Delivery vs dine-in menu: the statistics that decide which dish stays and which one goes

The food cost you track is not the cost of the channel

A dish running at 31% food cost does NOT cost the same in the dining room as it does in the app, and that gap decides whether the month closes in the black. Median food cost in full service reached 32.0% of sales in 2024, according to the National Restaurant Association's Restaurant Operations Report 2025, with 33.7% among houses under two million dollars in sales and 31.0% among those above that line. Those numbers describe the kitchen, not the channel. When that same lasagna priced at 42,000 pesos leaves through a marketplace charging 24% commission, with double-lid packaging running close to 5% and a 15% promotion the app demands to place it in the front carousel, gross margin collapses from roughly seventy cents on the dollar to barely over thirty. The decision these figures trigger together is simple: stop auditing dishes and start auditing routes to sale.

How much contribution margin actually survives per app order?

Between 25 and 35 cents of every dollar billed survive, against the 68 to 70 cents that same dish leaves when it goes out to a table.

The arithmetic allows no argument: start from the 32% median food cost the National Restaurant Association reports for 2024, subtract marketplace commission —the platforms' public range runs from 15% to 30%—, add 4% to 6% for packaging plus the promotional discount the app uses to buy you visibility, and the result lands near 31%. That Bogotá Italian sold 380 lasagnas a month through the app and kept 3,100 pesos per unit; by month's end, 1,178,000 pesos of contribution to justify a kitchen running flat out every Friday. There is no sales problem there. There is a pricing-by-channel problem. One dish can be a star in the dining room and a dog in the app at the same time, and the Kasavana and Smith matrix will tell you so only if you run it separately for each channel.

Menu engineering gets run twice, never once

The original logic ranks dishes by popularity and contribution margin in currency, and forty years later it remains the right tool; the mistake is feeding it a single data set when the cost structure has already split in two. With food cost by concept at 25% to 30% in quick service, 30% to 34% in casual and 34% to 40% in fine dining —the ranges the National Restaurant Association publishes—, a casual dining house pushing its 34% dish into delivery against 24% commission is giving product away. At Masterestaurant, Diego F. Parra insists on that double run before anyone touches a single price: the channel map first, the menu after. The pricing psychology that works on a printed menu falls apart inside the app, because there the anchor belongs to the market rather than to you. On paper you decide the typeface, the reading order, the expensive dish placed high as a reference and the removal of the currency symbol —tricks with decades of revenue management evidence behind them—.

On screen your price sits in a grid you do not control

In the marketplace your lasagna shows up in a grid beside nine competitors, sorted by an algorithm that rewards conversion and prep time, with a square photo and the price in plain sight. That context punishes high-ticket plates and rewards anything that reads as volume. The operational consequence is not raising everything 20% blindly, but choosing which dishes deserve screen space: low food cost with mid-to-high ticket, pizza being the canonical example —Sauce puts its food cost between 15% and 20% of menu price in 2025—. Two demand shifts are rewriting what belongs in the app, and neither one is about price. Seventy percent of Americans want to eat more protein in 2025, nearly twenty points above three years ago, according to the International Food Information Council's Food & Health Survey; alongside that, 49% plan to drink less alcohol this year, up 44% from 2023, per NCSolutions.

What the 2025 diner wants and your delivery menu still does not carry?

Translate it: the margin that used to arrive through the bottle of wine at table does not travel to a doorstep, and protein, your most expensive input, is exactly what the customer is hunting.

Food and beverage spending grew 3% year over year in the first half of 2025, according to Circana, so the money is there. The joint decision: build protein bowls and plates with cheap sides and make your margin there, not on drinks that delivery takes away from you. Switching suppliers was the number one strategy operators chose against rising costs —40% picked it in 2024, according to TouchBistro—, and even so it is the wrong lever for delivery. Negotiating two points off purchasing against a 32% food cost hands you back 0.64 margin points; repricing the digital menu to absorb a 24% commission hands you back ten times that. I got this wrong for years by recommending purchasing first, because it is what the owner already knows how to do and it delivers a fast win in week one.

Buying better will not fix a badly priced channel

But purchasing has a ceiling and the channel does not: cash flow remains the leading cause of financial stress and closure among small businesses, according to Inc., and a badly priced delivery channel drains cash every single day with the kitchen full. Fix the channel price first, then go back to the supplier list. If your app volume doubled tomorrow without touching the menu, you would fail faster, and that is the paradox growth keeps hidden. Go back to the Italian: 380 lasagnas a month at 3,100 pesos of contribution barely cover a slice of kitchen payroll. Take it to 760 and you will need another cook, a second packing station and more line hours, with fixed costs that climb in steps while unit contribution stays nailed in place. With optimal food cost at 28% to 35% —the National Restaurant Association range— and a 24% commission, every extra order eats capacity and returns crumbs.

What would happen if your delivery orders doubled tomorrow?

Growth only pays when contribution margin per order covers the marginal cost of producing it, and in a badly priced delivery channel it does not.

That is the tension: the channel handing you volume is the same one eating your cash. 32.0% median food cost in full service (National Restaurant Association, 2024): treat it as your baseline and calculate what is left AFTER commission, packaging and promotion, never before. Action: open your contribution-by-channel matrix this week. 15% to 30% marketplace commission (platforms' public range): it is the largest variable cost on your digital menu and it appears in no recipe card. Action: set delivery prices so contribution in currency matches the dining room's, even if the price climbs 18% to 25%. Forty percent of operators switched suppliers in 2024 (TouchBistro): most of them were optimizing the small cost. Action: before requesting a new quote, pull from the app the five dishes whose contribution per order sits below 25% of price and keep the menu that actually pays.

Where the two menus genuinely diverge?

Food cost is identical, the cost structure is NOT. A 30% food cost dish reaches the dining room with about 70 cents of gross margin per dollar;

that same dish on a marketplace charging 24% commission, 5% packaging and a 10% promotion lands at 31 cents. The decision lever is not the food cost percentage but contribution margin in currency, per dish and per channel. Price psychology works differently on a screen. On a printed menu the price sits inside a composition you control: typography, reading order, a high anchor up top, no currency sign. On the app that price sits in a grid beside nine competitors, sorted by an algorithm you do not control, and the market sets the anchor. That changes which dish belongs in which channel. Restaurant menu design in the room aims to steer the eye; in delivery it aims to survive a thumbnail. A mediocre photo in a 320-pixel tile sinks a profitable dish, while on the printed menu that same dish defends itself with written description.

Where the two menus genuinely diverge — in practice?

The sales mix decouples. When a restaurant measures both channels together, the blended mix hides that the digital channel concentrates the dishes with the weakest marginal profitability per dish.

Split the reports and four to eight references usually appear that must leave the app without touching the room. Delivery has no human suggestive selling. In the room a trained server lifts average ticket with a starter or dessert; on the app that job belongs to menu architecture —order, bundles, add-ons— and if nobody designed it, it simply does not happen.