How many new businesses survive one year, five years and ten years?

United States: the Bureau of Labor Statistics table

The Bureau of Labor Statistics (BLS) follows each year's new establishments in its Business Employment Dynamics data. Its questions and answers say the data come from the reports employers file for unemployment insurance, not from a survey. The tables are on the establishment age and survival page, last modified on 17 December 2025. Table 7 for the whole private sector has a block for each year's openings since the year to March 1994, counted every March up to 2025.

The rates have moved little in three decades. In the same table the share lasting one year has stayed between 75.2% and 80.9% for all 31 cohorts from 1994 to 2024. For every cohort old enough, the share lasting five years has stayed between 45.4% and 51.9%, and the share lasting ten between 32.4% and 35.3%.

The same questions and answers list self employed workers among the main exclusions from unemployment insurance. So a founder with a live product and nobody on a payroll is not in these figures at all.

United Kingdom: the Office for National Statistics release

The Office for National Statistics (ONS) publishes Business demography, UK once a year, from the Inter Departmental Business Register, the register of businesses signed up for VAT or PAYE. The latest edition covers 2024 and was released on 20 November 2025. It follows each year's births for five years, so there is no ten year figure. The UK totals are in Table 4.1 of the dataset released with it.

The quality report ONS publishes with these statistics says the register leaves out a sizeable population of businesses with low turnover and no employees.

GOV.UK's page on registering for VAT says a business must register once its taxable turnover for the last 12 months goes over £90,000, and may register below that by choice. So a product earning less than that, run by a founder with no payroll and no voluntary registration, is not in the UK table either. A company number alone does not put it there: the quality report says ONS removes companies that have neither VAT nor PAYE.

In the same bulletin ONS marks the latest two years in the publication as provisional and subject to revision. It also warns that many businesses registered at a single postcode have affected births, deaths and survival rates.

Why the two cannot be compared line for line

Each row comes from the BLS and ONS pages quoted above, chiefly the BLS questions on age and survival and the ONS quality report. So 77.9% and 93.4% are not two readings of the same thing. A UK business born at any point in 2023 is a one year survivor if it shows any turnover or employment in any part of 2024, and the UK rate then falls fast: 68.9% of businesses born in 2022 survived two years. At five years the order reverses, 51.4% against 38.4%, and that is not evidence that American businesses last longer. One table follows the locations of employers. The other follows every registered business, including those with no staff.

What the figures do not show

• Software products. Both tables count businesses, never products, and the nearest rows are whole industries. In the BLS table for the Information sector, 71.6% of establishments lasted one year, 45.7% five and 30.0% ten, for the same three cohorts. In Table 5.2a of the ONS dataset, 34.5% of the 21,730 businesses born in 2019 under computer programming, consultancy and related activities survived five years, and 44.8% of the 525 born under software publishing.
• Why a business stopped. Neither table has a column for the reason. An establishment that goes from some employees to none is a closing in the BLS definitions, and a business with no turnover and no employment in a year has not survived in the ONS ones. An owner who retires or takes a job looks the same as one who ran out of money. Closing is not the same as failing.
• What owners said. One survey did ask. In 1996 the US Census Bureau surveyed firms that existed in 1992, and about 62% of 78,147 firm surveys came back. A working paper by Brian Headd in the Census Bureau's research series reports that, of the firms opened from 1989 to 1992 and closed by 1996, 29.1% of owners said the business was successful when it closed. It is a survey, it is thirty years old, the paper is research and not official statistics, and it warns that firms which had closed were less likely to answer.
• How the business is doing. Surviving means still employing or still trading. The ONS quality report says firms that grow or shrink are not captured, and that the death of a sole trader and of a company employing dozens are treated equally.

What to do about it

The tables cannot say whether your product will last. They do show when the risk is highest: in the BLS table, 22.1% of the establishments opened in the year to March 2024 were gone a year later, while 92.8% of those that had reached nine years made it through a tenth. A table of closures is not something you can act on. Whether strangers find your product, pay for it and stay is.

• Check the product before the channel. Will your product sell sets out the six things that decide whether a stranger pays.
• Run one channel at a time, free ones first. Distribution channels takes each from its own official rules, among them Reddit, SEO and ChatGPT and AI search.
• Hold paid ads until strangers have used the product for free. Should I run ads before anyone has paid? says when a test is worth running.
• Learn the numbers that show whether customers stay and what each one costs: retention, churn and customer acquisition cost.
• For the reasons founders give when a company closes, read Why startups fail. Those are survey reports, not official data.