# Net Revenue Retention (NRR) > How to calculate NRR correctly, the benchmark by segment from 97 to 118 percent, and the marketing levers that move it, with the common calculation errors. Source: https://saas-marketing.net/guides/net-revenue-retention/ Topic: SaaS Metrics and Analytics Type: guide Published: 2026-09-11 Last updated: 2026-09-11 Publisher: SaaS Marketing (saas-marketing.net) License: CC BY 4.0. Quote or republish with attribution and a link to https://saas-marketing.net/guides/net-revenue-retention/ ## Short answer Net revenue retention measures how much recurring revenue a fixed cohort of customers generates a year later, including expansion, contraction and churn, and excluding any new logos. The formula is starting ARR plus expansion minus contraction minus churn, divided by starting ARR. Median B2B SaaS NRR sits near 101 percent, with enterprise around 118 percent, mid-market near 108 percent and SMB around 97 percent. Anything above 110 percent means the installed base grows without new sales. ## Key takeaways - NRR separates top from bottom quartile SaaS, roughly 113 percent against 98 percent in McKinsey's B2B software analysis. - New logos must be excluded from the numerator, or you are measuring growth and calling it retention. - Median NRR has declined from around 105 percent to roughly 101 percent since 2022 as seat expansion slowed. - Expansion ARR is roughly 40 percent of new ARR at scale, and almost nobody runs marketing campaigns against it. - Enterprise segments hit 118 percent, SMB sits near 97 percent, so a blended company number hides both stories. - Reactivations counted as expansion inflate NRR by two to four points and are the most common quiet error. --- NRR is the number that separates a top-quartile SaaS business from a bottom-quartile one, and marketing teams almost universally treat it as somebody else's problem. McKinsey's B2B software work puts the top quartile near 113 percent and the bottom near 98 percent. That 15-point gap compounds. Over four years it is the difference between an installed base that grows by half on its own and one that quietly shrinks. ## The formula, and the cohort window that decides whether it means anything Net revenue retention is starting ARR, plus expansion, minus contraction, minus churn, divided by starting ARR. New logos are excluded from both sides. ``` NRR = (Starting ARR + Expansion - Contraction - Churn) ÷ Starting ARR ``` Worked example. On 1 September 2025 you had 400 customers generating 12,000,000 dollars of ARR. Over the next twelve months those same 400 customers added 1,900,000 in expansion, downgraded by 450,000, and 680,000 churned entirely. You also signed 130 new customers worth 4,200,000, which does not appear anywhere in this calculation. NRR = (12,000,000 + 1,900,000 - 450,000 - 680,000) ÷ 12,000,000 = 12,770,000 ÷ 12,000,000 = 106.4 percent. The cohort window is what makes or breaks this. You freeze the customer list on day one and follow only those accounts for twelve months. The moment the denominator drifts, the number stops being retention and becomes a blended growth statistic. The [NRR and churn calculator](/calculators/nrr/) does the arithmetic if you want to check your own figures against a clean method. **113% vs 98%** Top-quartile versus bottom-quartile B2B software NRR ## Four calculation errors that inflate NRR Every one of these shows up in board decks, and every one flatters the number. Three of them are honest mistakes. One of them is usually not. **Including new customers.** If a customer signed in March and expanded in July, that expansion does not belong in a cohort that started in January. Including it can add five to ten points on a fast-growing base. This is the single most common error and the easiest to catch: ask whether any revenue in the numerator came from an account that didn't exist in the denominator. **Trailing-twelve-month windows on a growing base.** A rolling window that keeps absorbing new accounts systematically overstates retention while you're growing, because recent cohorts haven't had time to churn yet. Survival bias, dressed as a metric. **Counting reactivations as expansion.** A customer who churned in month three and came back in month nine is not expansion. They are a new logo with a familiar name. Netting this in typically adds two to four points and it is the quietest of the four. **Netting at company rather than cohort level.** Aggregating all expansion against all starting ARR across mixed cohorts blends segments with very different behaviour, so an enterprise base at 118 percent hides an SMB base bleeding at 92 percent. You learn nothing and act on nothing. Ask for NRR and GRR side by side, split by segment. If someone can only produce a single blended NRR figure, the underlying data model probably cannot support the split, which means nobody has actually looked at where the leakage is. The [NRR versus GRR comparison](/comparisons/nrr-vs-grr/) sets out why you need both. ## Benchmarks by segment, and why the blended number lies A company-wide NRR figure is an average of at least two very different businesses. Here is what the segments actually look like. Two structural points fall out of that table. Usage-based pricing produces the highest NRR because the customer expands without a purchase decision, which is how Snowflake and Datadog post consumption-driven expansion numbers that seat-based peers can't match. And SMB NRR near 97 percent is not a failure, it is the arithmetic of selling to businesses that shrink and close, so an SMB product with 100 percent NRR is genuinely excellent. The multi-year trend matters too. Median NRR across B2B SaaS has fallen from roughly 105 percent to around 101 percent since 2022, driven by headcount flattening and procurement teams running licence audits. If your NRR dropped three points last year, part of that is you and part of it is the market, and separating the two is the first analysis to run. There is segment-level data in the [NRR and churn benchmarks](/research/nrr-and-churn-benchmarks/). ## Why NRR belongs in the marketing scorecard Here is the argument. At scale, expansion ARR runs at roughly 40 percent of new ARR. So a company signing 20 million in new business is generating another 8 million from the installed base. In most orgs, zero marketing headcount is assigned to that 8 million. Customer success owns the relationship and the renewal conversation. That's right, and it should stay that way. But CSMs work an account list of 30 to 80 customers one at a time, which is the wrong instrument for programmatic demand generation. Announcing a new module to 3,000 existing accounts, segmenting by which of them have the adjacent use case, and running a nurture sequence to the non-users inside those accounts is a marketing job. It always was. The practical move is to add two lines to the marketing scorecard: expansion pipeline sourced by marketing, and product adoption rate for accounts touched by an adoption campaign against a matched control. Both are measurable inside a quarter. Both connect directly to NRR, and both sit alongside acquisition metrics like [CAC payback period](/guides/cac-payback-period/) and [LTV to CAC ratio](/guides/ltv-cac-ratio/) rather than replacing them. Expansion revenue typically carries a customer acquisition cost between one fifth and one third of new-logo CAC, because you already have the relationship, the contract and the security review. It is the cheapest revenue in the business and the least marketed-to. ## The four marketing levers that actually move NRR Not everything that touches customers moves the number. These four do, in rough order of impact per hour invested. **Running marketing against the installed base** The honest tradeoff: adoption campaigns cannibalise some expansion that would have happened anyway, so the uncontrolled numbers will look better than reality. Always run a holdout. A 10 percent holdout on a 3,000 account base costs you very little and is the only way to know whether the campaign did anything. Most teams skip it and then cite an inflated number for two years. Generic customer newsletters do not move NRR. Neither do quarterly business reviews as a blanket motion, once you are past a few hundred accounts. Both feel like customer marketing and neither is triggered by behaviour, which means neither reaches the right account at the moment the decision is live. ## Reading NRR alongside the rest of the metric set NRR on its own is easy to misread. Three pairings that give it meaning. | Pair it with | What the combination tells you | Warning sign | | --- | --- | --- | | GRR | Whether expansion is masking churn | NRR 120 percent, GRR 80 percent means a few whales are carrying you | | CAC payback | Whether growth is efficient as well as retained | Long payback plus high NRR means you can afford it, the reverse does not | | Quick ratio | Growth quality: new plus expansion over churn plus contraction | Below 2 means you are refilling a leaking bucket | | Logo retention | Whether revenue retention hides customer count decline | Revenue flat, logos down 15 percent means concentration risk | The [SaaS quick ratio](/glossary/saas-quick-ratio/) is the fastest of these to compute and the most diagnostic when NRR looks fine but growth has stalled. The [SaaS magic number](/guides/saas-magic-number/) tells you whether the sales and marketing spend behind that retention is justified. All of them sit in the broader [SaaS metrics and analytics](/saas-metrics/) framework, and a first pass on efficiency is easiest via the [CAC payback calculator](/calculators/cac-payback/). **NRR reporting audit** ## What to do next Recalculate NRR from a frozen cohort this quarter and compare it to whatever number is currently in the board deck. If the two differ by more than two points, find out which of the four errors caused the gap before you do anything else, because every downstream decision has been made on the wrong figure. Then pick one feature adoption campaign, one segment, one holdout group, and run it for 60 days. That single test tells you whether marketing can move the number in your business. If it can, the case for putting expansion pipeline on the marketing scorecard writes itself, and the 8 million that nobody is working stops being nobody's job. ## Frequently asked questions ### What is a good NRR for a SaaS company? It depends entirely on segment. Enterprise SaaS should target 115 percent or higher, mid-market around 108 percent, and SMB products are doing well at 97 to 100 percent because small businesses fail and downgrade at higher rates. Above 120 percent is exceptional and usually means usage-based pricing. Below 90 percent in any segment signals a product or fit problem rather than a marketing one. ### What is the NRR formula? Take the ARR of a fixed cohort of customers at the start of the period, add expansion revenue from those same customers, subtract contraction and churn from those same customers, and divide by the starting ARR. New customers acquired during the period are excluded entirely. Express it as a percentage. The cohort must be frozen at period start for the number to mean anything. ### What is the difference between NRR and GRR? Gross revenue retention excludes expansion, so it can never exceed 100 percent. It measures pure leakage: how much of the starting ARR you kept. NRR includes expansion and can exceed 100 percent. Look at both. A company with 120 percent NRR and 82 percent GRR is masking serious churn with a handful of large expansions, which is fragile. ### Should marketing own NRR? Marketing should own a share of it. Expansion ARR is roughly 40 percent of new ARR at scale, and the campaigns that drive adoption, seat growth and tier upgrades are marketing campaigns, run to an installed base rather than a cold market. Customer success owns the relationship and the save motion. Marketing owns the programmatic demand generation into existing accounts. ### Why is my NRR declining? Since 2022 the most common cause is seat compression: customers renewing at lower headcount after layoffs, plus procurement teams actively auditing licence counts. The second cause is that your expansion path is seat-only, so a flat-headcount customer has no way to spend more with you. Usage or outcome-based tiers give a customer a route to grow without hiring. ### How often should NRR be calculated? Quarterly for reporting, monthly for internal tracking, always on a trailing twelve month cohort basis. Monthly snapshots of a rolling window are noisy and, on a fast-growing base, systematically flattering. Pick the cohort that started twelve months ago, follow only those customers, and report that. Consistency of method matters more than frequency.