Delivery is no longer a side channel you bolt onto a working dining room. Three out of every four restaurant orders are now eaten somewhere other than your dining room, according to the National Restaurant Association’s May 2026 comments to the Federal Trade Commission (National Restaurant Association, retrieved 2026-09-26). For most operators, off-premise is the business, and the dining room is the part that got smaller.
The trouble is that delivery sales and delivery profit are two different things, and a P&L doesn’t separate them for you. A $60 order that arrives through a marketplace app and a $60 order rung up at the counter look identical in your sales report. One of them has already handed away $15 to $20 before you’ve touched a single ingredient.
Running a restaurant delivery operation profitably comes down to a short list of decisions you make once and then defend: what you’ll deliver, how far, who drives, how orders move through the kitchen, and which numbers you look at on Monday morning. Here are ten that decide the outcome. Where delivery intersects with the wider shifts in the business, our read on the restaurant industry trends that matter most in 2026 has the demand and labor picture behind these choices.
The Bottom Line
- Marketplace commissions run 15% to 30% of the ticket, and Rezku’s 2026 fee breakdown puts the effective all-in cost at 30% to 40% once processing, promotions and adjustments land. That is the entire net margin on a typical restaurant order.
- Set your zone in drive-time minutes rather than miles. Hot food degrades noticeably past the 25-minute mark, and the outer ring of a zone produces measurably weaker repeat business.
- Around 70% of U.S. restaurants now run at least some deliveries with their own drivers, usually as one leg of a mix rather than an all-or-nothing choice.
- Most kitchens lose delivery money on workflow, not on fees: unpacked orders, drivers waiting, and a hot bag that sat under a heat lamp for eleven minutes.
- The number to manage is contribution margin per delivered order. Order count and delivery revenue will both look great while you lose money.
Save 80% of delivery management time
We handle everything:
- Dedicated operations manager
- Real-time tracking dashboard
- Automated customer notifications
- Urgent issue resolution
1. Decide which menu items are allowed to travel
Menu triage is the cheapest profit lever in delivery, and almost nobody pulls it. Every dish on your menu has a travel tolerance, and the ones that fail arrive as refund requests.
Run your menu through three questions. Does it survive twenty minutes in a sealed container? Does it plate itself, or does it need a server’s hands? Is the margin high enough to absorb a packing cost and a delivery fee? A pasta dish that loses its texture in fifteen minutes does not belong in a delivery menu just because it sells well at table six.
Cutting items feels like losing sales. It usually isn’t. A tighter delivery menu speeds up your line, reduces the number of containers you stock, and removes the dishes that generate most of your complaints. Keep the full menu for the dining room and publish a shorter one for delivery. Customers will not notice; your refund rate will.
2. Draw your delivery zone in minutes, not miles
A radius is a circle on a map. A drive time is what happens to the food. A three-mile radius that crosses a bridge, a stadium or a one-way grid is not the same operation as three miles of suburban arterial.
Food type sets the ceiling. Pizza and hot entrees hold up for roughly 15 to 25 minutes of transit, non-pizza restaurant orders for 20 to 30, and sandwich, deli and barbecue formats tolerate 30 to 40 because the food survives the ride (Foxifood, retrieved 2026-09-26). Sushi and anything plated with sauce sit at the bottom of that range.
The other reason to keep the zone tight is repeat business. RadiusMapper’s analysis of zone performance found customers in the outer ring, past roughly 22 minutes of drive time, reordered at a 14% lower rate than inner-ring customers even when their satisfaction scores on the individual order looked similar (RadiusMapper, retrieved 2026-09-26). The far edge of your map buys you one order, not a customer.
| Zone | Drive time | What it’s good for |
|---|---|---|
| Inner | 0 to 10 minutes | Everything on the menu; best food quality and lowest cost per stop |
| Middle | 10 to 18 minutes | Profitable for most formats at normal ticket sizes |
| Outer | 18 to 30 minutes | Only with a minimum order, a real fee, or travel-tolerant food |
Review the map monthly with one question: which zones are we underwater on? Then shrink, raise the minimum, or raise the fee.
3. Know what third-party delivery commissions really cost you
Marketplace apps charge a headline commission of 15% to 30% depending on the plan you pick. DoorDash runs 15%, 25% or 30% by tier plus roughly 6% on pickup orders; Uber Eats runs 20%, 25% or 30% plus about 7% on pickup; Grubhub layers a 15% to 25% marketing commission on top of a delivery commission and per-order processing (Rezku, retrieved 2026-09-26).
The headline is not the number to plan against. Once payment processing, in-app promotions you agreed to, and refunds for missing items are counted, the same analysis puts the real cost at 30% to 40% of the order, meaning a restaurant keeps somewhere between $62 and $75 of every $100 that flows through the app.
Set that against the 3% to 5% net margin a typical full-service restaurant runs and the arithmetic is blunt: marketplace orders are a customer acquisition expense, not a profit center. That is a legitimate thing to spend money on. It is not a legitimate thing to do by accident on 60% of your volume.
4. Pick your mix of in-house drivers, marketplaces and courier partners
There is no single right delivery model, which is why the in-house-versus-third-party argument has run for a decade without resolving (QSR Magazine). Roughly 70% of U.S. restaurants now use in-house drivers for at least part of their delivery volume (TechRyde, retrieved 2026-09-26), and most of those also list on a marketplace.
| Model | Best at | Watch out for |
|---|---|---|
| Marketplace apps | Discovery and new customers | 30% to 40% effective cost; you don’t own the customer data |
| Your own drivers | Control, margin, and repeat customers at high order density | Idle paid time when density drops; hiring, insurance, vehicles |
| Courier platform | Covering peaks and thin nights without payroll | Needs enough notice and clean handoff to work well |
The practical answer for most independents is a mix: a marketplace for reach, your own ordering channel for the regulars, and an outside courier for Friday nights and catering drops. Platforms built for local food delivery, Metrobi among them, add multi-stop route optimization and let you work with the same drivers over time, which matters more for food than it does for parcels because the person carrying it learns your packing and your customers.
Whatever mix you land on, know the cost per stop for each leg of it. That single number is what lets you route an order to the cheapest channel that can still deliver it hot.
5. Give delivery orders their own lane through the kitchen
Most delivery problems are not delivery problems. They are expo problems that only become visible after the food leaves.
A delivery ticket has a different shape than a dine-in ticket. It fires all at once, it needs to finish at the same moment, and it has to be packed, checked, sealed and labeled before anyone can leave. If it shares a pass with dine-in plates, it loses every time the dining room gets busy, and the loss shows up as a driver standing in your doorway.
Three changes fix most of it. Fire delivery tickets as a complete order rather than by course. Put the packing station somewhere that isn’t the middle of the line. And create a staging shelf near the door that a driver can reach without crossing the kitchen, connected to the holding space so nothing sits out waiting to be moved.
The worst pickup experience in delivery is a driver arriving before the food is packed, which turns one late order into a late driver and then a late second order. If your building is still being drawn or gutted, these are cheap decisions to make now and expensive ones later; our restaurant construction checklist covers where the pickup counter, the staging shelf and the driver door belong on the plan.
6. Pack for a 25-minute ride, not for a two-minute walk
Packaging is the last thing your kitchen touches and the first thing the customer judges. Hot food needs to leave above 140°F to stay safe and palatable through transit, and hot and cold items should never share a bag, because one will always ruin the other (Sauce, retrieved 2026-09-26).
The mechanics are not complicated:
- Fried food needs airflow, so vented containers beat sealed ones every time.
- Sauces travel separately, in containers that survive being upside down.
- Hot bags work when they start hot; an insulated bag holds temperature, it does not create it (DoorDash Merchant Blog).
- Two bags per run, one hot and one cold, is standard practice for a reason.
- A tamper-evident seal and a printed label tell the customer the order wasn’t opened, and they cut your “missing item” claims.
One more thing pays for itself: photograph your packed dishes properly. SevenRooms’ 2026 research, drawing on 3,000 U.S. consumers and 500 operators, found that 87% of consumers have chosen what to order on a delivery app because of a photo or video (SevenRooms, retrieved 2026-09-26). The picture sells the dish; the packaging decides whether they order it twice.
7. Staff delivery peaks instead of staffing delivery hours
Delivery demand is spikier than dine-in demand. It arrives in a 90-minute wave, and the cost of covering it with a fixed schedule is paid idle time on either side of the wave.
Start by pulling your own order data by half-hour and day of week, then staff the peak and cover the shoulders differently. Options in order of cost: shift an existing employee into a packing role during the wave, add a driver shift that starts at the peak rather than the dinner hour, or hand overflow to an outside courier so you’re paying per delivery instead of per hour.
The same logic applies to dispatch. Sequencing stops by hand works up to about six or eight orders an hour, and then it becomes someone’s whole job. Routing software earns its keep at that point, particularly on multi-stop runs where the difference between a good sequence and a bad one is fifteen minutes of food quality.
8. Price delivery so the fee covers the stop
Free delivery is a marketing decision that is often made as an accounting accident. If a stop costs you $8 all-in and you charge $3.99, you have chosen to buy $4 of goodwill on every order. That can be a smart purchase, but only if you know you’re making it and know your ticket size covers it.
Three levers keep delivery pricing honest:
- A minimum order that puts the ticket above the point where the stop pays for itself.
- A fee that scales with the zone, not a single flat fee stretched across a map where costs differ by a factor of three.
- Menu pricing for delivery channels that accounts for commission, which most marketplaces now allow explicitly.
Delivery-channel pricing makes operators nervous, and it should be handled carefully: raise prices far above your dine-in menu and customers notice. A modest uplift on marketplace listings, though, is how a 25% commission stops eating the whole margin on the dish.
9. Track the four numbers that tell you whether delivery works
Delivery revenue and order count are the two most flattering numbers in the business and they tell you almost nothing. Four others tell you everything.
- Contribution margin per delivered order. Ticket minus food cost, minus packaging, minus commission or driver cost for that specific order. If it’s negative, you are paying to feed people.
- Cost per stop by channel. What a delivery costs through each of your legs, including paid idle time.
- On-time rate. The percentage of orders that arrive inside your promise. On-time performance is the number customers experience, and it’s the one that predicts reorders.
- Refund and remake rate, by menu item. This is your menu triage list writing itself.
Review these monthly against the zone map. Almost every operator who does this for the first time finds one zone, one channel and one menu item that are funding the whole problem.
10. Convert marketplace customers into direct customers
If marketplaces are your acquisition channel, then the whole point is to not pay for the same customer twice. That means moving repeat buyers onto your own ordering path.
The structural difficulty is that most restaurants can’t see who those people are. SevenRooms found that four in five operators cannot recognize the same customer across their on-premise and off-premise interactions, while 64% of consumers said they’d prefer a single app for delivery, pickup and reservations (SevenRooms, retrieved 2026-09-26). The gap between those two facts is where repeat revenue goes missing.
Practical moves, cheapest first: put a card in every bag with a reason to order direct next time, make your own online ordering as fast as the app’s, offer something on your channel you don’t offer on theirs, and capture an email or phone number at checkout so you can reach a customer without paying a toll. A direct order at 5% payment processing instead of 30% commission is the same food and roughly five times the margin.
Frequently asked questions
Is restaurant delivery profitable?
It is profitable per order when the contribution margin after food, packaging and delivery cost is positive, and unprofitable when it isn’t, which is entirely determined by channel mix, zone size and ticket value. Marketplace-only delivery at 30% to 40% effective cost rarely clears a typical 3% to 5% restaurant net margin without a price adjustment on the channel.
Should a restaurant use its own drivers or a delivery app?
Most use both, and roughly 70% of U.S. restaurants now run at least some volume with their own drivers. Apps buy discovery, own drivers protect margin once order density is high enough to keep them busy, and a courier partner covers peaks without adding payroll.
How far should a restaurant deliver?
Measure in drive time, not miles. Hot entrees hold up for about 15 to 25 minutes in transit and travel-tolerant formats like deli and barbecue stretch to 30 to 40. Beyond roughly 22 minutes, repeat order rates fall off even when the individual delivery went fine.
What’s the most common mistake in restaurant delivery operations?
Treating delivery as a sales channel instead of a production process. The money is usually lost inside the kitchen, in orders that get packed late, sit under a lamp, or go out in the wrong container, rather than in the commission line everyone argues about.
