Anyone can fill a restaurant once.

Restaurant repeat customer rate is the number almost no independent group tracks, and it is the one that predicts whether the business is still here in three years. Covers tell you what happened. Return rate tells you what is going to happen.

Start by throwing out the statistic everyone quotes. “Nine out of ten restaurants fail in the first year” is not a research finding. It traces to a 2003 American Express commercial, and a literature review by a hospitality researcher at Ohio State found no evidence for it anywhere. The real numbers are lower — and far more useful, because of where they cluster.

The failure curve is not where you think it is

Read the middle column. Roughly a quarter of restaurants fail in year one — bad, but survivable odds. Then the number doubles by year three. The steepest part of the curve is not the opening. It is the second and third years, which is precisely the window in which the novelty traffic runs out and the only thing still bringing people through the door is that they liked it enough to come back.

That reframes what an opening actually proves. A strong first ninety days measures curiosity, and curiosity is free — a new restaurant gets it whether it deserves it or not. Everyone comes once. The classic failure pattern in this business is not a bad opening; it is a very good one followed by eighteen quiet months.

What a point of return rate is actually worth

Return rate compounds, which is why small differences in it produce large differences in revenue. If a guest returns with probability r after any given visit, the expected number of visits from each guest you acquire is 1 ÷ (1 − r). That is the whole model, and you can check it on a napkin.

Put money on it. Take a restaurant acquiring 1,000 new guests a month — 12,000 a year — at a $65 average check. At a 25% return rate that is 16,000 visits and $1,040,000. At 40% it is 20,000 visits and $1,300,000. At 55% it is 26,667 visits and $1,733,000. The same marketing, the same door count, the same number of new people walking in: fifteen points of return rate is worth $260,000 a year, and thirty points is worth roughly $693,000.

Be honest about the model. It assumes a constant return probability, which overstates the long tail slightly, because real guests decay rather than flipping a fixed coin forever. The exact totals will be a little generous. The shape is not: return rate enters the revenue line as a multiplier, and every point of it is applied to every guest you have ever acquired.

How to measure repeat visits without a loyalty app

The usual objection is that you cannot measure this without a loyalty program, and it is wrong. Most groups are already sitting on the data in two systems they pay for.

• —Your reservation platform already matches guests on phone number and email. Count distinct guests with two or more visits in a rolling ninety days, divided by distinct guests in the ninety days before that. That is your return rate, and it costs an export.
• —Your POS almost certainly tokenizes cards. The same token appearing twice is the same guest, with no app and no sign-up — this is how you capture the walk-in half of the restaurant your reservation book never sees.
• —One question at the host stand, logged as one field: “First time with us?” Imperfect and self-reported, but a trend line built from it beats an exact number you never produce.
• —Count reservations booked at the door on the way out, as a share of covers. It is the cleanest single signal in the building and nobody records it.
• —Exclude third-party delivery from the calculation. It is a different guest exhibiting different behaviour, and blending it in will flatter or wreck the number without telling you anything about the restaurant.

The second visit is the leading indicator

If you track one thing, track the thirty-day second-visit rate by monthly cohort: of the guests who came for the first time in March, what share came back within thirty days? Run it every month and you have a leading indicator that moves before sales do.

This is what makes it operationally useful rather than academic. Sales are a lagging indicator — by the time comps go soft, the guests who decided not to return made that decision months ago, and you are now looking at the consequence rather than the cause. A cohort curve tells you what a menu change, a management change, or a service slip did to the guests who experienced it, while there is still time to respond.

What actually brings them back

In order, and the order matters. Food first: nothing downstream survives food that is not worth a second trip, and no amount of atmosphere or service recovery outruns a dish a guest would not order again. This is the one that is measurable in the kitchen, and it is where a return-rate problem should always be diagnosed first.

Then the story. A guest who cannot describe your restaurant to someone else in a sentence has no reason to bring anyone, and a returning guest almost always returns with a person they are showing it to. If your concept does not resolve into a single legible idea, you are relying on individual memory instead of word of mouth.

Then whether it is fun — which operators consistently treat as unserious and which behaves like a survival factor. Fun is what makes the visit an occasion rather than a transaction, and occasions get repeated on purpose. It is also the most operational of the three: it lives in pacing, in music levels, in whether the staff look like they want to be there on a Tuesday. None of that is a personality trait. All of it is a standard someone owns.

What to do with the number once you have it

Put it on the same report as sales and prime cost, monthly, by unit. The moment it sits next to the numbers that already get reviewed, it stops being a marketing curiosity and starts being an operating metric — and the units that are quietly living on new-guest traffic become visible while there is still time to do something about it.

Then hold someone accountable for it, because an unowned number does not move. The reason return rate is rarely tracked is not that it is hard to calculate. It is that it sits between marketing, which counts new guests, and operations, which counts covers, and neither one is measured on the guest coming back a second time.