The most cited statistic in startup writing answers a different question than the one most founders are asking. The 90% figure comes from Startup Genome, and what it counts is venture-backed startups that never deliver venture-scale returns. Under that definition, a profitable company doing $5 million a year is a failure if it raised a large round and never returned a multiple of it.
That is a real measurement of a real thing. It is a venture-outcome rate rather than a survival rate, and the two get quoted interchangeably. A founder who never raised anything reads "90% fail" as a statement about whether their product will still exist next year, which is not what the number was built to say.
Producing a rival number for "failure" would not help either, because failure has no agreed definition. Pivots, acquihires, dormant side projects and quiet abandonment all get counted or not depending on who is writing.
What can be measured is narrower and harder to argue with: whether a product that connected a payment account ever took money, and whether it still does.
Three states, 8,817 products
Every startup below connected a live payment account (Stripe, RevenueCat, Polar, Lemon Squeezy, Paddle or Dodo) to a public verification service. Lifetime revenue and current monthly revenue are both read from the payment processor.
| State | Startups | Share |
|---|---|---|
| Currently earning | 4,054 | 46.0% |
| Earned money, now at $0 | 2,705 | 30.7% |
| Never earned anything | 2,058 | 23.3% |
76.7% took money at least once. Of those, 40% no longer do.
That last figure is the closest honest analogue to a failure rate in this dataset. It does not mean the company is gone. A founder may have paused billing, or moved to a plan the processor does not report. It does mean the revenue that existed has stopped.
The median product that stopped had earned $136
For the 2,705 products that earned and stopped, lifetime revenue across their whole life:
| Lifetime revenue | Percentile |
|---|---|
| $20 | 25th |
| $136 | 50th (median) |
| $2,075 | 75th |
| $26,065 | 90th |
69% took less than $1,000 in total, ever. The median took $136.
This reframes what failure looks like at this end of the market. These are not businesses that scaled and collapsed. Most are products that found two or three paying customers, ran for a few months, and stopped. The break happened at the very beginning, in the gap between building something and finding anyone who wanted it.
That gap is cheap to investigate and expensive to skip. Demand signals are visible before you write any code, and a product that never finds its first customer usually could not have found one.
Surviving the first year does not move you out of the band
Grouping by founding year shows the three states moving:
| Founded | Startups | Never earned | Earned, stopped | Earning |
|---|---|---|---|---|
| 2026 | 3,088 | 36% | 27% | 37% |
| 2025 | 2,863 | 20% | 34% | 46% |
| 2024 | 770 | 9% | 37% | 54% |
| 2023 | 350 | 7% | 33% | 60% |
| 2022 | 130 | 7% | 28% | 65% |
| 2021 | 104 | 5% | 38% | 57% |
| 2020 | 79 | 3% | 38% | 59% |
Two readings, and the second is the interesting one.
"Never earned" collapses with age: 36% of 2026 products against 3% of 2020 products. Part of that is products still finding their first customer. Part is survivorship, since a product that never earns anything eventually stops being maintained and stops appearing.
"Earned, then stopped" does not collapse. It sits between 27% and 38% in every cohort from 2020 to 2026. Six years out, roughly a third of products that once took money have stopped taking it.
What this replaces
Instead of a venture-return rate borrowed from a population these founders are not in, three checkable statements about products that took real payments:
- 23.3% of products with a payment account never took a single payment.
- 30.7% took payments and no longer do.
- Of those that stopped, 69% had earned under $1,000 in total.
None of these is a failure rate. All three are measurable, and all three point at the same place: the difficulty is not scaling, it is the first customer.
For the products that do get past the first customer, the next threshold is $1,000 a month, which is where the median growth rate turns positive. Our pre-PMF benchmarks cover that band.
That is also where the revenue distribution lands. Of the same 8,817 startups, 54% are at exactly $0 a month right now and only 12.1% clear $1,000. The two readings describe one problem from different angles: most products never find someone willing to pay, and among those that do, most find too few to matter.
The practical consequence is about sequencing. A first customer is evidence that a problem is real and that someone will pay to have it solved, and it is available before the build rather than after it. Market validation and an honest read on willingness to pay are what stand between a product and a $136 lifetime.
Method, and what this data can't tell you
Source. Public revenue profiles from TrustMRR, where founders connect a payment provider so revenue can be verified independently. Snapshot of 8,817 startups, August 2026.
"Earned then stopped" is not "shut down." The signal is lifetime revenue above zero with current monthly revenue at zero. A founder who paused billing, moved to a processor outside the dataset, or disconnected the integration looks identical to one who quit. The figure is an upper bound on quitting.
"Never earned" includes the not-yet. A product that connected its payment account last month and launches next month sits in this bucket. That is why the 2026 cohort is so much higher, and it is why the cohort table matters more than the headline.
Survivorship affects the older cohorts. Products that disappear entirely stop being counted, so the 2020 and 2021 rows describe survivors rather than the original population. The true "earned then stopped" share for those years is higher than shown.
Cohort coverage. The founding-year table covers 7,384 of the 8,817 startups, since not every profile lists a year.
Selection. These founders chose to connect a payment account to a public verification service. That skews indie and bootstrapped, with no venture-funded companies. It is a census of a particular population rather than a sample of all startups.
No costs. Revenue is not profit. Nothing here says whether any of these products made their founder money.
On the 90% figure. The characterisation at the top is about what Startup Genome's number measures, not about whether it is accurate. Its framing is a venture-outcome one, and secondary write-ups describe it the same way. Where a shutdown rate is what you want, the U.S. Bureau of Labor Statistics series is the right one: roughly 20% of new establishments close within a year and about half within five.
Sources
TrustMRR — public revenue profiles The underlying data. Founders connect Stripe, RevenueCat, Polar, Lemon Squeezy, Paddle or Dodo so their revenue is read from the processor and published on a public profile. Every figure here is an aggregate computed over the 8,817 profiles visible in the August 2026 snapshot. The per-company figures belong to TrustMRR and to the founders who chose to publish them.
For failure rates, survival curves by business age and the reasons companies shut down, see our companion reference on startup idea validation statistics, which draws on CB Insights, the U.S. Bureau of Labor Statistics and Startup Genome.
Cite this
Of 8,817 indie startups with verified payment data, 23.3% never took a single payment and 30.7% took payments and no longer do. Of those that stopped, 69% had earned under $1,000 in total. — Scoutr, Startup Mortality Report 2026, from TrustMRR public revenue profiles
If you reference these numbers, a link back lets your readers reach both this analysis and the underlying profiles.