The Category Problem: Why Most Benchmarks Are Wrong

"A growth rate of 25% for a $2 million SaaS business is below the median, while growth of 25% for a $20 million SaaS business is above the median."

That line comes from SaaS Capital's 2025 growth benchmarks, built from an annual survey of more than 1,000 private B2B SaaS companies. Same number, opposite verdicts. The only thing that changed was which companies you compared it with.

That's the category problem. Most founders who cite a benchmark pick a real number from a real report. The trouble is that the report describes a different group of companies, and the difference usually flatters them.

This week's Friday Note covers what a category mismatch is, the five benchmarks that get misused most, a worked example with real figures, and a five-point test for a benchmark that will survive an analyst's first question.

A benchmark is only useful if the companies behind it are like yours. Growth rates should be compared by company size and retention by contract value, using the latest edition with its year stated. Read the method section before the headline, show where you sit in the spread, and say so when nothing matches exactly.

What a Category Mismatch Is

A category mismatch is when you compare your metric with a benchmark drawn from a population that doesn't match your company. The categories look close on a search results page. The behaviour underneath can be quite different.

It happens for an innocent reason. Founders search for "SaaS churn benchmark", find a respectable report, and take the headline median. That median pools tiny companies with large ones, cheap self-serve products with six-figure contracts, and early cohorts with mature ones. A pooled figure can hide almost anything.

Statisticians have a name for the extreme version. Simpson's paradox is the case where, as the Stanford Encyclopedia of Philosophy puts it, "an association between two variables in a population emerges, disappears or reverses" once you split it into subgroups. You don't need a full reversal for a benchmark to mislead. A pooled median that sits a few points away from your real peer group is enough to turn a weak number into a strong-looking one.

Why It's a Rules Problem Too

Comparing yourself with the wrong benchmark isn't only a diligence risk. In marketing, it can breach the rules on comparisons.

Rule 3.37 of the UK's CAP Code covers comparisons with competitors nobody can identify, which is what "beats the industry average" is. It says such comparisons must not mislead, and that "the elements of the comparison must not be selected to give the marketer an unrepresentative advantage." Choosing the benchmark that flatters you most is exactly that kind of selection.

For marketing aimed at other businesses, the Business Protection from Misleading Marketing Regulations 2008 set a similar bar. A permitted comparison has to compare products "meeting the same needs or intended for the same purpose", and objectively compare features that are "material, relevant, verifiable and representative".

A pitch deck to a private investor isn't an advert. But the test is a good one to borrow, because it's the test a careful reader applies anyway.

The Five Most Misused Benchmarks in Founder Decks

Five benchmarks cause most of the trouble. Here's each one, and the cut that makes it fair.

1. Growth rates pooled across company sizes. Small companies tend to post higher growth rates. SaaS Capital points to the "denominator effect", where dividing by a small number can produce very high growth rates. Its 2025 brief shows median growth falling as companies get bigger.

The same brief makes the point about the top end too. A $2 million company "needs to be growing more than 50% year-over-year to be in the top 25% of its peers, whereas that bar is 31% growth for a company with $20 million of ARR."

The bar moves with the funding market as well. In May 2023 Jason Lemkin, the SaaStr founder and investor, gave the example of a SaaS company growing 100% a year at $5m ARR whose growth slips to 80%:

2. Retention and churn without contract value. For retention, SaaS Capital is blunt about the right cut. In its 2025 retention benchmarks it says "benchmarking by ACV is the best starting point", because companies that share a similar selling price "have the most in common".

The medians look close, but the gap between a 98% and a 106% net retention business compounds every year. Churn has a second trap. Customer churn and revenue churn are different measures, and ChartMogul's metrics guide notes they "can differ significantly depending on how revenue is distributed across your customer base." Make sure you're quoting the same one the benchmark measured.

If those terms are fuzzy, ChartMogul's Nick Franklin runs through churn, gross retention and net retention, among others, in this 11-minute explainer. It's worth watching before you compare yourself with anyone.

Play Video (Loads YouTube)

3. CAC from pooled startup surveys. Customer acquisition cost depends on stage, channel, geography and price point. A pooled average that mixes a self-serve tool with an enterprise sales team tells you very little about either. If you can't find a cut that matches your sales motion, compare CAC payback against your own gross margin instead, and say why.

4. Engagement and satisfaction scores from consumer benchmarks. Usage ratios and satisfaction scores from consumer apps reflect different habits from a B2B tool people use for work. A consumer average is the wrong yardstick for a workflow product, in either direction. Find a benchmark built from products with the same usage pattern, or skip the comparison.

5. Market size from reports that aggregate adjacent markets. Analyst market sizes often add up total spend across a broad sector, much of which your product can't reach at your price. A deck that cites an industry report from years ago for its market size may be quoting a total addressable figure rather than the segment the company can serve. The fix is a current, scoped figure from a named source.

A Worked Example With Real Numbers

Here's the correction pattern applied to a typical growth claim. The company is made up. The benchmark figures are SaaS Capital's.

You might think the honest version hurts the deck. In my view it does the opposite. An investor who finds the flattering comparison on their own starts wondering what else was chosen to flatter. An investor who sees you've already done the honest comparison can move on to the next slide.

How to Find the Right Benchmark

The right benchmark comes from reading the method section first and the headline second. Before you cite any figure, answer four questions.

Who was surveyed? How many companies, and how were they recruited? A large, repeated survey with a stated method is worth far more than a one-off poll of a few dozen founders.

Is your exact type of company in the sample? Check the bands the report uses: company size, contract value, sales motion, customer type. Then match the cut to the metric. SaaS Capital's 2026 brief found that contract value doesn't appear to correlate with growth overall, even though it matters a lot for retention. The right category depends on what you're measuring.

When was the data collected? The publication date and the collection date are different. SaaS Capital's 2025 growth brief comes from a survey run in Q1 2025, and it reports growth for 2024.

What does the spread look like? A median on its own hides a wide range. Knowing where you sit in the distribution matters more than knowing the middle of it.

There's a catch with getting more specific. The narrower the cut, the fewer companies in it, and the less certain its numbers become. Pew Research Center makes the same point about polls: estimates for subgroups "have fewer cases, their margins of error are larger" (Pew, 2016). So use the narrowest cut that still has a decent sample, and say how big it is.

Recency: The Other Problem

A benchmark from the right category can still be out of date. Category mismatch gets the attention. Recency gets less, and it can move a number just as far.

Investors notice it too. Tomasz Tunguz, who founded the venture firm Theory Ventures, wrote this about the old Series A rule of thumb in July 2023:

SaaS Capital's own series shows how quickly a "normal" growth rate shifts. The pooled median fell by eight points across three editions of the same survey.

A founder raising in 2026 who quotes "the 30% SaaS median" is quoting a market that no longer exists. In our evidence grading, which you can see on our methodology page, a source older than 12 months grades C at best. It can still support a point, but the gap has to be stated.

If you want the latest numbers, ChurnZero walked through SaaS Capital's 2026 survey in this webinar, including where retention landed and how growth shifted across ARR bands.

Play Video (Loads YouTube)

A Framework for Defensible Benchmarks

A defensible benchmark passes five tests. I'd run every benchmark in a deck through them before it goes to an investor.

1. Named method. It comes from a research firm or institutional source that publishes how it collected the data.
2. Matching sample. The sample includes companies like yours, cut by the dimension that matters for that metric.
3. Current. It's the latest edition, with its year stated, and the data-collection date is clear.
4. Primary. You're citing the original report. Secondary summaries often misquote or strip the context.
5. Same definition. Your metric is measured the same way the research measures it. Free trial and free-plan users are out of customer counts, for example.

If a benchmark fails one test, fix it or caveat it. If it fails two, I'd take it out. A deck with fewer, honest comparisons reads as more confident than one stuffed with flattering ones.

For the bigger picture of what investors check, see how investors verify startup claims and what UK investors verify in a Series A data room. An earlier Friday Note on pre-empting investor objections pairs well with this one. And our post on what a 43% startup failure rate really means breaks one famous statistic down by stage, which is the same idea applied to failure rates.

Where ThriveFinity Fits

Checking benchmarks is a core part of Release Assurance. Every claim in scope, including each benchmark comparison, is checked against its source, given an A to E grade and tagged Verified, Supported, Flagged or Unverifiable. Weak ones come back rewritten, the way the worked example above was. A named verifier signs the Assurance Note. You can see the format in our anonymised sample, which includes a market-size claim that was marked Supported once a footnote naming its three sources was added.

If you'd rather check your own deck first, the free Series A claim checklist walks through the claims investors look at first. The free First Signal will look at one claim, such as your headline benchmark, in under an hour.

We're not the right choice if you need a bespoke benchmark built from primary research, or a formal valuation. We check whether the comparison you've made is fair and sourced. We don't create the dataset.

Pranav Unni

Founder · ThriveFinity Connect on LinkedIn →

Pranav Unni is the founder and lead verifier of ThriveFinity. He reads and signs every paid deliverable personally, and writes about go-to-market decisions for established B2B companies.