How long
businesses
actually last.

What the table says

The Bureau of Labor Statistics has published the survival of US private-sector establishments by opening year since 1994. It is one public table, it costs nothing to read, and almost nobody reads it. These are the figures taken from it on 12 October 2026:

• Of the establishments that opened in the year to March 2025, 77.9% were still there one year later, 769,449 of 980,333. The cohort before them did fractionally better, at 78.2%.
• Of those that opened in the year to March 2015, 50.2% were still there five years on, 340,281 of 677,876. The cohort that opened in the year to March 2019 reached 51.5%.
• Of that same 2015 cohort, 34.7% were still there after ten years, 235,071 of 677,876.

Those are separate cohorts rather than one group followed down the page, which is why the five-year and ten-year figures share a denominator and the one-year figure does not. The series runs from March 1994 to March 2025, so every one of these cohorts can be checked against the ones either side of it.

What the series counts as a death

This is the part that never survives the retelling. The Bureau counts an establishment as dead when it is one “with no employment or zero employment reported in the third month of four consecutive quarters following the last quarter with positive employment.”

Read that again slowly. It is a definition about employment reporting stopping. It says nothing about why it stopped, and it was never built to. The Bureau does not describe these establishments as failures anywhere in the series or its methodology, and the methodology does not ask the question.

What the statistic really tells us

“Half of businesses fail within five years” is the version in circulation. What the measurement actually supports is this: about half of new businesses are still employing people five years later.

An establishment can leave that series with nothing having gone wrong at all. It was sold. It merged into another. It restructured under a different identifier. The owner retired. Or the owner let the last employee go and carried on alone, which removes you from a series about employment while leaving you very much in business.

The table does not show how many closures were genuine business failures and how many had other causes. It was never designed to answer that question.

What a base rate is good for

A base rate describes a population. It is not a forecast about you and it cannot be, because it contains no information about your trade, your market, your costs or your customers.

It is still useful as context. When someone uses a failure statistic to sell you insurance, a course, a service, or a loan, ask where the figure came from and what it counted. In this case, it came from federal payroll data that never used the word “failure.”

The other thing worth taking from it is the first-year number, which is the one that gets quoted least. Roughly four in five new establishments are still there a year later, and that figure comes out of exactly the same table as the gloomier one everybody repeats.