# SaaS Growth Benchmarks: Match the Cohort First

> A benchmark helps only after you match the metric, cohort, period, and commercial profile.

- Author: Rishikesh Ranjan · Published: Aug 8, 2026 · Updated: Aug 29, 2026
- Type: Playbook
- Tags: Metrics, Frameworks, GTM
- Growth levers: Revenue (primary), also Acquisition, Retention
- ~2349 words

---

The first question behind a SaaS benchmark is not, "What is good?" It is, "Am I comparing the same thing?" A number can look precise and still be useless when its population, period, formula, or commercial context does not match yours.

A benchmark is a **comparison protocol**: metric definition, cohort, period, commercial profile, and statistic. It isn't a universal target. An all-SaaS median can orient you, but it can't tell a bootstrapped product what to fix next.

Scorecards make incompatible numbers look interchangeable. This memo keeps only source-scoped facts that survive a direct read of the underlying reports. When a narrow, comparable cut is unavailable, the honest answer is unavailable, then an internal cohort trend, not a neat range somebody made up.

> **The rule before any benchmark:** Don't ask whether you beat a SaaS median until you can state the numerator and denominator, period, cohort, ARR (annual recurring subscription revenue, a subscription run-rate metric rather than total recognized revenue) or ACV/ARPA, motion, funding, source date, and action you will take if the comparison is weak. If one of those fields is missing, the number is context at best.

## Choose the metric from the decision

Most benchmark pages begin with a giant matrix. That is backwards. Start with the operating decision in front of you. If you are deciding whether to spend more on acquisition, CAC payback belongs in the conversation. If the base is leaking, put gross revenue retention and net revenue retention next to each other. If a launch page is underperforming, define its conversion event before you compare a percentage.

The table below is the comparison header I would put in a monthly operating review. It's deliberately less sexy than a dashboard full of green and red cells. It does the work: it makes a bad comparison obvious before it turns into a bad target.

| Decision | Record before you compare | Reject the benchmark when |
| --- | --- | --- |
| Set a growth target | Growth definition, period, ARR, funding, geography, and GTM motion | It says only "SaaS growth" and names neither the population nor the period |
| Diagnose retention | NRR, GRR, or logo retention; starting cohort; period; expansion treatment | It calls revenue churn, customer churn, GRR, and NRR the same thing |
| Fund acquisition | CAC-payback formula, gross-margin policy, ACV, company type, and attribution treatment | It offers a month target without saying what sits in the numerator or denominator |
| Improve a landing page | Event, traffic source, configured goal, page type, and time window | It relabels a landing-page action as product conversion or customer conversion |
*Use this as an operating-review header. It is an original explanatory framework, not a source-derived benchmark table.*

Fill the header for one live decision, not every number in the dashboard. Put your actual result next to a candidate source and write down why the two belong together. That forces the useful question: if the comparison is weak, are we going to change spend, investigate a cohort, change a pricing assumption, or leave the target alone? A benchmark without a possible action is just reporting furniture.

1. **Name the decision: **write the next choice in plain language. "Fund paid acquisition next period" needs a different comparison from "find out why renewal revenue fell." The decision tells you whether growth, payback, retention, or conversion deserves attention.
2. **Freeze the calculation: **copy the numerator, denominator, included events, excluded events, and time window. A metric label is not a definition. "Churn" can mean customers lost, gross revenue lost, or net revenue lost after expansion. Those produce different diagnoses.
3. **Select one peer set: **match ARR or ACV/ARPA where the source permits it, then check funding, geography, commercial motion, and period. Record the mismatch you cannot remove. A close source with one visible mismatch is more useful than a generic source with none of its scope written down.
4. **Write the failure action: **state what you will inspect or change if the gap persists. For a retention gap, that may mean breaking out expansion from gross retention. For a payback gap, it may mean checking whether sales spend and new-customer ARR were counted on the same basis.

The comparison can fail in a good way. You may learn that no public cohort matches your product, which means you should not force a target this quarter. Mark the external comparator unavailable, keep the definition fixed, and compare your own cohorts over time. That isn't giving up on measurement. It's refusing to make a decision look evidence-backed when the evidence does not travel.

Keep the source date in the header too. A report can be well designed and still describe a different buying climate, pricing mix, or funding environment from the one you are operating in now. The date doesn't make an older number worthless. It tells you how much confidence to place in it, and whether the next job is finding a fresher cohort rather than changing the plan.

There is one practical consequence: a metric can be urgent without having a public benchmark. Activation, onboarding completion, feature adoption, PQL rate, and time to value often depend on a product-specific event definition. If your team hasn't agreed on that event, looking for an industry median is a distraction. Define the event, track cohorts for a few periods, then look for a source whose event really matches.

## Why 22% growth can mean three different things

SaaS Capital reports a 22% median growth rate in 2025 across more than 1,000 private B2B SaaS companies. Inside that same private-B2B survey, bootstrapped companies report 20% median growth and equity-backed companies report 25%. The useful part is not picking the highest number. It's recording the funding cut before you turn 22% into a board target.

Now put that beside two other reported figures. Benchmarkit's CY2024 growth cut has N=149 companies and a 26% median, with 30% for VC-backed companies and 13% for PE-backed companies. HSBC reports 22% median 2025 ARR growth for 50 UK Series A+ enterprise-software companies. The matching 22% labels are not evidence of one shared answer. They are evidence that labels travel farther than cohorts do.

| Source and period | Population and cut | Reported growth | What it cannot tell you |
| --- | --- | --- | --- |
| SaaS Capital, 2025 | 1,000+ private B2B SaaS companies; overall, bootstrapped, and equity-backed cuts | Median growth: 22% overall, 20% bootstrapped, 25% equity-backed | A target for public SaaS, consumer apps, or a company whose funding and motion do not match |
| Benchmarkit, CY2024 | N=149 growth cut; 583 participants overall; VC-backed and PE-backed cuts | Median growth: 26% overall, 30% VC-backed, 13% PE-backed | Numeric confirmation of the SaaS Capital result: year, sample, and funding labels differ |
| HSBC Innovation Banking, 2025 | 50 UK Series A+ enterprise-software companies tracked from 2023 to 2025 | Median ARR growth: 22% | A self-serve, PLG, or all-SaaS target |
*Different cohorts, not competing estimates of one SaaS median.*

Each row has a limitation that matters. SaaS Capital's lender-sponsored survey is self-reported, and its full sampling and weighting are not public. Benchmarkit's 2024 figures are independent context, not corroboration of SaaS Capital's 2025 numbers. HSBC is a small UK enterprise subset. The right move is to choose one primary peer set, note a second one only as context, and explain why the first set wins.

For an illustrative case, take a bootstrapped B2B product at $3M ARR, 20% year-over-year growth, and $18K ACV. The 20% SaaS Capital cut is a reasonable orientation for its funding status, not a verdict. It still lacks an ARR band, a geography match, and a motion match. A UK enterprise 22% figure can stay in the notes, but it shouldn't repaint the primary comparison green.

> **Steal this:** A number that matches your headline but not your cohort is not a benchmark. It is trivia with good typography. Pick the peer set first, then decide whether the gap deserves intervention.

## Pricing, ARPA, and billing are comparison fields

Raw dollar metrics travel badly. For example, a $30 monthly tool and a $30K ACV enterprise product can have wildly different CAC, sales effort, payback, and renewal behaviour while each is healthy for its own model. ARPA is average monthly revenue per account. ACV is the annual value of a contract. They overlap sometimes, but they aren't synonyms, and neither one is a shortcut for enterprise status.

[Benchmarkit's CY2024 N=149 growth cut](https://www.hibob.com/wp-content/uploads/2025-SaaS-Performance-Metrics-Benchmarks.pdf?utm_source=productgrowth.blog) reported 44% median growth for primarily usage-based pricing respondents and 25% for traditional subscription pricing respondents. Its report also warns that a growing AI-native mix can bias the comparison. That is a report about two pricing cohorts in one year. It isn't proof that PLG caused higher growth, or that a self-serve motion will be cheap for your business.

Put pricing and billing in the header because they can move with customer type and product value. Don't promote them into a causal story just because a report has a clean split. The more a conclusion depends on a hidden mix of segment, channel, and company stage, the more you should treat it as a hypothesis to test in your own cohorts.

When a source forces you to keep a caveat in the sentence, keep it. The caveat isn't legal padding. It tells you whether the action belongs in product, pricing, sales, or the measurement model. A metric that cannot separate those jobs is too blunt for an operating decision, even when it arrives in a polished annual report.

## Retention needs two numbers, not one

Net revenue retention and gross revenue retention sound close enough to collapse into one status light. Don't do that. NRR asks how much monthly recurring revenue (MRR) the starting cohort kept after expansion, contraction, and churn. GRR removes expansion from the calculation. A high NRR can coexist with a weak underlying base if upsells are covering losses.

| Retention measure | Formula | What it answers |
| --- | --- | --- |
| NRR | (Starting MRR + expansion MRR - contraction MRR - churn MRR) / starting MRR | Did the starting revenue cohort hold and expand? |
| GRR | (Starting MRR - contraction MRR - churn MRR) / starting MRR | How much starting revenue remained before any upsell? |
*ChartMogul documents year-over-year benchmark intervals; Benchmarkit uses a cohort and ARR framing. Period, cancellations, reactivations, usage revenue, and currency policies can still differ.*

Here is an illustrative diagnosis, not benchmark data. Start with $100 of cohort MRR. If $10 contracts or churns, GRR is 90%. If $20 expands, NRR is 110%. The revenue headline looks healthy, but the base still lost 10% before expansion. Compare that with 98% GRR and 110% NRR. Same NRR, very different product and customer-success question.

ChartMogul's H1 2024 platform analysis covers more than 2,500 SaaS businesses and excludes companies below $300K ARR from ARPA cuts. In its $500+ ARPA segment, only the top quartile reached at least 100% NRR. That scope matters: ARPA is not ACV, and $500+ ARPA does not mean enterprise. Treat it as a source-specific retention comparison, not a 100% target for every SaaS company.

| Source, period, and population | Reported result | Decision boundary |
| --- | --- | --- |
| ChartMogul, H1 2024; 2,500+ platform businesses; $500+ ARPA; companies below $300K ARR excluded from ARPA cuts | Only the top quartile reached at least 100% NRR | Use only when your ARPA and measurement policy are comparable. Do not relabel this as an all-SaaS target. |
*A source-scoped NRR observation, not a universal threshold.*

Billing is another field, not an answer. In ChartMogul's full-year 2024 mixed-billing cohort, companies at $250-500 ARPA had median NRR of 88% for annual-plan revenue and 76% for monthly-plan revenue. The analysis excludes companies below $300K ARR from ARPA cuts and excludes companies that rely entirely on one billing model. It reports an association. It doesn't prove that switching to annual billing caused a 12-point retention lift.

| Reported cohort | Reported NRR | What not to infer |
| --- | --- | --- |
| ChartMogul full-year 2024 mixed-billing companies at $250-500 ARPA; ARPA cuts exclude < $300K ARR; single-model companies excluded | 88% annual-plan revenue; 76% monthly-plan revenue | Annual billing caused the gap. Customer type, value, geography, discounts, and plan mix can all affect it. |
*An observational association with stated exclusions, not a causal annual-billing chart.*

## CAC payback is a formula before it is a number

CAC payback answers how many gross-margin-adjusted months it takes to recover sales and marketing spend from new-customer revenue. The formula sounds universal until you inspect what counts as new revenue. Benchmarkit's 2025 report says private-company payback usually compares sales and marketing expense with new-customer ARR, while public-company treatment can use net-new implied ARR that includes churn, downsells, and expansion. Those aren't interchangeable denominators.

| Question | Benchmarkit 2025 definition or warning | What must match before comparison |
| --- | --- | --- |
| Formula | ((Sales and marketing expense) / (ARR from new customers x gross subscription margin)) x 12 | Sales and marketing numerator, new-customer ARR denominator, and gross-margin policy |
| Company treatment | Private payback uses new-customer ARR; public payback can use net-new implied ARR that includes churn, downsells, and expansion | Company type and revenue treatment |
| Comparator | The report's payback chart has N=148 and says payback is highly correlated with ACV | ACV alongside formula and attribution policy |
*There is no portable month target until the formula, ACV, and company treatment align.*

An illustrative calculation makes the boundary obvious. Suppose the $3M bootstrapped company spends $180,000 on sales and marketing, adds $300,000 of new-customer ARR, and has an 80% gross subscription margin. Its calculated payback is nine months: ($180,000 / ($300,000 x 0.80)) x 12. That is a useful internal result. It becomes a valid external comparison only after the peer source uses the same revenue treatment, margin policy, ACV context, and company type.

This is where broad payback badges do real damage. A shorter period can look efficient because a team counts only new ARR, while another source looks slower because it nets churn and expansion into the denominator. Neither team needs a motivational label. They need the formula in the header and the cash implication behind it.

## Conversion starts with the event, not the rate

Unbounce reports a 3.8% median conversion rate for SaaS landing pages. Its dataset covers July 23, 2023 to July 23, 2024: more than 464M unique visitors, 57M conversions, and more than 41K landing pages. That is a large dataset. It's still not a generic SaaS funnel number.

| Reported number | Period and dataset | Configured page goal can be | Do not use it for |
| --- | --- | --- | --- |
| 3.8% median SaaS landing-page conversion | 2023-07-23 to 2024-07-23; 464M unique visitors; 57M conversions; 41K+ landing pages | A demo, free-trial signup, gated download, webinar signup, or trial-user payment | A visitor-to-signup, visitor-to-customer, free-to-paid, or product-funnel benchmark without an event match |
*Unbounce's methodology is same-owner method support, not independent numerical corroboration. Its landing-page event scope is the point.*

If your question is whether a 30-day trial converts to paid, 3.8% can't answer it. If your page's configured goal is a demo request, it may be useful orientation after you match the page type and traffic mix. The numerator and denominator deserve as much attention as the percentage, because they tell you which team can act on the result.

Mobile-app data is a useful exclusion test. [RevenueCat's 2026 report](https://www.revenuecat.com/state-of-subscription-apps?utm_source=productgrowth.blog) covers more than 115,000 subscription apps and $16B in revenue, with 2025 as its main metric period. It defines D35 download-to-paid as an install that produces at least one paid subscription within 35 days. That is a clean event for subscription apps. It isn't the same event as a B2B landing-page action, a trial start, or a sales-assisted customer. Keep it out of a B2B SaaS conversion table.

> **Unavailable is a valid result:** When you can't find a source that matches your event, cohort, and commercial profile, mark the external comparison unavailable. Track your own cohort over time and write down what would make you replace that status. An unsupported range doesn't become useful because it appears in a prettier table.

## Questions before you set a SaaS target

#### What is a good SaaS growth rate at my ARR?

There is no defensible answer from a blended SaaS median alone. SaaS Capital's 2025 private-B2B survey reports 22% median growth overall, 20% for bootstrapped companies, and 25% for equity-backed companies across more than 1,000 companies, but it doesn't create an ARR-specific target for every firm. Record your ARR, funding, geography, motion, period, and growth definition, then use the closest published cut as orientation. If that cut is missing, use your internal cohort trend instead of inventing a range.

#### Is 100% NRR a healthy target?

100% NRR means expansion offsets contraction and churn for the starting revenue cohort. It doesn't reveal gross retention, because GRR excludes expansion. In ChartMogul's H1 2024 platform data, only the top quartile in the $500+ ARPA segment reached at least 100% NRR, and that ARPA cut excludes companies below $300K ARR. Compare both NRR and GRR using the same cohort and period before calling 100% healthy or unhealthy for your company.

#### Why can two CAC payback benchmarks disagree?

They may use different revenue denominators, gross-margin policies, ACV cuts, or company types. Benchmarkit's 2025 definition uses sales and marketing expense, new-customer ARR, and gross subscription margin, while its public-company treatment can use net-new implied ARR that includes churn, downsells, and expansion. Put the formula and the population beside every payback figure before you compare month counts.

#### Can I use mobile-app conversion data for B2B SaaS?

Use it only when the event and product motion match. RevenueCat's 2026 subscription-app report defines D35 as installs that produce at least one paid subscription within 35 days. That is not a B2B landing-page conversion, trial-to-paid conversion, or visitor-to-customer measure. For a web SaaS landing page, use a source that names the page's configured action and its measurement period.

A benchmark earns attention when it changes a decision. That means it has to survive the boring questions first: what is counted, who is counted, when was it counted, and what would we do differently if the gap is real? Write those answers once in the comparison header. Then the next operating review has a number you can argue with for the right reasons.

**Next job: Build your comparison header.** Create one operating-review row with the metric definition, numerator and denominator, period and cohort, ARR, ACV or ARPA, motion, funding, benchmark source and as-of date, and action threshold. Before choosing a calculator or setting a target, fill the comparison header for one current metric in your next operating review.

---

All posts: https://www.productgrowth.blog/archive · Site: https://www.productgrowth.blog