The 90% failure claim has no source. Government data shows about 78% of new business locations survive year one and about half reach year five. Here is the data, what closing really means, and what owners report as their biggest problem.
No, 90% of small businesses do not fail. The claim has no known source. US Bureau of Labor Statistics (BLS) data shows 77.9% of new business locations were still open after one year, and about half of a typical class is still open after five years. That half that closes is real, but closing is not always failing.
Video: “Do 90% of Small Businesses Really Fail?” by Dooza on YouTube (@thedooza), published October 16, 2026. Narrated by an AI voice with AI illustrations; sources below. Watch on YouTube.
The version about restaurants appeared in an American Express ad in summer 2003. It aired during an NBC reality show about opening a restaurant, and said 90% of restaurants fail in the first year. That is according to Parsa and colleagues, in a 2005 paper in the Cornell Hotel and Restaurant Administration Quarterly.
The researchers asked American Express for its data. The company stated in writing that it could not provide data supporting the claim.
The idea is older than the ad. Entrepreneur ran a 2003 piece called "Urban Legend" on the same myth for all startups. In 2007, SBA economist Brian Headd told the magazine he knew of "no substantiated finding" that 9 in 10 fail in year one. Nobody could say where it started.
No. Parsa's team built a model of a market with 1,500 restaurants. They added normal new openings and market growth, then let 90% of new restaurants close in year one. After 20 years the market would have 254 restaurants left, down 84%. In about 94 years there would be none. Run the same model at a 30% first-year rate and the market grows to more than 3,000. The source is the same paper.
Restaurant counts were growing a few percent a year, not collapsing. A 90% first-year failure rate cannot be true and leave streets full of restaurants.
The best source is the BLS Business Employment Dynamics series. It tracks every new business location that has employees. These are establishments (locations), not owners, and businesses with no employees are not included.
| Years open | Still open: 2013 class (BLS) | Still open: 1994-2020 average (SBA Advocacy) |
|---|---|---|
| 1 year | 79.6% | Not listed here |
| 2 years | 68.9% | 67.7% |
| 5 years | 50.6% | 48.9% |
| 10 years | 34.7% | 33.7% |
Sources: BLS, 2013 class and SBA Office of Advocacy FAQ, March 2023.
The most recent data fits the pattern. Of locations opened in the year to March 2024, 77.9% were still open in March 2025. The 2013 class lost 20.4 points in year one, more than in any other year, per BLS. In that class, construction had the sharpest first-year drop of any industry.
So the honest summary is that about half of new employer locations are closed within five years. That part is true. The 90% part is not.
Restaurants are not special either. In a study of about 81,000 full-service restaurants in the western US, economists Luo and Stark found 17% of independent restaurant startups closed in year one, versus 19% for other service startups. The paper opens with the question "Do 9 out of 10 restaurants fail in their first year?" and answers "No." It is a preprint and has not been peer reviewed.
Not always. Brian Headd studied firms started between 1989 and 1992 using Census Bureau data. About half of new employer firms survived beyond four years. Owners of about a third of the firms that closed said the business was successful when it closed. Closing included planned exits, sales and retirement. See the published study; the earlier Census working paper adds that startups with ample capital, employees and education were more likely to survive.
That is the owner's own opinion, not an objective test. But it shows why "half are gone in five years" is not the same as "half failed". The Parsa paper makes a similar point about its own numbers. It followed 2,439 restaurants in Columbus, Ohio, from 1996 to 1999. First-year ownership turnover was 26.16%, and the three-year total was 59.74%. Turnover counts any change of owner, including a retirement or a new partner. It is not the same as closing.
Same restaurants, three different answers: 17%, 26% and 90%. The definition of "failure" moves the number a lot.
The Federal Reserve's 2026 Small Business Credit Survey asked employer firms about problems in the prior 12 months. It was fielded September 3 to November 14, 2025, with 6,525 responses. It is a convenience sample, not a random one, and it lists challenges, not causes of closure.
Figures are from the Fed report.
For venture-backed startups, a small slice of all business, the research firm CB Insights read shutdown stories. In its 2026 update, 70% ran out of capital. CB Insights calls that "almost always the final cause of death, not the root problem". Poor product-market fit, at 43%, was the next most common reason.
Yes, in record numbers. The Census Bureau's Business Formation Statistics show 5,671,836 business applications in 2025, against 3,498,990 in 2019. That is 62% more. Those are applications, and many never hire anyone. Still, if 9 in 10 businesses failed, it would be hard to explain why so many people keep applying.
Dooza is an AI-native company that builds AI products and services for small businesses, including Dooza Managed: an AI admin assistant whose work is checked by a real person, $200/month with the first 7 days free (see pricing). Custom AI agents are scoped on a free 30-minute call (book one).
The Fed survey puts reaching customers and keeping the business growing at the top of the list. Much of the work behind that is written admin that piles up on the owner. Dooza Managed takes that off your plate: email drafts and follow-ups, documents, spreadsheets and research. A real person checks the work. Start with the free week on pricing.
If you need something custom, book a free 30-minute call at dooza.ai/book and we will scope an AI agent around one workflow.
No. The 90% claim has no known source. BLS data for the 2013 class of new business locations shows 79.6% were open after one year, 50.6% after five and 34.7% after ten. These count locations with employees, and closing is not always failing.
It appeared in an American Express ad during an NBC show in summer 2003. When researchers asked for data, American Express stated in writing that it could not provide any, according to a 2005 Cornell paper by Parsa and colleagues.
About half. For the 2013 class, BLS reports 50.6% of new business locations with employees were still open after five years. The SBA Office of Advocacy long-run average for 1994-2020 is 48.9%.
Not always. In a Census-based study by Brian Headd, owners of about a third of closed firms said the business was successful when it closed. Retirements, sales and better job offers all count as closures.
In the Federal Reserve 2026 Small Business Credit Survey, 57% of employer firms named reaching customers and growing sales as an operating challenge, the most common one. The sample is a convenience sample, not a random one.
Automate your business with AI employees that work 24/7. Start with a refundable pilot: 100% refund within 14 days.
McDonald's ended its AI drive-thru test with IBM in July 2024. The likely reason was accuracy, not viral clips. Here is the timeline, the numbers, what rivals found, and the lesson for small businesses.
In "I Fell for 100 AI Scams," YouTuber yikes buys products pushed by AI-generated ads and tests every claim. Here is what held up, the warning signs to watch for, and how honest small businesses can use AI in marketing without looking like a scam.
Got busywork? Dooza it.
Start with a refundable pilot — 100% refund within 14 days. A Dooza engineer scopes it with you on a free 30-minute call. Pricing depends on the product; see pricing.
Refundable pilot · 100% refund within 14 days · No contracts
Dooza builds AI products and services for small businesses. Every product starts with a refundable pilot: 100% refund within 14 days.
Or start now: Dooza Managed, an admin assistant checked by a real person, first 7 days free, then $200/month. See pricing.