NRR (net revenue retention): formula, calculator and 2026 benchmarks
What is NRR?
NRR (net revenue retention) is the share of recurring revenue a company keeps from the customers it had a year ago, after expansion, downgrades and churn: NRR = (starting ARR + expansion − contraction − churn) ÷ starting ARR. Above 100%, existing customers grow revenue on their own. New customers are left out.
Key facts
- The SaaS Metrics Standards Board recommends a cohort: "The cohort method is the most accurate way to measure NRR and is the preferred approach in most situations." (SaaS Metrics Standards Board)
- The median private B2B SaaS company had NRR of 101% in SaaS Capital's 2025 survey of more than 1,000 companies. (SaaS Capital, 2025)
- The median public software company in Clouded Judgement's comps had net retention of 110% in September 2026, down from 120% in June 2022. (Clouded Judgement, Sep 2026)
- Across 25 public software companies, net dollar retention fell from 125% in 2022 to 112% in 2025, Tomasz Tunguz found. (Tunguz, 2026)
- "Net revenue retention" appeared in 36 annual reports filed with the SEC in 2025, up from 7 in 2020 and none in 2015 (our EDGAR full-text searches). (SEC EDGAR search)
NRR calculator
Take the customers you had 12 months ago. Enter their ARR then, and what happened since: expansion, downgrades and churn. Leave out anything from customers who joined during the year. The calculator gives NRR and gross revenue retention together.
Want NRR by cohort from your real billing data? CX Cash is being built to track expansion, contraction and churn by customer, month by month.
How to calculate NRR
NRR = (Starting ARR + Expansion − Contraction − Churn) ÷ Starting ARR
In words: take the recurring revenue of last year's customers today, including what they added and net of what they cut or cancelled, and divide by what the same customers paid a year ago.
- Starting ARR: Recurring revenue 12 months ago from the customers in the cohort.
- Expansion: Upsells, cross-sells and price increases from those customers.
- Contraction and churn: Downgrades from customers who stayed, and the revenue of customers who left.
Worked example
An illustrative cohort with $1.0M of ARR a year ago. Since then, those customers added $180,000 through upsells, cut $30,000 through downgrades, and $50,000 left with customers who churned. The results below are computed by the calculator's tested code.
| Input | Value |
|---|---|
| ARR from the cohort 12 months ago | $1,000,000 |
| Expansion ARR | $180,000 |
| Contraction ARR | $30,000 |
| Churned ARR | $50,000 |
Net revenue retention: 110.0% Gross revenue retention: 92.0%Cohort ARR today: 1,100,000
NRR = (1,000,000 + 180,000 − 30,000 − 50,000) ÷ 1,000,000 = 110.0%
NRR of 110% and GRR of 92%: the cohort lost 8% to downgrades and churn but grew 18% through expansion. At 110% a year, existing customers alone double revenue in about seven and a quarter years, without a single new customer (our arithmetic).
In a spreadsheet
- NRR, with starting ARR in B2 and expansion, contraction and churn in C2:E2: =(B2+C2-D2-E2)/B2
- Cohort NRR, with each customer's ARR a year ago in B2:B500 and today in C2:C500 (churned = 0): =SUM(C2:C500)/SUM(B2:B500)
- Annualize a six-month NRR in B2, as the Standards Board describes: =B2^2
- Years for existing customers to double revenue at the NRR in B2: =LN(2)/LN(B2)
How founders and investors read NRR
If you're the founder
- Above 100%, your installed base grows on its own, which makes every new customer worth more. Jason Lemkin's rule of thumb: "110% NRR: Your revenue doubles in 8 years even with no new customers" (SaaStr, 2021).
- Investors will ask how it's calculated. Use a cohort, compare per-customer revenue a year apart, and say how you treat usage and price increases.
- A high NRR can hide weak new sales. Lemkin again: "High NRR can mask mediocre growth in new logos and customers" (SaaStr, 2026).
Questions you'll be asked
- Is this cohort-based, and on ARR or revenue?
- What is your gross revenue retention?
- How much of expansion is price increases rather than more usage or seats?
- How does NRR differ by customer size or segment?
If you're the investor or LP
- Ask for the definition. Among public companies, Okta uses contract value, Twilio quarterly revenue, Snowflake a two-year cohort of product revenue, and Zoom reports enterprise customers only.
- Read it with GRR. Two companies at 110% NRR can lose 5% or 20% of revenue a year to churn and downgrades; GRR shows which.
- Expect it to lag. Jamin Ball: "Net retention is more of a lagging indicator" (Clouded Judgement, 2023).
Questions to ask the company
- How is NRR calculated, and has the method changed?
- What are GRR and logo churn for the same period?
- How concentrated is expansion among the largest customers?
- How much of NRR comes from usage that could fall back?
What is a good NRR?
100% is the line: below it, existing customers shrink. Private SaaS companies sit just above it at the median; public companies, which tend to sell larger contracts, around 110%. Larger contracts retain better.
Show the numbers
| Annual contract value | Median | 75th percentile |
|---|---|---|
| Under $12k | 98.0% | 106.0% |
| $12k–25k | 103.0% | 115.0% |
| $25k–50k | 102.0% | 111.0% |
| $50k–100k | 104.0% | 110.0% |
| $100k–250k | 102.0% | 109.0% |
| Over $250k | 106.0% | 110.0% |
Benchmarks
| Segment | Measure | Typical | Top quartile | As of | Sample | Source |
|---|---|---|---|---|---|---|
| Private SaaS companies · Self-reported surveys with different populations; compare a company with one of them, not across them. | ||||||
| Private B2B SaaS | Median NRR | 101% | — | Dec 2024 | More than 1,000 companies (SaaS Capital) | SaaS Capital, 2025 |
| Bootstrapped / equity-backed | Median NRR | 104% / 101% | — | Dec 2024 | SaaS Capital survey | SaaS Capital, 2025 |
| B2B SaaS and AI-native | NRR (formula not stated) | 102% (25th percentile 92%) | 110% | CY 2025 | 230 companies (Benchmarkit) | Benchmarkit, 2026 |
| Public software companies | ||||||
| Public SaaS comparables | Median net retention, disclosers only | 110% (25th percentile 102%) | 116% | 25 Sep 2026 | About 70 companies tracked (Clouded Judgement) | Clouded Judgement, Sep 2026 |
| Public SaaS comparables | Median net retention | 120% | — | Jun 2022 stale | Clouded Judgement comps at the 2022 peak | Clouded Judgement, Jun 2022 |
| 25 public software companies | Net dollar retention | 112% in 2025 (125% in 2022) | — | Mar 2026 | 374 quarterly observations (Tunguz) | Tunguz, 2026 |
| Rules of thumb | ||||||
| Bessemer | Good, better, best | 100%, 110%, 120%+ | — | Apr 2023 | VC rule of thumb | Bessemer, 2023 |
Contract size matters most. In SaaS Capital's data, NRR runs from a 98% median under $12k of annual contract value to 106% above $250k.
Billing-data samples run much lower than surveys. ChartMogul's data, which skews to small self-serve companies, put median B2B NRR at 82% in 2025 (ChartMogul), so don't mix its figures with survey medians.
Compiled Oct 2, 2026. Next review: December 2026, after SaaS Capital's 2026 retention report. Left out: High Alpha 2025 by ARR band (Marked proprietary and confidential; we don't reproduce its tables); ICONIQ 2026 (The report prohibits reproduction, and its sample is ICONIQ's portfolio); ChartMogul NRR in the tables (Billing data on a different basis from surveys; mentioned in the notes only). Download these benchmarks as CSV.
NRR in the wild: nine public companies, nine definitions
The latest figure each company reported, with the basis it uses. Twilio's own release warns that its measure "is not based on any standardized industry methodology". (Twilio, Aug 2026)
| Company | Reported | As of | Basis | Source |
|---|---|---|---|---|
| Snowflake | 126% | 31 Jul 2026 | Product revenue of a cohort, year 2 against year 1 | 10-Q |
| MongoDB | About 121% | 31 Jan 2026 | ARR, including annualized usage | 10-K |
| Datadog | In the low 120%s | 30 Jun 2026 | ARR, trailing 12-month weighted average | 10-Q |
| Cloudflare | 120% | Q2 2026 | Quarterly revenue × 4, same customers a year apart | 10-Q |
| GitLab | 118% | FY to 31 Jan 2026 | ARR | 10-K |
| Twilio | 116% | Q2 2026 | Revenue, quarter against the same quarter a year earlier | Release |
| CrowdStrike | 115% | 31 Jan 2026 | ARR, excluding incident response services | 10-K |
| Okta | 107% | 12 months to 31 Jul 2026 | Annual contract value | Commentary |
| Zoom | 98% | 31 Jan 2026 | ARR, enterprise customers only, 12-month average | 10-K |
From 98% to 126%, but not on one scale. Snowflake measures product revenue over two years; Zoom counts only enterprise customers; Twilio compares usage revenue quarter by quarter, so a price cut lowers it. Most use ARR and exclude new customers, the common ground of the definitions.
Figures and definitions are each company's own, as filed or published. SEC filings were read from archived copies because sec.gov blocks automated access; the links go to EDGAR.
NRR vs GRR, logo churn, quick ratio and ARR growth
| Metric | Answers | Ignores | Use it when | It misleads when |
|---|---|---|---|---|
| NRR | Revenue kept and grown from existing customers | New customers, and how expansion and churn net out | Judging the strength of the installed base | Expansion from a few large customers hides heavy churn |
| GRR | Revenue kept from existing customers, before expansion | Upsell and expansion | Measuring churn and downgrades on their own | Read alone in a land-and-expand business |
| Logo churn | Share of customers lost | How big the lost customers were | Many small customers, or product fit | A few large customers drive revenue |
| SaaS quick ratio | MRR added per $1 of MRR lost | The size of the base | Monthly growth efficiency | Growth is small relative to the base |
| ARR growth | Total recurring revenue growth, new customers included | Where growth came from | Measuring scale | Strong new sales hide a leaky base |
NRR − GRR = expansion ÷ starting ARR, so NRR is GRR plus expansion. ARR growth over a year is roughly NRR − 100% plus new customers' ARR as a share of starting ARR.
Common mistakes: how NRR gets flattered
- Including new customers. NRR measures last year's customers only. CrowdStrike's 10-K, for one, "excludes revenue from new subscription customers in the current period".
- Averaging percentages instead of summing revenue. Tomasz Tunguz showed that the average of per-customer rates and the growth of the total give different answers. SaaS Capital's advice: "use per-customer data to compare periods, and only sum the customers after you've compared periods."
- Using bookings or cash. Multi-year prepayments and billing timing distort it. The Standards Board says to use revenue, "not bookings, billings, or cash accounting".
- Annualizing a short period by multiplying. A quarter's NRR should be compounded, not multiplied by four: "it can be annualized by taking the result to the appropriate power".
- Counting price increases as product success. A list-price rise lifts expansion without customers using more. Report price-driven expansion separately.
- Comparing across definitions. Public companies use ARR, contract value or revenue, and point-in-time or averaged rates. Compare like with like.
- Quoting the highest of several rates. One company can report several. Fastly's 10-K for 2022 gave 122.7%, 119.1% and 110.7% for the same customers at the same date. The highest, Fastly disclosed, leaves out revenue lost from former customers (Fastly, 10-K).
- Reading a threshold as a number. "Greater than 130%" is a floor. Zoom reported 129% for January 2022 against "greater than 130%" a year earlier, for customers with more than 10 employees. SEC staff asked for the actual figure in September 2022. Zoom's reply gave 164%, so the drop was 35 points (our arithmetic) (SEC staff letter, 2022; Zoom reply, 2022).
Where NRR came from and how it's regarded today
| When | What happened | Source |
|---|---|---|
| Sep 2026 | The median public software company in Clouded Judgement's comps reports net retention of 110%. | Clouded Judgement, Sep 2026 |
| Mar 2026 | Tomasz Tunguz finds net dollar retention fell from 125% in 2022 to 112% in 2025. | Tunguz, 2026 |
| 2025 | "Net revenue retention" appears in 36 annual reports filed with the SEC, against none in 2015. | SEC EDGAR search |
| May 2024 | Ray Rike finds growth and the Rule of 40 now explain valuation multiples better than NRR does. | Rike, 2024 |
| Jul 2024 | The public median slides to 110% in Clouded Judgement's comps. | Clouded Judgement, Jul 2024 |
| 2023 | The SaaS Metrics Standards Board publishes a cohort-based NRR standard. | SaaS Metrics Standards Board |
| Jun 2022 | The public median peaks at 120%. | Clouded Judgement, Jun 2022 |
| Jan 2020 | The SEC asks companies to define key metrics like NRR, explain them and disclose changes in method. | SEC, 2020 |
| Dec 2017 | Christoph Janz notes that most public SaaS companies report churn as dollar-based net retention. | Janz, 2017 |
| Jun 2016 | Twilio's IPO prospectus reports a dollar-based net expansion rate of 155% for 2015. | Twilio, 2016 |
| Oct 2015 | Instructure's S-1 is the first public IPO filing to use "net revenue retention rate". | Instructure, 2015 |
| Mar 2014 | SEC staff ask Zendesk to treat its dollar-based net expansion rate as a non-GAAP measure. | SEC staff, 2014 |
| Mar 2012 | Eloqua's IPO filing is the earliest SEC filing to use "net dollar retention". | Eloqua, 2012 |
How it's regarded today
- Lower than at the peak. Lemkin: "high NRR has gotten rarer than it used to be. Median NRR for public B2B companies is now around 108-110%" (SaaStr, 2026).
- The bottom quartile is at breakeven. Tunguz: "The 25th percentile fell from 106% to 101% in a single quarter, now touching the breakeven line" (Tunguz, Mar 2026).
- Still a top predictor. Kyle Poyar: "The two strongest predictors of long term and profitable growth are CAC payback period and net revenue retention (NRR)" (Growth Unhinged, 2025).
- Usage makes it volatile. Consumption-priced companies annualize usage into the base, and Confluent warns its method "may result in increased volatility in NRR".
- 2025 B2B SaaS Retention Benchmarks (Research Brief 32) (SaaS Capital, 2025). NRR and GRR for more than 1,000 private B2B SaaS companies by contract value, funding and contract length.
- The Sword of Damocles in Software (Tunguz, 2026). Public software net dollar retention fell from 125% in 2022 to 112% in 2025.
- Simpson's Paradox in Measuring Net Dollar Retention Rate (Tunguz, 2020). How the way you aggregate customers changes NRR (Simpson's paradox).
- Full-text searches for NRR's names in 10-K filings, by year (SEC EDGAR search). When NRR's different names entered SEC filings, and how often they appear.
What experts say about NRR
“Net retention is more of a lagging indicator, so I expect this downward trend to continue”
“The 25th percentile fell from 106% to 101% in a single quarter, now touching the breakeven line.”
“But here's the 2026 reality check: high NRR has gotten rarer than it used to be. Median NRR for public B2B companies is now around 108-110%.”
Where they disagree
A growth analyst and a benchmarker on how much NRR matters to a company's value.
“The two strongest predictors of long term and profitable growth are CAC payback period and net revenue retention (NRR).”
“In May, 2024 you can see that both Revenue Growth Rate (R^2 = .33) and Rule of 40 (R^2 = .28) have more impact on EV:NTM Revenue multiples than Net Revenue Retention with an R-Squared of .18.”
NRR FAQ
What is net revenue retention?
NRR is how much recurring revenue you have today from the customers you had a year ago, as a share of what they paid then. It includes their upsells and expansion, minus downgrades and cancellations, and leaves out new customers.
How do you calculate NRR?
Take last year's customers. Add up their ARR today, counting churned customers as zero, and divide by their ARR a year ago. Or use the shortcut: (starting ARR + expansion − contraction − churn) ÷ starting ARR.
What is a good NRR?
Above 100% means existing customers grow on their own. Private B2B SaaS medians were 101% to 102% in 2025, the private top quartile starts at about 110%, and the public median was 110% in September 2026. 120%+ is best in class.
Does NRR include new customers?
No. NRR follows a fixed group of customers from a year ago. Revenue from customers who joined since is new business, measured by ARR growth, not retention.
What is the difference between NRR and GRR?
GRR counts only what you keep, before any expansion, so it can't exceed 100%. NRR adds expansion. The gap between them is how much existing customers grew.
Is NRR the same as net dollar retention?
Yes. Net dollar retention (NDR), dollar-based net retention and net expansion rate are names for the same idea, though each company defines its own version.
How long does it take to double revenue from existing customers?
At 110% a year, a little over seven years; at 120%, about four (our arithmetic). Lemkin's rule of thumb is that at 110% revenue "doubles in 8 years even with no new customers".
Related metrics and guides
Sources
Every link was opened and checked. Archived copies guard against links that move or disappear.
Primary data
- Full-text searches for NRR's names in 10-K filings, by year. US Securities and Exchange Commission (EDGAR), Queried 2 Oct 2026. Our queries; counts are documents matching the exact phrase, and coverage starts in 2001.
- Eloqua prospectus (424B4). Eloqua (SEC filing), 2 Aug 2012. The earliest use, an S-1/A of 30 Mar 2012, was found by EDGAR search; the final prospectus is linked. Archived copy
- Instructure registration statement (Form S-1). Instructure (SEC filing), 9 Oct 2015. Read from the archived copy. Archived copy
- Twilio registration statement (Form S-1/A). Twilio (SEC filing), 13 Jun 2016. Read from the archived copy. Archived copy
- Twilio Announces Second Quarter 2026 Results. Twilio, 6 Aug 2026. Company newsroom copy (also filed with the SEC); no archived copy found.
- Form 10-Q for the quarter ended 31 July 2026. Snowflake (SEC filing), 4 Sep 2026. Read from the archived copy. Archived copy
- Form 10-K for the fiscal year ended 31 January 2026. MongoDB (SEC filing), 11 Mar 2026. Read from the archived copy. Archived copy
- Form 10-Q for the quarter ended 30 June 2026. Datadog (SEC filing), 6 Aug 2026. Read from the archived copy. Archived copy
- Form 10-Q for the quarter ended 30 June 2026. Cloudflare (SEC filing), 6 Aug 2026. Read from the archived copy. Archived copy
- Form 10-K for the fiscal year ended 31 January 2026. GitLab (SEC filing), 17 Mar 2026. Read from the archived copy. Archived copy
- Form 10-K for the fiscal year ended 31 January 2026. CrowdStrike Holdings (SEC filing), 5 Mar 2026. Read from the archived copy. Archived copy
- Q2 FY2027 posted earnings commentary. Okta (investor relations), 26 Aug 2026. No archived copy found.
- Form 10-K for the fiscal year ended 31 January 2026. Zoom Communications (SEC filing), 27 Feb 2026. Read from the archived copy. Archived copy
- Form 10-K for the fiscal year ended 31 December 2025. Confluent (SEC filing), 11 Feb 2026. Read from the archived copy. Archived copy
- Response letter to SEC staff comments on its Form 10-K for fiscal 2022. Zoom Video Communications (SEC correspondence), 19 Oct 2022. Read from the archived copy. Archived copy
- Form 10-K for the fiscal year ended 31 December 2022. Fastly (SEC filing), 27 Feb 2023. Read from the archived copy. Archived copy
- 2025 B2B SaaS Retention Benchmarks (Research Brief 32). SaaS Capital, Sep 2025. Lender's survey; chart figures read from the PDF; no archived copy found.
- 2026 SaaS and AI Metrics Benchmarks. Benchmarkit, 1 Jun 2026. Sponsored report; formula not stated. Archived copy
- The SaaS Retention Report: The AI churn wave. ChartMogul (Kyle Poyar), 10 Dec 2025. Vendor billing data, skewed to small self-serve companies; not comparable with surveys. Archived copy
Regulation
- Commission Guidance on Management's Discussion and Analysis of Financial Condition and Results of Operations (Release 33-10751). US Securities and Exchange Commission, 30 Jan 2020. sec.gov blocks automated access; read from the archived copy. Archived copy
- Staff comment letter to Zendesk on its draft registration statement. SEC Division of Corporation Finance, 13 Mar 2014. Read from the archived copy. Archived copy
- Staff comment letter to Zoom Video Communications on its Form 10-K for fiscal 2022. SEC Division of Corporation Finance, 28 Sep 2022. Read from the archived copy. Archived copy
Practitioner
- Net Revenue Retention (NRR). SaaS Metrics Standards Board, Undated (first archived Feb 2023). Industry body, not an accounting standard setter. Archived copy
- Clouded Judgement 9.25.26 - Own the Interaction Layer. Jamin Ball (Altimeter), 25 Sep 2026. Newsletter; Altimeter trades in public securities; quartiles read from its table.
- Clouded Judgement 6.3.22. Jamin Ball (Altimeter), 3 Jun 2022. Newsletter. Archived copy
- Clouded Judgement 7.5.24 - Declining Growth. Jamin Ball (Altimeter), 5 Jul 2024. Newsletter. Archived copy
- Clouded Judgement 12.8.23 - Q3 Earnings Preview. Jamin Ball (Altimeter), 8 Dec 2023. Newsletter. Archived copy
- The Sword of Damocles in Software. Tomasz Tunguz, 7 Mar 2026. Author's analysis of public filings. Archived copy
- Simpson's Paradox in Measuring Net Dollar Retention Rate. Tomasz Tunguz, 3 Feb 2020. Archived copy
- GRRumbling About Retention Metrics, or, Pitfalls in Measuring SaaS Churn. Randall Lucas (SaaS Capital), 8 Apr 2025. Archived copy
- How public SaaS companies report churn, and what you can learn from them. Christoph Janz (Point Nine Capital), Dec 2017. Archived copy
- State of the Cloud 2023. Bessemer Venture Partners, 11 Apr 2023. Archived copy
- We All Sort of Underestimate the Power of 110%+ NRR. Jason Lemkin (SaaStr), 30 Apr 2021. Archived copy
- High NRR is Magic. It Can Also Hide a Lot of Problems.. Jason Lemkin (SaaStr), 21 May 2026 (updated). Practitioner opinion; no archived copy found.
- What's going on in SaaS: The 2025 SaaS Benchmarks Report. Kyle Poyar (Growth Unhinged), 12 Nov 2025. Newsletter. Archived copy
- Net Revenue Retention Trends. Ray Rike (The SaaS Barometer), 27 May 2024. Newsletter; no archived copy found.
Changes to this page
- · Major · Rewritten and moved here from our earlier article on retention. Benchmarks and filings compiled 2 October 2026.
- · Major · First published as "Net Revenue Retention: Four Settings Behind Every Reported Rate".
Cite this page
Dominique Bouillet, "NRR (net revenue retention): formula, calculator and 2026 benchmarks", CX Cash, updated Oct 6, 2026, https://cxcash.com/metrics/nrr
CX Cash builds software for founders and investors. This page is education, not investment advice. Third-party figures link to their source, and our own arithmetic and illustrative examples are labelled as such.