What do the figures really say about new businesses, apps and startups?

The rules this shelf keeps

• No number without a link. Every figure sits in a sentence with a link to the table, release or report it was read on, and that source is listed again at the foot of the page with the day it was read.
• Primary sources only. The statistical office, the regulator, or the company reporting on itself. Never an article about a report.
• A survey is called a survey. Every time, with who ran it, how many people answered and when. A survey is never set beside an official table as if they were the same kind of thing.
• Where sources disagree, the page says so. And says why: different countries, different definitions, different years.
• No source, no claim. A figure that could not be traced to a primary source was left out, however often it is repeated elsewhere.

Five figures, and who counted each

The five cannot be set against each other. Two are official tables for two countries, two are companies reporting on their own platforms, and one is a count of what founders wrote after closing. Each page says what its figure does and does not count.

How to read a startup statistic

Four questions take most of the air out of a frightening number.

1. Who counted? A statistical office counting registrations, or a company counting the essays it could find.
2. What was counted? A business, an establishment, a venture backed startup and an app are four different things. The official tables count the first two, in every industry.
3. What counts as failing? In the official tables a business that closed has only stopped trading or employing. An owner who retired or took a job looks the same as one who ran out of money.
4. When, and where? A figure for one country and one year is often quoted as if it were a law of nature.

Take the best known one. No table on this shelf shows that nine startups in ten fail. The Bureau of Labor Statistics table has 34.7% of establishments opened in the year to March 2015 still open ten years later. And the widely quoted reasons come from counts such as CB Insights' report of March 2026, which read the public accounts of 385 venture backed shutdowns, not from any official record. Business survival rates and Why startups fail have both in full.

What to do about it

A table of closures is not something you can act on. Whether strangers find your product, pay for it and stay is.

• Check whether the product can sell before you market it: will my app make money and product market fit.
• Pick one channel where your buyers already are: distribution channels takes each one from its own official rules.
• Learn the handful of terms that show whether customers arrive, pay and stay: marketing basics.
• See the ways of getting help side by side, with what each costs: compare.

Every page of figures