The Startup Statistics Fact-Check

Usually credited to

Harvard Business Review / "a U.S. Census study"; occasionally misattributed to the Kauffman Foundation

Actual source

Azoulay, Jones, Kim & Miranda, "Age and High-Growth Entrepreneurship," American Economic Review: Insights 2(1) (2020) — U.S. Census Bureau administrative data on 2.7 million founders of employer firms started in the United States between 2007 and 2014. The mean founder age across all startups was 41.9. The 45.0 figure is the mean age only for founders of the top 0.1% of firms by five-year employment growth (about 1,700 firms); the top 1% averaged 43.7 and the top 5% averaged 42.1. The finding first circulated via the authors' July 2018 HBR article, whose headline used the 45 figure.

Verdict

The number is real but applies to a much narrower group than most citations imply. In the study, 45.0 is the average founding age among the top one-in-a-thousand fastest-growing new firms; the average across all 2.7 million founders was 41.9. The HBR headline (“The Average Age of a Successful Startup Founder Is 45”) is accurate on its own terms — “successful” meant the extreme upper tail — but the qualifier is routinely dropped, producing the false claim that the average founder is 45. The underlying data also predates the post-2014 founding waves.

How it spread

Azoulay, Jones, Kim & Miranda (AER: Insights, 2020) → HBR summary headline (July 2018) → statistics roundups and news articles that drop the top-0.1% qualifier.

What to say instead

A study of U.S. Census data on firms founded 2007–2014 (Azoulay et al., published 2020) found the average age of startup founders was 41.9; among the fastest-growing 0.1% of those firms, the average founder age was 45.

Usually credited to

CB Insights

Actual source

CB Insights, "The Top 20 Reasons Startups Fail" (original post-mortem analysis) (2014) — A tally of reasons cited in 101 public startup-failure post-mortems — essays self-published by founders. “No market need” was noted in 42% of them, as one of multiple reasons per company (totals far exceed 100%). The sample was self-selected (founders who chose to write publicly about their failure), not a representative survey of startup outcomes.

Verdict

CB Insights itself no longer stands behind this number. The URL that hosted the post-mortem series now carries a 2024 replacement analysis of 431 VC-backed companies that shut down since 2023, with a different framing: 70% ran out of capital (which CB Insights describes as the final symptom, not the root cause), 43% cited poor product-market fit, 29% bad timing, and 19% unsustainable unit economics. The 42% figure continues to circulate widely despite being superseded by its own publisher.

How it spread

CB Insights post-mortem analysis (2014, 101 companies) → “startup statistics” roundups (Founders Forum, Failory, and thousands of others) → still cited as current despite CB Insights replacing the analysis at the same URL in 2024.

What to say instead

In CB Insights' current analysis of 431 VC-backed startups that shut down since 2023, poor product-market fit was the leading root cause of failure, cited in 43% of cases; the older “42% no market need” figure came from a since-replaced analysis of 101 self-published post-mortems.

Usually credited to

CB Insights

Actual source

CB Insights, "The Top 12 Reasons Startups Fail" (2021 update of the post-mortem series) (2021) — The 2021 update of CB Insights' post-mortem series, tallying 111 self-published startup post-mortems written since 2018. “Ran out of cash / failed to raise new capital” ranked first at 38%, with “no market need” second at 35%. The original page is gone (it now redirects to the 2024 replacement analysis); this entry was verified against an Internet Archive snapshot, with the 38%/35% percentages corroborated by contemporaneous coverage of the report. Same self-selected-sample caveat as the original series.

Verdict

This is the second-generation version of the same retired series as the “42% no market need” claim — in the 2021 update the ranking flipped, putting cash first at 38%. CB Insights replaced the whole series in 2024 with an analysis of 431 VC-backed shutdowns since 2023, which reports that 70% of failed companies ran out of capital but explicitly reframes that as the final symptom rather than the root cause. Writers citing 38% (or the 2014 version's 29%) are quoting a number its publisher has withdrawn.

How it spread

CB Insights “Top 12 Reasons Startups Fail” (2021, 111 post-mortems) → statistics roundups citing “38% run out of cash” → superseded by CB Insights' 2024 analysis published at the same URL.

What to say instead

In CB Insights' current analysis of 431 VC-backed startups that shut down since 2023, 70% ran out of capital — but CB Insights describes running out of money as the final symptom, with poor product-market fit (43%) the leading root cause.

Usually credited to

Startup Genome (or cited with no source at all)

Actual source

Startup Genome, "Startup Genome Report: Premature Scaling" (v1.2, edited March 2012) (2011) — Nothing, in the case of the 90% figure: the sentence “More than 90% of startups fail, due primarily to self-destruction rather than competition” appears in the report's introduction as an unsourced framing assertion, not a research finding. The report itself measured premature scaling in roughly 3,200 self-reported high-growth internet startups using the Startup Genome Compass benchmarking tool. The original host domains are dead; the PDF survives only on third-party mirrors (the linked mirror was verified in 2026).

Verdict

The most common attribution — “according to Startup Genome, 90% of startups fail” — credits a study that never measured failure rates; the line was scene-setting in a 2011 report about premature scaling. No primary study supports 90% as a general business failure rate. Federal Business Employment Dynamics data show about 22% of new U.S. employer businesses close within their first year and roughly half within five years; a cohort takes about three decades to approach 90% closure (87.4% of establishments opened in the year ending March 1994 had closed by March 2025). The figure may loosely describe venture-scale return outcomes, but as a general claim it is folklore.

How it spread

Startup Genome report introduction (2011, unsourced assertion) → tech press and Forbes contributors (“according to a report by Startup Genome, 90% of startups fail”) → cited bare in statistics roundups as a general truth.

What to say instead

There is no primary study behind “90% of startups fail.” U.S. Bureau of Labor Statistics data show about one in five new employer businesses closes within its first year and roughly half close within five years.

Usually credited to

Usually uncited or credited to secondary roundups (Exploding Topics, Zippia); occasionally "a Harvard study"

Actual source

Gompers, Kovner, Lerner & Scharfstein, "Performance Persistence in Entrepreneurship," Journal of Financial Economics 96(1) (2010) — Venture-capital-backed U.S. companies first funded between 1986 and 2000, with “success” defined as going public or filing to go public by December 2003 — a definition that excludes acquisitions. The published abstract reports a 21% success rate for first-time entrepreneurs, versus 30% for entrepreneurs whose prior venture went public and 22% for previously failed entrepreneurs. The widely quoted 18% figure appears in pre-publication working-paper versions and in specific regression specifications of the paper (17–18%), not in the published abstract. The raw, unadjusted success rate on first ventures was 25.3%.

Verdict

The study is real and well regarded, but the claim as circulated has two problems. First, vintage: outcomes were measured through December 2003, under an IPO-centric definition of success — the venture market and exit environment have changed beyond recognition since. Second, the commonly cited 18% does not match the published paper's headline figure of 21%; it traces to earlier drafts. Quoting any of these numbers as a current founder success rate, undated, misleads readers.

How it spread

Gompers, Kovner, Lerner & Scharfstein (NBER working paper 2006; JFE 2010) → Zippia and Exploding Topics → DemandSage, Embroker, Failory → cited undated as a current success rate, usually with the working-paper-era 18% figure.

What to say instead

A Journal of Financial Economics study of VC-backed companies funded 1986–2000 (success defined as an IPO or IPO filing by December 2003) found first-time founders succeeded about 21% of the time, versus 30% for founders with a prior IPO — figures now more than two decades old.

Usually credited to

U.S. Bureau of Labor Statistics

Actual source

U.S. Bureau of Labor Statistics, Business Employment Dynamics — establishment age and survival tables (Table 7) (2025) — Survival of all private-sector establishments (business locations with employees) by opening year, measured March to March. In the most recent data, 22.1% of establishments opened in the year ending March 2024 had closed by March 2025, and 48.6% of the March 2020 cohort had closed within five years; long-run first-year closure rates cluster around 20–25%. Figures are computed from BLS Table 7 (published as data tables, not prose) and were cross-checked against independent analyses of the same table. “Establishments” are locations, not firms, and cover every kind of employer business, not just startups.

Verdict

This one holds up: the claim tracks the federal data closely (the latest cohort's first-year closure rate is 22.1%, and about 20% has been a fair long-run summary). The common failure mode is misapplication, not misquotation — the BLS series covers all new employer businesses, from restaurants to landscaping firms, so quoting it as a statistic about venture-backed startups transplants it to a population it does not measure. It is also frequently mixed on the same page with the incompatible “90% of startups fail” claim.

What to say instead

U.S. Bureau of Labor Statistics data show that about one in five new private-sector businesses closes within its first year (22.1% for the cohort opened in the year ending March 2024), and roughly half close within five years — figures that cover all employer businesses, not just venture-backed startups.

Usually credited to

Zippia (often uncredited, via DemandSage and other statistics roundups)

Actual source

Zippia, "Founder Demographics and Statistics in the US" (resume-profile analysis; yearly data tables end at 2021) (2021) — A jobs-site analysis of resume profiles: Zippia cites a database of 30 million profiles and roughly 26,430 people “currently employed” with the job title “Founder.” The headline gender split (64.3% male / 35.7% female) is identical to the page's 2021 table row, and every year-by-year table on the page (gender, ethnicity, unemployment) stops at 2021, although the page is stamped “[2026]” and “Updated January 8, 2025.” It measures people whose resumes list “Founder” as an employed job title — the page files the role under its production-and-manufacturing job category — not startup founders as an economic population.

Verdict

The numbers are quoted faithfully from Zippia's page, but the page itself is the problem: the headline figures come from data tables ending in 2021 while the page's year stamp rolls forward automatically, and the underlying dataset is a resume-title scrape rather than any census of startup founders. These figures cannot support claims of the form “X% of startup founders are Y,” yet they anchor the demographics sections of several top-ranking startup-statistics pages.

How it spread

Zippia resume-profile tables (data ending 2021, page year-stamp rolls forward) → DemandSage, Embroker, and other “startup statistics” pages → cited as current startup-founder demographics.

What to say instead

Zippia's jobs-site analysis of roughly 26,000 resume profiles listing “Founder” as a job title (yearly data through 2021) found 64.3% were men and 62.8% white; there is no current, representative demographic census of startup founders that supports these as 2026 figures.

Usually credited to

National Foundation for American Policy / Stuart Anderson

Actual source

National Foundation for American Policy, "Immigrants and U.S. Billion-Dollar Companies" (Stuart Anderson) (2026) — Privately held U.S. companies valued at $1 billion or more as of April 2026 — 775 companies, venture-financed, not yet listed on a stock market, tracked via CB Insights. NFAP found 455 of the 775 (59%) were founded or cofounded by at least one immigrant; approximately 66% were founded by immigrants or the children of immigrants; and the 455 immigrant-founded companies had a collective value of $5.0 trillion.

Verdict

Current and correctly cited when the 59% figure is used with its June 2026 vintage. The commonly seen 55% is not wrong so much as stale — it is the figure from NFAP's 2018 and 2022 editions of the same analysis (in 2018, 50 of 91 U.S. unicorns). The definition matters when quoting: the population is privately held billion-dollar companies (unicorns), not all startups or all public companies.

How it spread

NFAP policy briefs (2018 and 2022 editions, 55%) → June 2026 update (59%); the older 55% still circulates in roundups that have not caught the update.

What to say instead

A June 2026 National Foundation for American Policy analysis found immigrants founded or cofounded 59% (455 of 775) of U.S. privately held startup companies valued at $1 billion or more as of April 2026, up from 55% in NFAP's 2018 and 2022 reports.

Usually credited to

Ali Tamaseb, "Super Founders"

Actual source

Ali Tamaseb, "Super Founders: What Data Reveals About Billion-Dollar Startups" (PublicAffairs) (2021) — A hand-collected dataset of roughly 30,000 data points (65 factors per company) on the 200+ U.S. startups founded between 2005 and 2018 that reached billion-dollar valuations, compared against a random control group of VC-funded startups from the same period that did not. Median founder age at founding for the billion-dollar group was 34, with a range of 18 to 68. The primary source is the book itself (published May 2021, no stable public URL); the linked CNBC article quotes the finding directly from the book and was used for verification alongside the author's own published summaries.

Verdict

Accurately cited: the book does report a median founding age of 34 for unicorn founders, and Tamaseb's methodology (case-control, hand-collected) is described transparently. The caveat writers should carry is the data window — it closed at the end of 2018, so the figure describes a 2005–2018 founding cohort and predates the current AI startup wave. It is a well-sourced historical statistic, not a live number.

What to say instead

Among U.S. startups founded between 2005 and 2018 that reached billion-dollar valuations, the median founder age at founding was 34 (Ali Tamaseb, Super Founders, 2021) — a cohort statistic whose data window closed in 2018.

Usually credited to

Carta

Actual source

Carta, Founder Ownership Report 2026 (2026) — Fully diluted equity held by founding teams of startups on Carta's cap-table platform, based on priced rounds raised from 2021 through 2025. Median founding-team ownership was about 56% after a seed round and 36% after a Series A. At Series B, the median AI founding team held 27.3% of fully diluted equity versus 21.8% for non-AI teams. The sample is Carta-platform companies — largely Delaware C-corp, venture-track startups — so bootstrapped and non-VC businesses are not represented.

Verdict

Current and correctly cited; the figures above were verified directly against Carta's published report page. The structural caveat is selection: Carta's data describes companies on its platform, which skews to U.S. venture-track startups. Within that population it is the most authoritative ownership dataset available, refreshed annually.

What to say instead

Per Carta's Founder Ownership Report 2026 (priced rounds raised 2021–2025 on Carta's platform), the median founding team retains about 56% of fully diluted equity after a seed round and 36% after a Series A; at Series B, median AI founding teams hold 27.3% versus 21.8% for non-AI teams.

Usually credited to

Startup Genome

Actual source

Startup Genome, "Startup Genome Report: Premature Scaling" (v1.2, edited March 2012) (2011) — Self-reported benchmarking data from roughly 3,200 high-growth internet startups that used the Startup Genome Compass tool in 2011. Key finding #6 states that solo founders took 3.6x longer to reach “scale stage” than a founding team of two, and were 2.3x less likely to pivot. “Scale stage” is the tool's own stage classification (based on self-reported progress), not revenue, funding, or survival. The original host domains are dead; the report survives only on third-party mirrors (the linked mirror was verified in 2026).

Verdict

The finding is real and traceable, but it is fifteen years old, drawn from a self-selected sample of internet startups that opted into a benchmarking tool and self-reported their stage, and it has not been replicated on modern data. It is often quoted as settled evidence that solo founding is a mistake; meanwhile solo founding has become markedly more common (Carta's 2026 report puts solo founders at about 36% of new startups formed on its platform in 2025). Treat the 3.6x as a dated, sample-specific result, not a law.

How it spread

Startup Genome Report (2011, Compass benchmarking data) → startup-advice posts and statistics roundups (“solo founders take 3.6x longer to scale”) → cited undated, usually linking to dead or mirrored PDFs.

What to say instead

A 2011 Startup Genome analysis of roughly 3,200 self-reported internet startups found solo founders took 3.6x longer to reach its “scale stage” benchmark than two-person founding teams — a fifteen-year-old finding from a self-selected sample that has not been replicated on current data.