What it actually costs to run a single-location restaurant in 2026, what the failure statistics really say, and why the median kitchen is already operating past the point where the numbers work.
Each one is drawn from published industry data. Where a source reports a range, the range is shown rather than the flattering end of it.
Median labour of 36.5% of sales plus median food cost of 32.0% gives a prime cost of 68.5%. The band operators are advised to hold is 55% to 65%. This is not a warning about the weakest operators, it is the middle of the distribution.
42% of operators began 2026 without a profit. Against full-service net margins of 3% to 5%, there is no room in the model for an error of more than a couple of points.
The widely repeated claim is that 90% of restaurants fail in year one. The measured figure is about 17%, inside a reported range of 14% to 30%. The real attrition is slower and later, not immediate.
At three years, 61% of independents have closed against 57% of franchised units. A four point gap does not support the story that independence is the risk factor.
Third-party commissions are advertised at 15% to 30% per order and land at 30% to 40% once processing, marketing and menu adjustment are counted. That is charged against a business earning 3% to 5%.
Prime cost is food plus labour, and it is the only number in a restaurant that decides whether the rest of the model has room to work. Operators are told to hold it between 55% and 65% of sales. Above 65%, profitability becomes difficult to sustain regardless of how well everything else is run.
The figures below open on the published medians for full-service independents. Move them and watch where the line goes.
The model holds fixed costs at 28% of sales, which is the residual the published figures imply once prime cost and net margin are both accounted for. It is deliberately a simple model, and it holds sales volume constant. That last point matters: the delivery slider shows what the channel costs in margin, not what it adds in orders. The published 3% to 5% industry net margin is measured on restaurants already running delivery, so it sits closer to this model's zero-delivery case than to its opening one. The purpose here is to show how little slack sits between the median operator and a negative number, not to forecast a P&L.
The claim that 90% of restaurants close in their first year has been repeated for three decades without a source. The measured first-year rate for independent full-service restaurants is about 17%, and the National Restaurant Association puts the range at 14% to 30%. Attrition is real, but it is slower and later than the myth.
Close in the first year. Repeated everywhere, attributed to nothing.
Close in the first year, inside a reported range of 14% to 30%. About 56% are still trading at year five.
The second myth is that independence is itself the risk. At three years, one dataset puts closures at 61% of independents against 57% of franchised units. The four point gap is real but small, and it does not carry the weight the story puts on it.
Where the sources disagree, and we are not going to hide it. That 61% three-year closure figure cannot be reconciled with the separate finding that about 56% of restaurants are still trading at year five. One implies far steeper attrition than the other. The curve above follows the year-one and year-five series because they come from the same measurement basis; the three-year independent-versus-franchise comparison is reported as it stands, for what it says about the gap between the two groups rather than about the level.
Third-party platforms advertise commissions of 15% to 30% per order. The number that reaches the operator's accounts is 30% to 40%, once payment processing, promoted placement and the menu markup needed to stay competitive on the platform are counted. That charge lands on a business whose whole net margin is 3% to 5%.
One qualification, stated plainly. This is fully-absorbed costing: it charges the ticket its full share of fixed costs. On a genuinely marginal order, where the rent and the salaried kitchen are already paid for by dine-in, the ticket clears rather than loses. That is exactly the argument operators use to justify the channel, and it holds right up until third-party stops being marginal. The figures below are the reason the share matters more than the rate.
This is why the operators handling the channel well treat it as acquisition rather than revenue. The discipline reported among them is to cap third-party at 20% to 40% of order volume and move the rest onto a direct channel, where the commission does not apply.
Nothing in the data suggests survival is a matter of concept, cuisine or luck. It is a matter of holding four numbers.
Not as an aspiration. As the condition under which every other decision has room to be wrong.
Between 20% and 40% of orders, treated as customer acquisition. Beyond that share, the commission is buying volume at a loss.
About 56% reach year five. Of those, most of the attrition is already behind them. The dangerous window is years two and three, not year one.
At a 3% to 5% net margin, one point of food cost is a fifth to a third of the entire profit. Most operators do not know which of their three big numbers is the one out of line.
Every figure in this report is drawn from published industry data. Where sources disagree, the report shows the range. Where a figure is a median rather than a mean, it is labelled as one.
Disclosure. Kelvey Research is an invented publisher. This report was produced as a public example of an interactive research document, and it holds no panel, no proprietary dataset and no client. Every statistic above comes from published industry sources and is reported as found. The kitchen model on page 3 is a simplified representation, not a forecast, and it holds fixed costs at a constant 28% of sales.