Case Study 1: The Statistic That Ate an Industry

How "90% of restaurants fail in the first year" became universal knowledge without ever being true

Background

There is no other industry in which a completely unsupported number is quoted as confidently, as often, or by as many credentialed people as "ninety percent of restaurants fail in the first year."

You will hear it from bankers who are about to decline your loan. You will read it in the opening paragraph of business-press articles about restaurant openings. You will hear it from culinary instructors, from landlords, from your relatives when you tell them what you are planning, and — this is the part that ought to be alarming — from restaurant consultants who charge money for advice.

It is usually attributed to research at a large American university. When people go looking for that research, they find nothing. What they find instead is a citation chain: an article citing a book citing an article citing "a study," with the trail terminating in air. Researchers who have worked on restaurant survival have publicly noted that they cannot locate the source either, and that no study they are aware of supports the figure.

The actual research

The best-known empirical work on this question comes from H.G. Parsa and colleagues, published through Cornell. Rather than repeating folklore, they did the unglamorous thing: identified restaurant openings in a defined market, tracked them over a multi-year window, and counted what happened.

What they found, in round terms:

• Roughly 26–27% of restaurants did not survive their first year.
• Roughly 60% were gone within three years.
• Failure was defined to include closure or change of ownership — a definition that captures founders selling a going concern or retiring alongside genuine business failures.

Subsequent work in the same vein has produced figures in similar territory, varying with market, period, and definition. The honest summary is: about a quarter in year one, approaching six in ten by year three, with meaningful uncertainty around both.

That is a hard business. It is not a 90% business.

Why the myth persisted

Four mechanisms kept it alive, and each one is instructive.

It confirms an existing belief. Everyone already knows restaurants are risky. A statistic that agrees with a widely held prior is not scrutinized; it is nodded at. This is a general hazard in an industry with limited public data, and it will bite you elsewhere — on "acceptable" food cost, on what a location "should" produce, on what staff "expect" to be paid.

It is useful to the people repeating it. A lender declining an application, a landlord justifying a higher security deposit, or a consultant selling a survival package all benefit from a scarier number. Nobody was lying; the figure simply cost nothing to repeat and paid something to whoever repeated it.

Nobody has an incentive to check. Restaurants that fail do not publish post-mortems. There is no central registry. The data is genuinely expensive to collect — which is precisely why Parsa's work is notable and why so little of it exists.

The correction sounds like special pleading. An operator who says "actually it's only 26%" sounds like someone talking themselves into a bad decision. This is the trap, and it is why the framing in Chapter 1 matters so much: the correct response is not "it's less risky than you think." It is "the risk has a different shape than you think, and the shape is the actionable part."

What it shows

The shape of the risk is the whole lesson.

A 90% year-one failure rate would describe a business whose outcome is essentially random — struck by lightning. If nine of ten competent operators fail immediately, skill is not the operative variable and there is no point building systems.

The real curve says something entirely different. About 26 of 100 restaurants are lost in year one. Roughly 34 more are lost across years two and three. Most of the casualties are businesses that worked, at least for a while — that opened, served guests, got reviews, hired people, and then declined over eighteen to thirty months.

That is a bleeding pattern. Bleeding is:

• Slow — which means there is a window in which it is diagnosable.
• Countable — which means measurement finds it.
• Caused — which means it has mechanisms, and mechanisms have countermeasures.

Every technique in this book depends on that being true. If restaurant failure were random, weekly inventory counts would be superstition.

Outcome

The myth is still in circulation and probably always will be. But its correction has become reasonably well established in the trade press and in hospitality education, and the practical effect has been a shift in emphasis: away from "restaurants are a gamble" and toward the specific, identifiable operating failures that actually close them — cost control, labor management, and cash.

That shift is what made this book possible to write. A book about how to manage a coin flip would be short.

Lesson

Check the number. This is the discipline the case is really teaching, and it applies to far more than failure rates.

The restaurant industry is dense with confidently repeated figures that nobody has verified: what food cost "should" be, what a square foot "should" produce, what turnover "always" is, what a delivery platform's commission "really" costs you. Some are roughly right. Several are badly wrong. Almost none of them are right for your restaurant, which has a specific menu, a specific labor market, a specific rent, and a specific guest.

An operator's job is not to memorize benchmarks. It is to measure their own business and compare. Benchmarks tell you where to look. Only your own numbers tell you what is true.

Discussion questions

1. The chapter argues that correcting the 90% figure makes the situation more alarming rather than less. Reconstruct that argument in your own words. Do you find it persuasive?

2. Identify three other numbers commonly quoted in the restaurant industry that you have heard stated as fact. For each, how would you go about verifying it — and what would "verified" even mean?

3. The Parsa definition of failure includes change of ownership. Argue both sides: why is that the right definition for studying restaurant survival, and why might it overstate failure from an operator's point of view?

4. A lender cites the 90% figure while declining your application. You know it is wrong. Does correcting them help your case? Draft what you would actually say — and consider whether the lender's underlying caution might be reasonable even when their statistic isn't.

5. Suppose the real first-year failure rate were 5% rather than 26%. Which arguments in Chapter 1 would still hold, and which would collapse?

6. This case study is about an industry believing something convenient and unverified. Where else in restaurant operations does that pattern show up — and what does it cost when it does?