Every number on this page has a source
Startup Statistics: Birth Rates, Survival, and Why Companies Fail
Founders trade in round numbers: 90% fail, half die in year three. Most of those claims trace back to one study with a fuzzy definition. This page keeps to figures you can trace — the BLS Business Employment Dynamics series, which counts every employer business in the United States, and CB Insights' March 2026 post-mortem of 385 VC-backed failures. Tables, definitions, and release dates included, so each number can be checked at the source.
Last checked against sources on October 26, 2026
The short answer
What do the actual numbers say?
More companies are being started than ever — 1,065,228 US business births in 2022, the highest in the 30-year series — but each one starts smaller: the average new establishment had 7.3 employees at birth in 1999 and 3.5 in 2023. Survival has barely moved: between 49.8% and 57.3% of cohorts are still operating five years later. Among VC-backed companies, running out of capital (70%) and poor product-market fit (43%) lead the post-mortems.
Five things the tables show
What to take from this
- Business births are at a 30-year record. Competition for attention and capital is not shrinking.
- New businesses start smaller every decade — three or four people instead of seven — which changes what a validation plan has to cover.
- Roughly half of businesses survive five years, and that held steady across cohorts from 1994 to 2018. The base rate is neither doom nor a coin flip.
- Among VC-backed failures, 70% ran out of capital and 43% never found product-market fit. A company can carry more than one cause, so these overlap.
- The median funded startup that died shut its doors 22 months after its last raise. Money buys time, and the time runs out on a schedule you can see coming.
Source: BLS Business Employment Dynamics
Business births hit a record while the average start got smaller
The table below comes from the BLS twentieth-anniversary release of Business Employment Dynamics (January 2024). A birth counts an establishment that appears on the payroll for the first time that year; jobs at birth divided by births gives the average size. Note the directions: after the 2010 trough, births climbed to all-time highs in 2022 and 2023, while the average establishment was born with fewer and fewer people.
| Year | New establishments | Jobs at birth | Average size |
|---|---|---|---|
| 1994 | 569,387 | 4,127,123 | 7.2 |
| 1999 | 650,661 | 4,730,629 | 7.3 |
| 2010 | 560,391 | 2,509,568 | 4.4 |
| 2019 | 770,523 | 3,116,850 | 4.0 |
| 2022 | 1,065,228 | 3,697,739 | 3.5 |
| 2023 | 1,054,052 | 3,710,694 | 3.5 |
Figures read from the chart-data tables of the BLS Spotlight release in October 2026. The series begins in 1992 and is anchored to the Census Bureau's Business Register, so it counts real employer businesses, not survey estimates.
Source: BLS cohort survival tables
Five-year survival has stayed in a narrow band for 25 years
A cohort is every business born in one year; the rate is the share still on a payroll five years later. Selected cohorts from the same release:
| Birth cohort | 5-year survival |
|---|---|
| 1994 | 54.3% |
| 2001 | 52.3% |
| 2003 | 55.3% |
| 2006 | 49.8% |
| 2010 | 56.0% |
| 2018 | 57.3% |
The band runs from 49.8% (the 2006 cohort, which met the financial crisis head-on) to 57.3% (2018). Even recessions move it a few points — which makes "90% of startups fail" impossible to reconcile with any government release.
Source: CB Insights post-mortem study, updated March 2026
Why funded startups actually shut down
CB Insights examined 431 VC-backed companies that closed after 2023 and could classify a cause of death for 385 of them. One company can list more than one, so the shares add up past 100%. Among failures blamed on poor product-market fit, two-thirds had never found a market at all.
| Cause of death | Share of failures |
|---|---|
| Ran out of capital | 70% |
| Poor product-market fit | 43% |
| Bad timing | 29% |
| Unsustainable business model / unit economics | 19% |
These are VC-backed companies with a named post-mortem, not the general business population. Don't read the 70% as your startup's chance of dying — read it as a list of questions a funding plan has to answer.
The shape of the collapse
How much was raised, and how long it lasted
From the same 385 classified failures:
- The 385 companies raised $17.5 billion in equity between them.
- The median raise was $11 million; the mean was $48 million — a handful of megafauna drag the average up.
- The median company closed 22 months after its final raise.
- Nearly a quarter survived more than three years past their last round.
- About 50,000 companies tracked by CB Insights have taken no outside funding since early 2023 — the unfunded majority, invisible in post-mortems because nobody writes them up.
- "Ran out of capital" at 70% does not mean the account emptied the day after a raise; 19% of the same sample died of unit economics that never worked at any scale.
Where the failures clustered
Three sectors account for the largest share of classified, funded failures:
Healthcare and biotech — 62 companies (14%)
The biggest single cluster and the most capital-hungry: those 62 raised roughly $5.1 billion together. Long regulatory timelines mean the burn happens before revenue can exist.
Fintech — 57 companies (13%)
Nearly 60% of these failures were outside the United States, where licensing regimes and incumbent banks move at a different pace than a Silicon Valley deck assumes.
Food and agriculture — 54 companies (13%)
About a third of them were alternative-protein companies, including two that raised nine figures each before closing. Consumer enthusiasm in surveys did not turn into repeat purchases.
What to do with a base rate
What these numbers mean for what you build next
A base rate is a planning tool, not a verdict. The steady message across thirty years of BLS data: about half of new businesses are still hiring after five years — and funded companies that die usually die of cash, which is a symptom. The 43% product-market-fit share is the leading root cause hiding underneath.
The practical translation: test demand before committing to the burn. A week of customer interviews and a landing-page smoke test is cheap; a 22-month glide path from a raise you cannot extend is not. If you work in healthcare, fintech, or food-tech, the tables above say your failure mode is specific and known — write it into the plan.
Maintenance
When these numbers change
BLS publishes Business Employment Dynamics twice a year, with a lag of several quarters; CB Insights updated the post-mortem dataset on March 5, 2026. Every figure on this page is re-checked against the primary sources on each update, and the date above shows the last pass. If a number here no longer matches its source, the sources section has the live links.
Sources
- BLS Spotlight: Business Employment Dynamics, twentieth anniversary (January 2024) — Birth counts, jobs at birth, and cohort survival tables cited on this page.
- BLS Business Employment Dynamics program home — The ongoing semiannual release and its methodology documentation.
- CB Insights: Top Reasons Startups Fail (updated March 5, 2026) — Post-mortems of 431 VC-backed shutdowns, 385 with classified causes.
- US Census Bureau: Nonemployer Statistics — The companion series for one-person businesses; referenced here for definition, not quoted for figures.
Questions these tables raise
Frequently asked questions
startupStatistics.faqQ1Q
startupStatistics.faqQ1A
startupStatistics.faqQ2Q
startupStatistics.faqQ2A
startupStatistics.faqQ3Q
startupStatistics.faqQ3A
startupStatistics.faqQ4Q
startupStatistics.faqQ4A
startupStatistics.faqQ5Q
startupStatistics.faqQ5A
Check the numbers against your own idea
The base rate tells you what typically happens to companies like yours. A Yibud report tells you which assumption in yours is likeliest to break first.
Run a validation report →