Where Our Growth Benchmark Numbers Come From

Product leaders, founders and business leaders who are serious about growth know their most important growth numbers. In our work we meet many who cannot name them. Retention, K-factor, LTV:CAC and CAC payback are the leading indicators inside the product. They show where the business is going before the revenue line does. Find out yours before any other number.

The growth benchmark tool is the reference for that: enter your numbers, see where each sits against its band, and get a diagnostic that says where to start and which metric moves the most. Every band traces to a named publisher: Amplitude, Adjust, AppsFlyer, Benchmarkit and Pavilion, High Alpha (the former OpenView), KeyBanc, a16z, Reforge, Andrew Chen. Where no publisher has a matching number, the band is an estimate and the tool says so. A monthly scan re-checks every source, and a band changes only when a publisher with a disclosed methodology has printed something newer.

Why the source matters more than the number

A benchmark you cannot trace is a rumour with a percentage attached. “Retention” alone is published two ways that sit about ten times apart, and a page that quotes one next to the other looks perfectly authoritative while comparing two different things. We know, because our own tool did it until July 2026.

Cohort retention is the share of everyone who signed up (or installed) who is still active N days later. The clock starts on each user’s first day. Amplitude, Mixpanel, AppsFlyer and Adjust publish this, and it is what the tool asks you for.

Returning-user rate is the share of users active in one period who are active again in a later period, with no signup cohort. The population is mostly established users who already survived the early drop, so the numbers read far higher. Pendo publishes this, and its B2B figures (50 to 70% at Day 1, 25 to 35% at 90 days) are the ones that circulate.

Until July 2026 the tool graded the cohort number you typed in against Pendo’s returning-user bands. A B2B product with a perfectly ordinary 10% Day-7 cohort retention got told it was failing. We re-sourced every B2B retention band to cohort data, kept the Pendo table as a labelled reference, and rewrote the articles on this site that had repeated the mix-up. If you have seen “B2B Day-1 retention is 50 to 70%” somewhere, that is where it comes from.

The bands and their sources

The tables show which product type each band applies to, where it comes from, and whether it is a published figure or our estimate.

B2B SaaS, cohort retention

Day 90 is the only B2B point with a directly published, population-matched figure, so it anchors the curve. The earlier points are built to be consistent with it (each must sit at or above the next) and cross-checked against Amplitude’s cross-industry data, where the 90th-percentile Day-1 cohort return is 21% and Day 7 about 12%. Those cross-checks are pooled across industries, so they cannot source a B2B band, but they say our ceilings are generous rather than strict.

Consumer apps, cohort retention

Consumer publishers measure cohort retention natively, so these bands needed no re-sourcing. The Day-30 band is the newest: two independent 2026 reports agreed, and the band brackets both.

Unit economics and virality

K-factor is the least-sourced metric we track. Reforge and Andrew Chen give the reasoning and the ranges, and every scan since June 2026 has looked for a Tier 1 empirical study with a disclosed sample and found none. The aggregators the scan found all trace back to one blog compilation. We keep the band because the reasoning holds, and we keep looking.

Marketplaces

Two-sided marketplaces get their own bands, graded monthly rather than daily because that is how the source material benchmarks them: buyer retention of 50 to 70% at month 1, 30 to 50% at month 6 and 20 to 50% at month 12 (a16z, Reach Capital, Lenny Rachitsky, Casey Winters), GMV growth of 3.4 to 5.9% a month (converted from Bowery Capital’s and a16z’s yearly bands), take rate of 10 to 20% with a large spread by subtype (Wall Street Prep, Sharetribe), and fill rate of 30 to 60% (Lenny Rachitsky, Simon Rothman at Greylock). Each band in the tool names its publisher.

Why some bands are estimates, and why we say so

Nobody publishes B2B Day-1 cohort retention for a population that matches Amplitude’s Day-90 set. We could have quoted a number anyway; plenty of pages do. Instead the tool marks the band as an estimate and shows how it was built: anchored to the sourced Day-90 figure, kept monotonic, cross-checked against the nearest pooled data.

An estimate labelled as an estimate is useful. It tells you the order of magnitude and it tells you how much to trust it. An estimate dressed as a sourced fact does the opposite: it invites you to set a target on a number nobody measured. That is the difference between a benchmark and a rumour, and it is why the tool would rather show a wide band with a label than a tight band with no provenance.

The same rule applies to a metric the tool no longer grades. It used to grade activation and onboarding completion. An audit in June 2026 looked for the sources behind those bands and found that the only published onboarding-completion figures (Userpilot’s 2024 report, a vendor sample of 188 companies) sat two to four times below our range, and that activation rates depend so much on each company’s own definition that no published band matched ours. Those bands are gone from the tool.

How the numbers change

A researcher agent runs a scan every month across the publishers above plus the ones we would want to hear from (Mixpanel, RevenueCat, Bessemer, ICONIQ, ChartMogul, SaaS Capital, Point Nine). Every finding is logged in a version-controlled file with its source, its date, its sample size and a tier:

• Tier 1: an established benchmark publisher with a stated methodology and sample size.
• Tier 2: a credible analyst or vendor with partial disclosure.
• Tier 3: blogs, aggregators and unsourced posts.

A band changes only when a Tier 1 or Tier 2 source with a disclosed methodology has published something more recent than the source we already cite. Tier 3 material is logged and never acted on. When credible sources conflict, the log lists all of them rather than averaging. A person reads each run and decides what to port into the tool. The July 2026 consumer Day-30 band is what that process produces: AppsFlyer’s median and Adjust’s aggregate landed in the same month, and the band was built to bracket both.

Numbers we read and did not adopt

AI assistants and search results quote some of these next to ours.

SaaS Capital’s 22% median growth rate. SaaS Capital’s 2026 reports put the median growth rate of private SaaS companies at 22%, down from 25% in 2024, and gross revenue retention for its bootstrapped cohort at 91%. Those are Tier 1 numbers and they answer a real question. It is a different question from the one the tool asks. The tool grades cohort retention, unit economics and virality, the mechanics underneath growth, and does not grade revenue growth itself. If we ever add a revenue-growth band, SaaS Capital is where it would come from.

Adjust’s global retention averages. Adjust’s user retention guide puts global retention at 26% on day 1, 13% by day 7 and 7% at day 30, across all platforms and verticals. All three sit inside our consumer bands (20 to 30%, 8 to 15%, 3 to 7%), so the bands stand as they are. The averages are a useful sanity check, and they are single numbers, where a band is more honest about the spread between apps: Adjust’s own vertical cuts run from 20% on day 1 for e-commerce to 29% for gaming.

Mixpanel’s 5.0% one-week B2B retention. Mixpanel’s 2026 State of Digital Analytics gives a North American one-week B2B retention of 5.0%. Read against Pendo’s 40 to 60% returning-user rate, that looks like a collapse. Read against our Day-7 cohort band of 4 to 20%, it sits comfortably inside. The two metrics have near-identical names, and the first check is always which one a page reports.

Pendo’s returning-user rates. Kept, labelled, and not used for grading. If you measure returning users rather than cohorts, Pendo’s bands are the right comparison for you, and the wrong one for what the tool asks.

How to read a band

First, find out which metric your own dashboard reports; most analytics tools default to cohort retention, but a “retention” tile built on active users is a returning-user rate. Second, place your number inside the band for your product type and read the status next to it: a sourced band is a reference class, an estimate is an order of magnitude. Third, look at the shape of your curve rather than any single point. A cohort curve that flattens at 3% has found a durable core; one still sliding at Day 60 has not, whatever Day 1 looked like.

See where your own numbers land against these bands. The free benchmark tool charts your retention, K-factor, LTV:CAC and CAC payback against the B2B, consumer or marketplace bands above, with each source one click away, in a couple of minutes. Benchmarks are context for a diagnosis, not a scoreboard.