What Is Net Revenue Retention (NRR)? Formula, 2026 Benchmarks, and What Good Looks Like
NRR is existing-customer ARR after expansion, contraction, and churn. Benchmarkit's 2026 private median is 102%; top quartile is 110%. Usage-based prints 108%, seat-based 98%.
Net revenue retention (NRR) — also called net dollar retention (NDR) — measures how much recurring revenue you keep from the customers you already had at the start of a period, after expansion is added and contraction and churn are subtracted. New logos never enter the formula. An NRR above 100% means the installed base grew without a single new customer. Below 100% means you must buy new logos just to stand still.
Growth Centr publishes evergreen, research-backed analysis on SaaS unit economics for founders, marketers, and operators who need a board number they can decompose, not a blended average that hides the pricing model they actually sell on.
Benchmarkit's 2026 SaaS and AI-Native Metrics (CY-2025 actuals, 342 B2B SaaS and AI-native companies; NRR n = 230) puts the private-company median at 102%, just above the break-even line, with the top quartile at 110%. Usage-based books print 108%; seat-based print 98%. Aleph's write-up of the same book is the clean public formula: NRR = (Starting ARR + Expansion − Contraction − Churn) ÷ Starting ARR. Public SaaS is a different clock: Blossom Street Ventures' Q1 2026 screen of post-2017 IPOs that still disclose the metric puts median NDR at 110%, stable for six quarters after falling from 123%+ in H1 2022.
Key Takeaways
- NRR = (starting ARR + expansion − contraction − churn) ÷ starting ARR. New logos are excluded. Above 100% means the existing base grew on its own.
- Benchmarkit CY-2025: private B2B SaaS median 102%, 25th percentile 92%, top quartile 110%. GRR sits at 84% — expansion is covering a 16-point hole.
- Pricing is the structural split: usage-based median 108% (75th percentile 155%); seat-based 98%, below break-even. The $25K–$50K ACV band leads at 105%.
- Public SaaS median NDR is 110% (Blossom Street, Q1 2026, 27 disclosers). That is not a rebuttal of 102%. It is what the number looks like after the IPO filter.
- 100% is the floor, 110%+ is top quartile on the private book, 120%+ is best-in-class. Sub-$5M ARR median is 94% — 110% is a scale target, not a seed-stage score.
How to Calculate NRR
NRR = (Starting ARR + Expansion ARR − Contraction ARR − Churned ARR) ÷ Starting ARR × 100
Lock a cohort on day one. Track only those accounts for twelve months. Everything that happens to them — seats added, usage grown, modules sold, price increases, downgrades, cancellations — goes in. Everything that happens with a customer who was not there at the start stays out.
Starting ARR. Recurring revenue from customers who were paying at the beginning of the period. Same definition of “recurring” you use for ARR everywhere else. Professional services, one-time implementation, and usage that is not contracted do not belong in the denominator.
Expansion. Upsell, cross-sell, seat growth, usage growth, and price increases from that same cohort. Reactivations of customers who were in the starting set and left, then came back, are expansion (or a separate reactivation line you add back). They are not new logos.
Contraction. Downgrades and usage declines from customers who stayed. This is the silent leak. Seat-based companies seeing AI-driven headcount cuts feel it here before they feel it in logo churn.
Churned ARR. Recurring revenue from customers who cancelled entirely. Logo churn and revenue churn are not interchangeable — a $5K SMB cancel and a $200K enterprise cancel move NRR very differently. Pair this with SaaS churn rate benchmarks rather than treating NRR as a substitute for logo math.
What never enters. New-logo ARR acquired during the period. That is the most common way a finance deck “beats” 110%. If new business leaks into expansion, NRR inflates and a retention problem hides behind a growth print.
Worked example. Starting ARR $10M. Expansion $1.8M. Contraction $0.6M. Churned $0.9M. NRR = (10 + 1.8 − 0.6 − 0.9) ÷ 10 = 103%. The installed base grew 3% with zero new customers. GRR on the same numbers is (10 − 0.6 − 0.9) ÷ 10 = 85%. Expansion papered over 15 points of leak.
NRR vs GRR vs Logo Churn
Three metrics, three jobs. Report them together or expansion will hide a hole.
| Metric | Formula idea | Can it exceed 100%? | 2026 private median | What it answers |
|---|---|---|---|---|
| Logo (customer) churn | Accounts lost ÷ starting accounts | No | ChartMogul monthly: 6.5% below $300K ARR, 3.1% at $8–15M | Are we losing customers? |
| Gross revenue retention (GRR) | (Starting − contraction − churn) ÷ starting | No. Caps at 100%. | Benchmarkit 84% (from 88%) | How much of the original dollars did we keep? |
| Net revenue retention (NRR) | (Starting + expansion − contraction − churn) ÷ starting | Yes | Benchmarkit 102% | Did the existing base grow or shrink? |
The gap between GRR and NRR is the expansion engine. A 102% NRR on an 84% GRR means expansion is doing 18 points of work to cover churn and downgrades. That is a growth strategy sitting on a retention problem — fine if you see both numbers, fatal if the board only sees NRR.
ChartMogul's customer-churn book (2,500+ private SaaS businesses on a billing platform) makes the logo link explicit: companies with NRR ≥100% run a median 1.7% monthly logo churn. Companies with NRR below 60% run 7.3% — more than double. High NRR is not a different metric from churn. It is what churn looks like after expansion has had its say.
ChartMogul's billing population is earlier-stage and more self-serve than Benchmarkit's finance-survey book, so the absolute NRR medians are not interchangeable. ChartMogul's 2025 AI retention study (3,500 companies scraped and classified; annualized rates for businesses at ≥$250K ARR) puts median NRR at 82% for B2B SaaS, 49% for B2C, and 48% for AI-native products — with AI-native GRR at 40%. That is a different sample and a different clock from Benchmarkit's 102%. Both are real. Neither is your number until you match population, period, and formula.
2026 Benchmarks: Private vs Public, and Why They Disagree
Two clocks, two populations.
Private B2B (Benchmarkit / Aleph, June 2026, CY-2025, n = 230). Median NRR 102%. 25th percentile 92% (from 95% the year before — worse than the 2023 trough). Top quartile 110%. The 75th percentile eased from 110–111% in CY-22/23 to about 108%. Expansion is harder to come by: buyer budget scrutiny, slower upsell cycles, and deferred commitments as buyers evaluate AI-native alternatives in parallel.
By growth rate, same book:
| 2025 growth rate | Median NRR | What it is telling you |
|---|---|---|
| >50% | 111% (75th percentile 155%) | Fast growers are not new-logo machines. Product adoption and a post-sale motion are doing real work. |
| Population median | 102% | Above break-even, thin margin for error. |
| <10% | 92% | A shrinking installed base compounds the new-logo problem. You are buying growth at the expensive end of CAC. |
By ARR and ACV:
| Cut | Median NRR | Note |
|---|---|---|
| >$100M ARR | 103% (75th 115%) | Scale: more product surface, stronger CS, multi-product cross-sell. |
| $20M–$50M ARR | 101% | Expansion infrastructure is being built; it has not fully paid off. |
| <$5M ARR | 94% | Too little installed-base depth and CS capacity for systematic expansion. |
| $25K–$50K ACV | 105% | Highest band in the book. Annual expansion reviews, multi-product motion. |
| $10K–$25K ACV | Below 100% | First year this band dropped under the line. |
| Sub-$5K ACV | 98% | SMB expands slowly and churns more. |
By pricing model. This is the structural finding, not a rounding error. Usage-based median 108%, 75th percentile 155%. Seat-based median 98%. Hybrid (fixed fee plus usage) sits in between with a revenue floor plus consumption upside. Every year the usage-based cohort compounds its installed base roughly 10 points faster than the seat-based cohort. Seat count is under pressure as customers use AI to do more with fewer people. That is the same pricing split that shows up in SaaS pricing models and in the Rule of 40 — hybrid subscription-plus-usage leads the 75th percentile there at 43%.
Public SaaS (Blossom Street Ventures, Q1 2026). Every SaaS company that IPO'd since October 2017, still public and not distressed: 44 names, 27 that disclosed NDR in the quarter. Median 110%, average 110%. Down from 123%+ in H1 2022. Stable near 109% for six quarters. Q4 seasonality — the old pattern of a Q4 drop because renewals cluster then — has not shown up in 2023, 2024, or 2025. Public books run older, more annual, and more enterprise than a 230-company private survey. They are what NRR looks like after the $20M–IPO filter has already been applied, not a target for a $6M ARR seat-based SMB.
AI-native on a billing platform (ChartMogul, 2025). Median NRR 48%, GRR 40%, improving from 27% GRR in January to 40% by September as early “tourists” left. Price is the cut that actually predicts durability: AI products above $250/month print 85% NRR (essentially B2B SaaS); $50–$249 print 61%; below $50 print 32%. Annual plans run 10–20 points higher NRR than monthly. Treat experimental AI spend and production spend as different cohorts or you will average a wrapper with a workflow product and fire the wrong team.
Why NRR Moves Valuation — And Why Expansion Is the Cheap Lever
NRR is the compounding term in the SaaS growth equation. A $10M ARR book at a steady 120% NRR and zero new logos is ~$12.0M, $14.4M, $17.3M, $20.7M, $24.9M over five years. The same book at 92% NRR is $9.2M, $8.5M, $7.8M, $7.2M, $6.6M. You do not outrun that with a demand-capture hire.
It is also the cheap dollar. Benchmarkit median expansion CAC is $0.80 of S&M per $1 of new ARR versus $1.63 for a new logo — 53% cheaper. Expansion already supplies 40% of net-new ARR at the median, 59%+ at the top quartile, 62% above $100M ARR. Read that next to B2B SaaS CAC benchmarks: blended CAC ratio $1.30, 16-month payback. A 92% NRR is a CAC problem wearing a retention badge. You will pay the new-name tax twice.
A product-led motion helps NRR only if activation turns into expansion. Benchmarkit's PLG cohort shows a 36% median expansion share of new ARR — below hybrid (54%) — which is the tell that many PLG books have not built the commercial overlay or in-app upsell that converts usage into dollars. Cheap self-serve growth with ugly GRR raises this year's new-logo print and next year's NRR miss.
What a Good NRR Is in 2026
100% is still the floor. 110% is still roughly top quartile on the private book. 120%+ is still best-in-class. Plan to the row you are actually in.
| Your situation | A "good" 2026 print | What to manage instead if you miss |
|---|---|---|
| <$5M ARR, still finding the ICP | Trajectory toward 100%. Median peers sit at 94%. | Logo churn, onboarding, GRR. 110% is a scale number. |
| $20–50M ARR, mixed growth | 100–105% and rising. Median peers ~101%. | Split GRR vs expansion. If GRR is 80% and NRR is 102%, you have an expansion story sitting on a leak. |
| $50M+ ARR, raising or planning a process | 110%+ on a labeled twelve-month cohort, same definition as the buyer. | Public comps sit near 110% NDR. Do not switch from logo-weighted to revenue-weighted mid-process. |
| >50% growth, any scale | Median 111%. 155% at the 75th is usage-plus-expansion, not a slogan. | Keep new-logo quality from poisoning GRR. Fast growth with 7% monthly logo churn is borrowed. |
| <10% growth | Median 92%. You will not grow out of this with more S&M. | Expansion quota on the existing base, then pricing architecture. |
| Usage-based or hybrid | Median 108%; 75th 155%. 100% is underperformance for this model. | Draw the line between new ARR and expansion ARR cleanly or the number is fiction. |
| Seat-based | Median 98%. Holding 100% is outperforming the cohort. | Packaging, workflow depth, and a path off pure seats — not an NRR OKR on the same price card. |
| AI-native, <$50/month | ChartMogul median 32% NRR. Do not benchmark against B2B SaaS 102%. | Move upmarket or treat the book as trials, not ARR. >$250/month AI is the 85% NRR row. |
A 94% is not a crisis at $3M ARR. A 94% is a problem at $80M ARR with a 2027 IPO story. Same number, different job.
A 30-Day Operator Test
Do not set “hit 110% this year” as an OKR until the cohort matches the book.
Days 1–7 — Lock the definition. Write starting ARR, what counts as expansion vs new ARR (especially under usage-based pricing), the period (trailing twelve months, not a single strong quarter annualized), and whether reactivations sit in expansion. Recompute the last four quarters. If finance and CS cannot produce the same NRR in one sitting, stop targeting and reconcile.
Days 8–14 — Split NRR into GRR plus expansion. Put last year's GRR and the expansion add-back in two columns. A 102 made of 96 GRR + 6 expansion is a different company from 84 GRR + 18 expansion. The first needs more packaging and usage. The second needs to stop the leak — onboarding, ICP, involuntary churn — before you hire another expansion AE.
Days 15–21 — Benchmark the right row. Use Benchmarkit if you are private B2B. Use a public NDR screen if you are modeling a listing. Use ChartMogul if you are early, monthly, and on a billing platform. Do not average them. Put ARR, ACV, pricing model, and growth band next to the number. If you are under ~$5M ARR, park 110% and put GRR and logo churn on the slide instead.
Days 22–30 — Move one lever, then re-read. One of: put expansion quota on the existing base (the $0.80 CAC dollar); stop selling seats into accounts that are about to compress headcount; or shift a slice of the book from monthly to annual (ChartMogul: 10–20 point NRR lift). Recompute on the same cohort definition. If NRR rose because you started counting new logos as expansion, you did not improve the business — you changed the formula. If NRR rose because expansion lifted the numerator at a stable GRR, keep going.
Methodology
This is a 2026 planning brief, not a survey we ran. The private-company NRR median (102%), 25th percentile (92%), top quartile (110%), ARR and ACV splits, usage-versus-seat gap (108% / 98%), expansion CAC ($0.80 vs $1.63 new-name), expansion share of net-new ARR (40%), and GRR (84%) are from Benchmarkit's 2026 B2B SaaS and AI-Native Performance Benchmarks (1 June 2026; 342 companies; NRR n = 230; CY-2025 actuals), cross-checked against Aleph's public NRR write-up of the same book. Logo-churn-by-NRR-band (1.7% monthly at ≥100% NRR vs 7.3% below 60%) is from ChartMogul's customer-churn benchmarks (2,500+ private SaaS businesses). AI-native vs B2B vs B2C NRR (48% / 82% / 49%), the $250/month price cut (85% / 61% / 32%), and the 10–20 point annual-vs-monthly lift are from ChartMogul's 2025 AI retention study (3,500 companies; ≥$250K ARR filter). Public median NDR (110%, 27 of 44 post-2017 IPOs that disclosed) is from Blossom Street Ventures, Q1 2026. No statistic appears here unless it was on a page we fetched. Private finance-survey medians, billing-platform medians, and public-company NDR are not interchangeable.
Read Next
- SaaS Churn Rate Benchmarks 2026: What's a Good Rate by Segment
- B2B SaaS CAC Benchmarks 2026: What It Costs to Acquire a Customer
- What Is the Rule of 40? Formula, 2026 Benchmarks, and When It Matters
- SaaS Pricing Models Guide with AI Optimization Tactics
- What Is Product-Led Growth (PLG)?
FAQs
1. What is net revenue retention (NRR)?
It is the recurring revenue you keep from customers who were already paying at the start of a period, after adding expansion and subtracting contraction and churn. New logos are excluded. Above 100% means the existing base grew without any new customers.
2. What is a good NRR in 2026?
100% is the floor, 110%+ is top quartile on Benchmarkit's private book, and 120%+ is best-in-class. The 2025 private median is 102%. Public SaaS that disclose NDR sit at a 110% median. Sub-$5M ARR companies print a 94% median — 110% is a scale target.
3. How do you calculate NRR?
(Starting ARR + expansion − contraction − churned ARR) ÷ starting ARR × 100, on a fixed cohort, usually trailing twelve months. Do not let new-logo revenue leak into expansion, especially under usage-based pricing.
4. What is the difference between NRR and GRR?
GRR excludes expansion and cannot exceed 100%. NRR includes expansion and can. Benchmarkit 2025: GRR 84%, NRR 102%. The 18-point gap is your expansion engine — and the size of the leak expansion is covering.
5. Why is usage-based NRR higher than seat-based?
Usage expands automatically as customers consume more, without a renegotiation. Benchmarkit: 108% median (155% at the 75th) versus 98% for seats. Seat counts are under pressure as buyers use AI to do more with fewer people. Holding 100% on seats is outperforming that cohort; printing 100% on usage is underperforming it.
Disclaimer: This content is provided for informational purposes only and does not constitute financial, investment, or operating advice. NRR figures reflect publicly reported research as of August 2026, from studies with different sample frames (private finance survey vs billing-platform vs public filings), period definitions, and ARR filters. Benchmarkit numbers are finance-survey medians on CY-2025 actuals (n = 230 for NRR). ChartMogul numbers are billing-platform aggregates. Blossom Street numbers are public-company disclosures. Treat them as directional peer checks, not board targets without your own cohort data.
