Delivery Is a Second Business. Price It Like One.

Ask an operator what third-party delivery costs and you’ll get a number: fifteen, twenty-five, thirty percent, depending on the tier they signed. That number is accurate and it is not the answer. Restaurant third-party delivery cost is the sum of a commission you can quote and four costs you probably can’t — packaging, refunds and remakes, the expo minutes pulled out of your dine-in service, and the price you chose not to raise on the app. Add them up honestly and delivery stops looking like incremental revenue on your existing P&L. It looks like what it is: a second business with different unit economics, running through a kitchen you already paid for.

Delivery is a second business running through the same kitchen

The accounting habit that causes the damage is treating delivery as a line inside the restaurant’s P&L instead of a channel with its own. When delivery sales land in the same revenue bucket as the dining room, the blended numbers stay comfortable: food cost holds, sales are up, and nobody asks which channel earned what. Run the two apart and the picture separates fast. A dine-in cover and a delivery order can carry identical food and produce contribution margins twenty or thirty points apart, because the delivery order pays a commission, wears packaging, absorbs a share of refunds, and consumes kitchen labor the dining room needed at exactly the same moment. Same food, same building, two different businesses. The operators who make delivery work are the ones who model it that way before they scale it, not after a slow quarter forces the question.

What does third-party delivery actually cost a restaurant?

Start with the visible half, because it’s the only half most operators can quote. The major marketplaces publish tiered commission structures on their own merchant pricing pages, and in the United States the standard marketplace tiers have settled into roughly the same shape: about 15% at the entry tier, about 25% in the middle, and about 30% at the top, with pickup orders running far lower — around 6%. The tiers buy visibility and delivery radius, not better food handling. Rates vary by market and by negotiated agreement, and a handful of cities have legislated caps — New York’s, at 15% for delivery plus 5% for other services, is the most cited and has been contested in court — so check your own market and your own signed agreement rather than the headline number.

What matters more is that almost no operator actually pays the headline rate. The effective rate is the one to compute, and it’s a single division: total platform fees for the period divided by gross platform sales for the same period. Promotional funding, sponsored-listing spend, and the discounts you agreed to underwrite all settle on the same statement as the commission. An operator on a 15% tier who runs two promotions a quarter and buys placement during slow weeks can land north of 25% effective without ever changing plans. That number is knowable this afternoon, from statements you already have.

The arithmetic on one $60 order

Worked once, all the way through, so you can rerun it with your own inputs. Take an illustrative full-service unit with food cost at 31% of sales — mid-range for full service — card processing at 3%, packaging that runs $2.25 for a bagged and sealed delivery order, and a fully loaded kitchen wage of $20 an hour. Assume three of every hundred delivery orders end in a refund or remake the restaurant absorbs. Now run the same $60 of menu items through two channels.

Dine-in first. A $60 check carries $18.60 of food at 31%, and $1.80 of card processing at 3%. Sixty dollars minus $18.60 minus $1.80 leaves $39.60 — a contribution margin of 66.0% before any fixed cost. Hold that number, because it’s what the same food earns walking out the front door.

Now the same $60 through a marketplace at the 30% tier, priced at parity with the dining room. Commission takes $18.00, so $42.00 is remitted to you. The food is identical: $18.60. Packaging adds $2.25. The refund allowance is three percent of the $42.00 you would have kept, or $1.26, spread across every order. Packing, bagging, sealing, labeling, and staging runs four minutes of kitchen labor at $20 an hour, which is $1.33. Work it down: $42.00 minus $18.60 is $23.40; minus $2.25 is $21.15; minus $1.26 is $19.89; minus $1.33 is $18.56. Contribution margin: 30.9%. The same $60 of food that earned $39.60 in the dining room earned $18.56 on the app — $21.04 less per order, a 53% haircut, with no card fee to blame it on.

Now change one decision. Mark the delivery menu up 15% and the order becomes $69.00. Commission at 30% takes $20.70, remitting $48.30. Food cost does not move, because it’s the same food: $18.60. Packaging stays $2.25. The refund allowance becomes 3% of $48.30, or $1.45. Labor stays $1.33. That leaves $48.30 minus $18.60 minus $2.25 minus $1.45 minus $1.33, or $24.67. The markup added $6.11 of contribution per order — a 33% improvement from a pricing decision that costs nothing to implement. It also leaves delivery $14.93 per order behind the dining room. Both facts are true, and operators tend to believe only one of them at a time.

Contribution margin, channel by channel

Run every channel through the same arithmetic and the ranking holds across essentially every full-service model we’ve built this for in engagements we run. The table uses the same illustrative inputs throughout — $60 of menu items, 31% food cost, $2.25 of packaging on anything that leaves in a bag, four minutes of pack labor at $20 an hour on those same orders, and a 3% refund allowance on platform orders. Card processing at 3% applies where you settle the payment yourself; on marketplace orders it already sits inside the commission.

Two things fall out of that column. First, pickup is where the recoverable margin lives: moving an order from platform delivery at the top tier to platform pickup is worth $13.97 on a $60 order, and moving it to your own first-party ordering is worth $17.46 — which is why a pickup incentive printed on every delivery bag outperforms almost any promotion a platform will sell you. Second, the tier you sit on is worth $8.73 per order between 15% and 30%. If you’re on a top tier because someone signed for the visibility during a slow stretch and nobody revisited it, that’s the first conversation to have.

Is third-party delivery profitable? Incrementality decides it

Everything above is per-order. The channel-level answer depends on a question the platform dashboards will never ask you: how much of this business would have happened anyway? An order from a suburb you don’t draw from is new money. An order from the regular four blocks away who used to sit at your bar on Tuesday is the same money, routed through a 30% toll. The first grows the business. The second shrinks it while the top line goes up.

Annualize it on the same illustrative unit. Say it does $3M a year and delivery is 8% of sales — $240,000 of menu value, at an average $60 order, which is 4,000 orders. At the parity contribution of $18.56, the channel produces $74,240. That’s the number a dashboard would show you, and it looks like a good business. Now net out cannibalization. Assume 30% of those orders — 1,200 of them — would have come through the door anyway. As dine-in, those 1,200 covers would have contributed $39.60 each, or $47,520. Subtract: $74,240 minus $47,520 leaves $26,720 of genuinely new contribution. The channel booked $240,000 in sales and delivered $26,720 in new money — $6.68 per delivery order, and about 11 cents on every delivery dollar. Still positive. Nothing like what the revenue line implies.

Change the cannibalization assumption and the conclusion changes with it, which is the point. At 10% the channel throws off $58,400 of new contribution and deserves more investment. At 50% it destroys $4,960 a year while adding $240,000 to the top line. On these inputs the break-even sits near 47% — below it delivery builds contribution, above it delivery buys revenue with margin. Nobody can hand you that percentage. You estimate it from your own data: delivery orders by daypart against dine-in covers by daypart, delivery ZIP codes against your trade area, and how many delivery guests also appear in your dine-in loyalty data. An estimate you built beats a number you assumed.

Pull these before you renegotiate

Most operators enter a platform conversation with one number — the tier — and leave with a slightly better version of the same deal. The leverage is in arriving with the channel’s whole P&L. Every line below already exists in a settlement statement, an invoice file, or a POS export you own.

• —Delivery net sales by platform, twelve months, broken out from dine-in. If the two are blended into one revenue line, nothing below is computable.
• —Effective commission rate: total platform fees divided by gross platform sales, per platform. Compare it against the tier you think you’re on.
• —Promotional and sponsored-listing spend, listed separately from commission. This is the line that turns a 15% tier into a 25% reality.
• —Refund and adjustment dollars per platform, as a percentage of gross platform sales — from the settlement statements, not the dashboard summary.
• —Packaging cost per delivery order: twelve months of container, bag, sauce-cup, and label spend divided by delivery order count. Most operators have never computed this one.
• —Minutes per order to pack and stage, timed on a real Friday rather than estimated on a Tuesday. Multiply by your fully loaded kitchen rate.
• —Delivery orders by daypart against dine-in covers by daypart. Where the two peaks overlap, you are paying one expo to serve two businesses.
• —Delivery menu prices against dine-in prices, item by item, with the date each was last changed.
• —Your best available read on incrementality — delivery ZIP codes against your trade area, and the share of delivery guests who also appear in dine-in loyalty data.

Bring that page to the conversation. It moves the discussion off a percentage point of commission and onto the real question — whether the channel earns its place in your kitchen — and platforms behave differently with an operator who can show the arithmetic than with one asking for a discount.

Decide it as a business, not a default

Delivery earned its way into most restaurants during a period when nobody had the luxury of running the numbers, and a lot of operators have never gone back and run them. That’s the whole problem. Delivery isn’t automatically wrong; on the arithmetic above it’s often modestly right, and for some concepts and some trade areas it’s clearly right. What it never is, is free — and it is never simply additive. It’s a second business with its own commission structure, its own packaging cost, its own defect rate, its own labor draw, and its own pricing decision, running through a kitchen sized and staffed for the first one. Model it separately, price it deliberately, and watch its contribution the way you watch prime cost, and it can be a real line of business. Blend it into the dining room’s P&L and you’ll grow revenue for years without ever finding out what it cost you.