What percentage of startups fail? The numbers by year

About one in five new US businesses closes within its first year, and about half are gone within five years. Those are the counts of the US Bureau of Labor Statistics (BLS). It follows new private business locations with paid staff, each March, from the year they open. Of the 677,876 locations that opened in the year to March 2015, 34.7 percent were still open in March 2025, ten years on.

The often-quoted 90 percent appears in writing about high-growth startups and their investors, where failure can mean falling short of a projection. For all private businesses together, no yearly group in the BLS table has lost 90 percent. Of the locations that opened in the year to March 1994, 87.4 percent had closed after 31 years.

0%25%50%75%100%100.0%79.6%69.1%61.4%55.4%50.2%46.4%43.6%40.3%37.4%34.7%

012345678910

	Opened in the year to March 2015
0	100.0%
1	79.6%
2	69.1%
3	61.4%
4	55.4%
5	50.2%
6	46.4%
7	43.6%
8	40.3%
9	37.4%
10	34.7%

About half were gone after five years and about two thirds after ten.

Source: US Bureau of Labor Statistics, Business Employment Dynamics, Table 7 (survival by opening year), data to March 2025

How many startups fail in the first year?

The first year takes the largest share. Of the 988,310 locations that opened in the year to March 2024, 77.9 percent were still open in March 2025. The other 218,861, or 22.1 percent, had closed.

That share has held steady for three decades. Across the 31 yearly groups the BLS has followed through a first year, between 19.1 and 24.8 percent closed in it. The worst first years came for the groups of the year to March 2008, at 24.8 percent, and March 2001, at 24.3 percent. The year to March 2022 brought 1,065,228 openings, the most in the series, and 23.7 percent of them closed within a year.

What percentage of small businesses fail within five years?

About half. Across the 27 groups the BLS table follows for five years, a mean of 49.3 percent were still open at that point. The range is narrow, from 45.4 percent for the year to March 2006 to 51.9 percent for March 2018. Of the 767,573 locations that opened in the year to March 2020, 51.4 percent were open in March 2025.

How many survive ten years or longer?

About one in three reaches its tenth year. In the BLS table, a mean of 33.9 percent did, across 22 groups, never below 32.4 or above 35.3 percent. After that the losses slow down. One location in five is still open after 20 years. The first group in the series, from the year to March 1994, still had 12.6 percent of its 569,387 locations open in March 2025.

The mean covers every yearly group the table follows that far: 31 groups after one year, seven after 25 years. We calculated it from Table 7 of the BLS data.

The BLS puts the pattern in one line: “Survival rates follow a similar path, regardless of the birth year”. The groups of 1994, 2005 and 2015 kept 33.6, 33.8 and 34.7 percent of their locations after ten years.

Do 90 percent of startups fail?

For all new businesses together, the government table has never reached that figure. The 1994 group had lost 87.4 percent after 31 years. Only two sectors passed 90 percent, and only after more than two decades. In information, the group of the year to March 2001 had lost 90.5 percent after 22 years. In transportation and warehousing, the 1994 group lost 90.4 percent in 31 years.

The two oldest texts with the figure that we could read date from 2011. The first is a March 7, 2011 article from Harvard Business School. In it, senior lecturer Shikhar Ghosh gave three failure rates for start-ups, each for a different meaning of failure:

Ghosh’s 90 to 95 percent counts companies that made a projection and fell short of it. Closing a business is a different event, and that is what the government tables count.

Five months later, on August 29, 2011, Steve Blank put the figure in a blog post. Its second sentence reads: “More than 90% of startups fail, due primarily to self-destruction rather than competition”. That sentence cites no data. The post introduces the Startup Genome Project, which had gathered data on more than 3,200 “high growth technology startups”, in the project’s own words.

Why do sources give different numbers?

Mostly because they count different units. The BLS counts establishments, which the Census Bureau defines as “a fixed physical location where economic activity occurs”. A company with three shops runs three of them. The BLS dates each one by its first quarter with paid staff. Its FAQ takes a chain store as the example. A new branch “would be a new opening under the establishment methodology while it would be an expansion under the firm methodology”.

The Census Bureau’s Business Dynamics Statistics count firms instead. A new firm there is one whose locations are all new that year. In 2023 it counted 508,777 new firms, while the BLS counted 1,049,106 new locations in the year to March 2023. The odds of survival still come out close. We set the two series side by side over the same opening years:

These are means over the groups that opened from 1994 to 2022 (one year), to 2021 (two years) and to 2018 (five years). We calculated them from BLS Table 7 and the Census Bureau’s 2023 firm age table.

Startup or small business: which number applies to you?

Both tables follow only businesses with paid staff. The BLS data leave out self-employed workers and locations with no employees, and the Census series leaves out the self-employed too. Such employer businesses are a minority. By the Census Bureau’s count, they made up 21.6 percent of all US businesses in 2023. The SBA’s Office of Advocacy counts 36,207,130 small businesses, or 99.9 percent of all businesses.

The tables also put every new restaurant, shop and contractor beside the companies built to grow fast. They do not mark which is which.

Is closing a business the same as failing?

Many closures are planned. Brian Headd’s 2003 study in Small Business Economics used Census Bureau data. It found that “about a third of closed businesses were successful at closure”. It concludes that many owners “may have executed a planned exit strategy”, sold “a viable business” or “retired from the work force”.

Timothy Bates studied small businesses started from 1989 to 1992 and closed by 1996. Their owners “often described their firms as “successful”” when they closed, and were moving on to more attractive alternatives. His Census Bureau paper appeared in October 2002.

The BLS defines a closing by employment alone: a location with staff in one quarter and none in the next. Its tables cannot tell a failure from a retirement or a planned exit.

Which industries have the highest failure rate?

The sector matters far more than the year of opening. In the BLS tables, ten-year survival ranged from 24.5 to 53.0 percent across sectors, against 32.4 to 35.3 percent across opening years. Mining, quarrying, and oil and gas fared worst, and agriculture, forestry, fishing and hunting best:

Each share is the part still open in March 2025. Five years means the locations opened in the year to March 2020; ten years, those of March 2015. The sector names are the BLS industry names, shortened into plain words. We calculated the shares from the 19 sector versions of Table 7.

The first year shows the same spread. Information kept 71.6 percent of its new locations through year one, and agriculture 85.7 percent. Restaurants and hotels came second, at 85.3 percent, though ten years on they sit in the middle.

What do these numbers mean for a founder?

The odds improve with every year a business lasts. On the table’s averages, a location that survives its first year has an 86.0 percent chance of lasting the second. By year five, the chance of lasting one more year is 91.0 percent; by year ten, 93.7 percent; by year twenty, 95.3 percent. Year one is the steepest, with 21.3 percent of a new group closing.

The sector shapes the long run more than the timing does. A new location in information or mining faces worse ten-year odds than one in utilities or farming. These are odds for locations with paid staff, though. A founder working alone sits outside the tables. A founder raising money against a plan faces the investors’ measure above, where missing the plan counts as failing.

What do people get wrong about these numbers?

• Half in the first year. Half is the five-year figure. About one in five new locations closes in year one.
• Ninety percent. That figure counts missed projections, in Ghosh’s definitions. For all private businesses, no yearly group has lost 90 percent, even after 31 years.
• Every closure a failure. About a third of closed businesses were successful at closure, in Headd’s study.
• A location for a company. The BLS counts each new branch of a chain as an opening. The Census Bureau counts firms and finds about half as many new ones.
• Small business numbers for everyone. The tables follow businesses with paid staff, about one in five US businesses.

Where do these numbers come from?

The BLS builds Business Employment Dynamics from the Quarterly Census of Employment and Wages. That census rests on the reports most employers file each quarter for unemployment insurance. The age data measure change from March to March. A location is born on its “date of first positive employment”. Each March, the table counts how many of a group are still open. Table 7 follows 32 yearly groups, from the year to March 1994 to the year to March 2025.

The data leave out self-employed workers, private households, government and locations with no employees. In the first quarter of 2025 they covered 9.5 million private locations. The newest locations are small. Those that opened in the year to March 2025 averaged 3.6 workers, half the 7.2 of the 1994 group.

The latest year came with the BLS release of December 17, 2025. The next release, which adds March 2026 and revises earlier years, is due on October 28, 2026.

Every survival number on this page is our own calculation from the tables. For each group we took the share still open by years since opening, then the mean, lowest and highest. As a check, the 2013 group’s 34.7 percent after ten years matches the BLS’s own figure.

Founders can add their company to TheCompanyIndex.com for free, and a person reads every entry before it goes live.

What this page could not check

• The tables follow only businesses with paid staff. About four in five US businesses had none in 2023, and no survival table here covers them.
• No table separates companies built to grow fast from other new businesses.
• A closing is defined by employment alone, so the tables cannot tell a failure from a retirement or a planned exit.
• The Small Business Administration’s survival figures sit in PDF files that could not be read as text, so none of them appear here.
• The 2011 Startup Genome report is a PDF that could not be read as text either. Whether it prints the 90 percent is unchecked.
• Only the abstract of Headd’s 2003 paper was read. The Harvard article does not describe the data behind Ghosh’s figures.
• The BLS revises these tables every year. Each number here is due for a check after its release of October 28, 2026.

Sources

• US Bureau of Labor Statistics, Business Employment Dynamics, “Table 7. Survival of private sector establishments by opening year”, total private and 19 industries, data to March 2025, read October 24, 2026.
• US Bureau of Labor Statistics, “Frequently Asked Questions” on establishment age and survival, last modified February 16, 2022.
• US Bureau of Labor Statistics, “Business Employment Dynamics, First Quarter 2025”, news release USDL-25-1582, December 17, 2025.
• US Bureau of Labor Statistics, Business Employment Dynamics home page, notice of the release of October 28, 2026, read October 24, 2026.
• US Bureau of Labor Statistics, The Economics Daily, “34.7 percent of business establishments born in 2013 were still operating in 2023”, January 12, 2024.
• US Bureau of Labor Statistics, “Entrepreneurship and the U.S. Economy”, last modified April 28, 2016.
• US Census Bureau, Business Dynamics Statistics 2023, firm age table, released September 25, 2025; “BDS Methodology”, revised June 18, 2026; “About this Program”.
• US Census Bureau, Travis Shoemaker, “Number of U.S. Nonemployers Grew Faster Than Employer Businesses Nearly Every Year From 2012 to 2023”, July 30, 2025.
• US Small Business Administration, Office of Advocacy, “Frequently Asked Questions About Small Business 2026”, February 3, 2026.
• Carmen Nobel, “Why Companies Fail”, on the research of Shikhar Ghosh, Harvard Business School Working Knowledge, March 7, 2011.
• Steve Blank, “The Startup Genome Compass”, blog post, August 29, 2011.
• Startup Genome, “Premature Scaling: A Deep Dive”, September 2, 2011.
• Brian Headd, “Redefining Business Success: Distinguishing Between Closure and Failure”, Small Business Economics, volume 21, issue 1, August 2003.
• Timothy Bates, “Analysis of Young Small Firms That Have Closed: Delineating Successful from Unsuccessful Closures”, US Census Bureau, Center for Economic Studies, working paper CES-02-24, October 2002.