARR gives 126%. That is not NRR. It is a different metric wearing NRR's name, and it is 25 points more flattering. This is the most common way NRR gets inflated, and it rarely happens on purpose: someone pulled total ending ARR instead of cohort ending ARR.
101% or 126%, differing only on whether new customers were excluded. Before comparing your NRR to anyone else's, confirm you both fixed the cohort and left new logos out.
Gross vs net retention, and why you need both
NRR of 101% reads as stable. GRR of 88% is less comfortable: the company lost 12% of its starting base over the year, and expansion covered it up.
- Gross retention is capped at 100% and can only go down. It measures whether customers stay and keep paying what they paid, the honest read on whether your product is holding.
- Net retention has no ceiling, because expansion has none. It measures the trajectory of your base, not its stability.
A high NRR on a weak GRR is fragile: a few growing accounts are paying for a leaky base, and it holds until those accounts stop expanding. That is why a serious board reporting package carries both numbers.
What over 100% actually means
NRR above 100% means the cohort you started with is worth more now than it was then: expansion out-ran everything you lost. The company grows with zero new sales, compounding without paying acquisition cost.
It also has a second-order effect. Revenue from an existing customer generally carries far lower acquisition cost than revenue from a new one, so a base that expands on its own structurally lowers the cost of growth. That connects NRR to gross margin and to cash efficiency. For that side, see burn multiple, where net new ARR already bakes retention in, and good burn rate.
The definition traps that make NRR non-comparable
Two companies can report the same NRR and not be measuring the same thing. The choices that move it:
- The window. Annual is standard. A monthly NRR annualized, or a trailing-twelve-month figure, will not equal an annual cohort figure from the same data. Short windows flatter you: churn takes time to surface.
- Cohort definition. Fixing the cohort on day one and following it twelve months is clean. Re-setting it monthly and chaining results smooths over bad months.
- The revenue base. Recurring revenue only, or services, usage overages, one-time fees too? Overage-heavy bases swing NRR on seasonality alone.
- Churn timing. Recognized at notice, at contract end, or when the paid term runs out? Each choice pushes churn into a different window.
- Segment mix. A blended NRR across an SMB base and an enterprise base describes neither.
- Exclusions. Some exclude customers below a revenue floor, non-core products, or accounts lost to acquisition. Each one moves the number, almost always upward.
None of these are fraud. They are choices, rarely disclosed, so the number arrives without its definition and comparison becomes meaningless. Pick a definition once and calculate it the same way every period as part of your monthly close.
What good actually looks like
"120% is best in class" circulates constantly with no source attached. It traces to Bessemer Venture Partners' State of the Cloud 2023, which recommends founders steer toward net revenue retention of 100% (good), 110% (better), and 120%+ (best). Read the framing: those are fundability goalposts Bessemer suggests aiming at, not a measurement of what companies achieve. An aspiration became a "benchmark."
The measured data is more sober. In the 2024 KeyBanc Capital Markets and Sapphire Ventures SaaS Survey, median net dollar retention among private SaaS respondents was 102% in 2022, 102% in 2023, and an expected 101% for 2024. The top quartile reported 110%, 108%, and an expected 109%. Median gross dollar retention ran 89%, 89%, and an expected 91%, on roughly 48 respondents: directional, not definitive.
SaaS Capital's 2025 benchmarks, measuring December 2023 against December 2024, agree: for private B2B SaaS companies with $25,000 to $50,000 average contract values, median NRR was 102%, top quartile 111%, bottom quartile 97%.
Two independent surveys put the median near 102% and the top quartile around 108-111%, so the repeated 120% bar sits above both. Treating an aspiration as an expectation makes a healthy business look broken on its own board deck. Your own consistent trend line tells you more than any cross-company comparison.
Most NRR figures in circulation have no traceable publisher, no year, and no stated definition. If you cannot see who measured a number and how, it is not a benchmark.
Getting the number right
NRR is only as good as the revenue data underneath it. If recurring revenue is not cleanly separated from one-time revenue, if expansion is indistinguishable from new business, or if churn is recorded whenever somebody remembers to, the metric will be confidently wrong in ways that are hard to see. That is a bookkeeping problem before a metrics problem.
Consistent bookkeeping and a disciplined monthly close make NRR trustworthy enough to steer with. A fractional CFO can set the definition, hold it stable, segment it where a blended figure would mislead, and present it with the gross number beside it. For the short definition, see net revenue retention in our glossary.
This post is educational and general: how you define and report NRR depends on your contracts and billing model.