Two figures published this year both claim to say how fast a B2B SaaS company grows. SaaS Capital’s 2026 edition of its annual survey, drawn from more than 1,000 private B2B SaaS companies, puts the median 2025 growth rate at 22 percent, down from a population median of 25 percent in 2024. ICONIQ’s September 2026 scaling report puts median year-over-year growth at 115 percent for the cohort it calls Pacesetters once they pass 100 million dollars in ARR.
Neither number is wrong. They describe two populations that no longer behave like one market. That is the live problem with SaaS growth benchmarks heading into 2027: the headline figure most operators quote in board decks averages two groups moving at different speeds, and knowing which group you are in changes almost every decision that follows. This is not a story about SaaS shrinking: the fast cohort is growing faster than any private software cohort on record, and the mechanics separating it from the rest are legible. It is a story about benchmark hygiene.
Two true numbers, two different populations
Start with what each sample is, because the samples explain the spread. SaaS Capital’s figure comes from its fifteenth annual survey of more than 1,000 private B2B SaaS companies, reporting calendar 2025 growth; the page states no fielding window. Inside that survey the split is already visible: bootstrapped companies reported median growth of 20 percent, down from 23 percent in 2024, while equity-backed companies reported 25 percent, unchanged. Only 7.3 percent reported flat or negative growth in 2025. Narrow the lens again and the number drops: in a companion post published on 24 April 2026, SaaS Capital put median growth for bootstrapped companies between 3 million and 20 million dollars in ARR at 15 percent, with median net revenue retention of 103 percent.
ICONIQ’s figure comes from a different pool: quarterly operating and financial data from 137 software companies, being ICONIQ’s own venture and growth portfolio companies where data was available plus 11 public companies it selected on IPO performance criteria, through Q2 2026. This is a venture firm reporting on its own holdings, and it says so. Each data point in its charts is one fiscal quarter for one company, so sample counts look larger than the company counts behind them.

A third population makes the point from outside the United States. Monterro’s 2026 Nordic B2B Software Benchmark Report, its fifth edition, surveyed 236 Nordic B2B software companies and reported average revenue growth of 25 percent in 2025, down from 32 percent in 2024, and average net revenue retention of 104.4 percent. Three credible benchmark sets, three answers.
The split is widest in net revenue retention
Growth rates are noisy. Net dollar retention separates the two populations most cleanly, because it strips out new logos and asks one question: is the installed base getting bigger or smaller. ICONIQ’s 2026 cut is stark. Above 100 million dollars in ARR, median net dollar retention outside the Pacesetter group was 99 percent, while the Pacesetter median was 130 percent. Below 100 million dollars the figures were 102 percent and 120 percent. One group holds a base that shrinks without new logos. The other compounds on the base alone.

Treat those Pacesetter cells with care. Four and five company-quarters is a thin base for a median, and this is a vendor reporting on companies it has invested in. The direction shows up in filings that have nothing to do with ICONIQ. Snowflake’s Form 10-Q for the quarter ended 31 July 2026 reports a net revenue retention rate of 126 percent as of that date. Asana’s Form 10-Q for the same quarter end reports a dollar-based net retention rate of 97 percent overall, and 98 percent for customers spending over 100,000 dollars on an annualised basis. Both are well past 100 million dollars in revenue. The two are not computed the same way, Snowflake measuring product revenue over a trailing two-year window and Asana taking a four-quarter average, so read them as two points on a distribution, not a head-to-head.
Monterro found the same shape with a different cut. Companies actively monetising AI reported average net revenue retention of 116 percent, against 104 percent across its full sample of 236. Same market, same year, twelve points apart.
Public markets stopped averaging the two groups
Public investors made this separation explicit first. ICONIQ tracked median enterprise value to next-twelve-month revenue multiples for public enterprise software companies that listed between 2013 and 2025, split by growth rate. In Q2 2026 the 20 percent-plus growers traded at 9.2 times, against 2.7 times for everyone else. The fast group is 14 companies. The slow group is 73.

Averaging across that divide produces a number that describes nobody. The Bessemer Cloud Index, as displayed on 13 October 2026, showed an average revenue multiple of 8.8 times and average revenue growth of 20.6 percent. Useful at market level, poor as a comparison set at either tail.
What the fast cohort is actually doing
The more useful part of ICONIQ’s report is the mechanics behind the gap. Three are concrete enough to test against your own numbers. The first is where new revenue comes from. In 2026, Pacesetters took 74 percent of gross new revenue from new logos below 100 million dollars in ARR and 71 percent above it, against 54 percent and 46 percent for other companies in the same bands. They expand better and still source most of their growth from accounts they did not have last year.
The second is pricing structure. ICONIQ reports that 70 percent of the Pacesetter group uses hybrid or consumption-based pricing, 60 percent are AI-native, and its median age is six years from founding. The buy side is moving the same way: in Ramp’s Summer 2026 Business Spending Report, built from over 70,000 businesses using Ramp Bill Pay and corporate cards, usage-billed products account for 36 percent of embedded AI spend, up from 24 percent a year earlier. That is Ramp’s own customer telemetry, not an independent survey.
The third is spend. Pacesetters below 100 million dollars in ARR spent 284 percent of revenue on operating expenses in 2026, against 124 percent for others in that band. Most operators skip this when they copy the playbook. The fast cohort buys its growth with capital most companies do not have, which is why capital efficiency metrics belong next to the growth number, not underneath it. Discipline arrives later: Pacesetters above 100 million dollars in ARR ran a burn multiple below 1 times in 2026.
Headcount follows the same sorting. ICONIQ’s median year-over-year headcount change in 2026 was 146 percent for companies growing 100 percent or more, and 2 percent for those growing under 25 percent, across 76 company-quarters. Two different assumptions about next year, which is why revenue per employee is the more comparable operating metric.
Your buyers are sorting too
The split is not confined to vendors. It shows up in how much customers spend, and is sharper there. Ramp’s Summer 2026 report found that the top 1 percent of AI adopters on its platform spend more than 7,400 dollars on AI per employee per month, the top 10 percent spend 650 dollars, and the median adopter spends 11.95 dollars. Again, that is Ramp’s own card and bill-pay telemetry, not an independent benchmark. Model expansion revenue off a median customer and you are modelling the wrong customer by three orders of magnitude.
Adoption splits by company size in a way that would surprise most enterprise-first go-to-market plans. The Ramp AI Index as published on FRED, last updated 29 September 2026, put the share of businesses with an observed AI payment in August 2026 at 66.8 percent for medium-sized businesses, 62.5 percent for small businesses and 59.0 percent for large businesses. Ramp defines small as one to 24 employees, medium as 25 to 99, large as 100 or more. The largest buyers are the slowest adopters here, the opposite of how most SaaS pipeline models are built. The index covers only Ramp’s own customers and only observed payments, so free tools and personal-card spend are invisible to it.
The trend holds even if the level is specific to Ramp’s base. Axios reported in January 2026, using the same index and a sample it put at roughly 50,000 companies, that 47 percent of US companies paid for AI subscriptions that month, up from 26 percent a year earlier.
Buying behaviour is shifting at the same time. ICONIQ’s State of GTM in 2026, built from a January 2026 survey of go-to-market executives at more than 150 B2B software companies, reported that free trial or proof-of-concept to paid conversion improved 14 percentage points year over year, against 3 points for new lead to MQL, 3 points for MQL to SQL, 3 points for SQL to closed won and 2 points for demo to closed won. The 2025 sample was 167 respondents, the 2026 sample 147.
Picking the right comparison set
None of this makes benchmarking useless. It makes the choice of comparison set the first decision rather than an afterthought. Four rules follow. Match the funding structure first. SaaS Capital separates bootstrapped from equity-backed and finds a five-point median gap, and the bootstrapped median falls further inside the 3 million to 20 million dollar ARR band. A bootstrapped company held to a venture-backed median gets a target its balance sheet cannot fund.
Check whether the benchmark is a vendor’s own customers. ICONIQ reports on its portfolio. Ramp reports on businesses that use Ramp. Both are useful, both disclosed, neither a market census. Jason Lemkin’s 18 September 2026 write-up of the ICONIQ index reproduces the same cells, including median revenue per employee of 655,000 dollars and a top quartile of 890,000 dollars above 100 million dollars ARR.
Read the n, not the headline. A 130 percent median resting on four company-quarters is a signal, not a benchmark. The 99 percent beside it rests on 32 company-quarters.
Pair every growth comparison with a margin comparison. The fast cohort’s growth comes with 284 percent operating expense ratios at small scale. Take one half of that trade and not the other and Rule of 40 planning goes wrong.
Frequently asked questions
What is a good SaaS growth rate in 2026?
It depends on which population you belong to. SaaS Capital’s 2026 survey of more than 1,000 private B2B SaaS companies puts the median 2025 growth rate at 22 percent, bootstrapped companies at 20 percent and equity-backed at 25 percent. Monterro’s survey of 236 Nordic B2B software companies reported average 2025 growth of 25 percent. ICONIQ’s Pacesetter group, measured across its own portfolio plus 11 selected public companies, showed a median of 115 percent above 100 million dollars in ARR. Match the sample to your funding, stage and geography first.
Why do SaaS growth benchmarks vary so much between reports?
Because the samples differ in funding mix, size band, geography, period and who collected them. A survey of bootstrapped companies between 3 million and 20 million dollars in ARR produced a 15 percent median in the same year a venture portfolio produced 115 percent at scale. Some publish medians, others averages. Some count companies, others company-quarters. Check that both measure the same statistic on the same companies over the same window.
Is net revenue retention above 100 percent still normal?
At scale it is no longer automatic. ICONIQ’s 2026 data shows median net dollar retention of 99 percent for companies above 100 million dollars in ARR outside its Pacesetter group, against 130 percent inside it. Public filings show a similar spread: Snowflake reported 126 percent as of 31 July 2026, Asana 97 percent overall for the same quarter end using a different method. Treat 100 percent as a threshold to defend, not a floor you inherit.
Should I benchmark against public SaaS multiples or private survey data?
Use private survey data for operating targets and public multiples for valuation context, and do not blend them. Public enterprise software companies growing above 20 percent traded at a median of 9.2 times forward revenue in Q2 2026, against 2.7 times below that line, so a market average hides the distinction being priced. Private surveys sit closer to your operating reality, but neither tells you what a buyer will pay.
How do I know if my company is in the fast cohort?
Three tests from the 2026 data are easy to run. First, the share of gross new revenue from new logos: the fast cohort sat at 71 to 74 percent, the rest at 46 to 54 percent. Second, the share of revenue tied to consumption or hybrid pricing: 70 percent of the fast cohort uses one or the other. Third, your operating expense ratio, because the fast cohort below 100 million dollars in ARR spent 284 percent of revenue to hold that rate. If the first two look right and the third does not, you have the motion without the funding.
The short version
The median SaaS company is still a real thing. It is no longer a useful description of the market it sits in, because the distribution has pulled apart far enough that the centre describes neither tail. Treat SaaS growth benchmarks as a single number and you will keep setting targets drawn from a population you are not in. Name your cohort first, check the sample behind every figure, and read growth next to the spend that bought it.







