Every operator I have worked with has the same conversation about third-party delivery at some point. Sales are up. Contribution is flat. The delivery line on the P&L is comfortable enough that nobody looks under it. Then somebody runs the actual math on a single order and the room gets quiet.
The math is not complicated. It is just a lot of small subtractions that hide inside three or four line items on a monthly rollup. I want to walk through a real $22 order the way I walk through it with a general manager on a Monday, and then talk about when marketplace delivery actually pays for itself and when it is quietly moving your margin from a good channel to a worse one.
The $22 order, one line at a time
Let us use a $22 order because it is close to the average marketplace ticket for a fast casual concept in most metros, and because the numbers scale cleanly. Same food, same kitchen, same guest. The only thing changing is where the order came from and how it left the building.
Fig. 1 · Every line comes out. What is left is what you keep.
Here is the same picture as a P&L snippet, the way it would sit on a weekly review sheet:
- Gross order · $22.00
- Marketplace commission at 27.5 percent · -$6.05
- Delivery packaging · -$0.75
- Cost of goods sold at 30 percent · -$6.60
- Incremental prep and expedite labor · -$1.10
- Contribution before fixed cost · $7.50 (34.1 percent)
Compare that to the same $22 ticket coming through the front door. No commission. No delivery packaging. The same $6.60 in food cost, and roughly the same labor because your line is already staffed for the shift. Contribution on the dine-in version is about $10.80, or 49 percent.
The gap is 15 points of contribution margin on the same food. Not on some other menu. On the same $22 bowl of the same thing, moving through the same kitchen.
Where the commission number actually comes from
People argue about the commission line more than any other, so it is worth being precise. There are two commission structures across DoorDash, Uber Eats, and Grubhub, and they behave very differently on your P&L.
Marketplace plans, 25 to 30 percent
The platform delivers the order for you. Their driver, their app, their support. Commission is typically 25 to 30 percent depending on tier and city, plus 1 to 5 additional points if you want the featured placement that actually generates order volume. This is the tier most operators are on because it is the tier that produces the sales line they can see.
Delivery-only or self-delivery plans, 12 to 18 percent
You take the order through the platform but you provide your own driver. Commission drops to roughly 15 percent. You now own the driver labor, insurance, and vehicle cost. For a restaurant with 20 or more delivery orders a day at peak, this often nets ahead of the marketplace plan. For a restaurant with five delivery orders on a slow Tuesday, it does not.
Pickup plans, 6 to 8 percent
The guest orders through the platform and picks it up in your restaurant. No delivery at all. Commission is 6 to 8 percent. This is the least discussed and most profitable channel on the platforms, and most operators are not featuring it because it is not defaulted on the platform side.
The move most operators should make on commission is not to leave the platform. It is to run the mix analysis and shift a portion of orders from marketplace to delivery-only or to pickup where the guest tolerance allows it. A 5-point shift in channel mix is often worth more than any menu price change.
The lines nobody puts on the delivery P&L
Packaging and incremental labor are the honest lines. Refunds and cannibalization are the ones that hide.
Refunds and chargebacks: 3 to 5 percent of gross
The marketplace platforms all run a low-friction refund process for the guest. Missing item, wrong item, cold food, late delivery. In most restaurants the refund rate on marketplace orders runs 3 to 5 percent of gross orders. Roughly half to two thirds of those refunds get billed back to the restaurant depending on the platform and the reason code.
On a $22 average ticket and a 4 percent refund rate with 60 percent restaurant liability, you are giving back about $0.53 per order in refund cost. That is another 2.4 points of margin, and it does not show up in your commission line. It shows up in a separate reconciliation report that most general managers do not open.
Cannibalization: 30 to 50 percent
This is the one operators do not want to run because the answer is uncomfortable. When a guest orders through DoorDash instead of walking into the restaurant, some of those orders are net new and some of them are just channel shift. Honest measurements across three concepts I have worked with put the cannibalization rate at 30 to 50 percent.
Delivery is not additive if half of it would have walked in. It is a margin transfer from a 49 percent contribution channel to a 34 percent contribution channel, dressed up as growth.
You cannot measure cannibalization perfectly. You can measure it well enough. Run a 30-day window before you turn on a new platform, compare dine-in traffic to a matched 30-day window after. If dine-in dropped 20 percent while delivery added 40 percent of previous dine-in volume, you just gave up 20 percent of your best channel to grow your worst one.
When third-party actually pays
None of this is an argument for pulling delivery. Delivery earns its place on the P&L when it is doing work no other channel can do. Three cases hold up consistently.
Off-peak fixed-cost coverage
Your rent, utilities, and salaried kitchen labor are running from open to close whether you have a guest or not. Between 2pm and 5pm most casual restaurants are burning fixed cost. A delivery order in that window at 34 percent contribution is pure absorption. That $7.50 of contribution is paying rent that would otherwise get paid by nobody.
Cuisine or format that does not serve dine-in well
Wing shops, poke, salad, and certain ghost brands were built for delivery. Their kitchen throughput, packaging, and menu design assume the order is leaving the building. For these formats delivery is not the low-margin channel. It is the channel.
New market entry or trial
When you open a new unit or launch a new concept, the platforms are a discovery mechanism. You are paying commission to acquire guests you can then convert to your direct channels through packaging inserts, loyalty tie-ins, and offer codes. This works only if you are running the direct-channel conversion play. If you are not, you are paying acquisition cost forever.
The Hana test we ran across 21 units
At Hana Group, running 21 franchise units inside Walmart, Sam's Club, Whole Foods, and Target across six states, we tested a delivery configuration change on eight units for a quarter. We moved half of platform volume from marketplace to pickup and delivery-only tiers, and we cut two items with refund rates above 7 percent. Blended commission dropped from 25.8 percent to 20.2 percent. Refund liability dropped by roughly a third. Contribution on the delivery channel lifted 4.1 points. Top-line delivery sales dropped 6 percent because we lost some of the featured placement volume, but the units kept the contribution dollars because the mix was cleaner. That is the shape of a delivery decision that actually works. You are trading gross sales for contribution dollars, on purpose, with numbers.
When third-party quietly hurts you
The mirror of that is the situations where delivery is actively subtracting from the P&L, even though the top-line sales look fine.
- Peak-hour delivery that adds incremental labor. If you are hiring a second expo or adding line hours because delivery volume is stacking on top of dine-in volume between 6pm and 8pm, that incremental labor cost usually eats the entire 34 percent contribution.
- Delivery orders that displace dine-in tables. When your dining room is at capacity and delivery orders are getting fired ahead of tables, you are literally converting 49 percent margin traffic into 34 percent margin traffic in real time.
- Menu items that do not travel. Items that arrive at temperature or texture the guest complains about drive the refund rate on their category to 8 or 10 percent, which turns the math negative on that item.
- Refund reason codes you are not auditing. A general manager who is not reading the weekly reason code report is not catching the delivery driver who is stealing items or the kitchen station that is missing sides.
What the fix actually looks like on a Monday
When I sit down with a group that has a delivery mix problem, the first thing that goes on the whiteboard is not a strategy. It is a channel-level contribution table. Every channel, every commission tier, every packaging cost, every incremental labor assumption, in one view. Once the operator can see the 15-point contribution gap on a single sheet, the conversation changes on its own.
The second thing is a menu-level filter. Delivery menu is not the same menu as dine-in. It is smaller, priced 8 to 12 percent higher, weighted toward items that travel, and it removes the two or three items whose refund rate is dragging the whole channel down. This alone recovers 2 to 4 points of contribution on the delivery channel in a quarter, without pulling off any platform.
The third thing is a channel-mix target. Not a directive. A target. If pickup can grow from 8 percent of platform orders to 20 percent, and delivery-only can grow from 10 percent to 25 percent, the blended commission drops from 26 percent to 20 percent. That is 130 basis points of contribution recovered on the same volume.
The point
Third-party delivery is neither the villain nor the growth engine most conversations make it out to be. It is a channel with a specific cost structure, and it earns its place on your P&L when it is covering fixed cost you were going to eat anyway. It steals from your P&L when it is crowding out higher-contribution channels or driving incremental peak-hour labor you would not otherwise pay.
Run the $22 order on your own numbers. Not the industry averages. Your commission tier, your packaging cost, your food cost, your incremental labor. Then run the same order through your dine-in P&L. The gap is the real cost of the channel, and it is the number the general manager and the owner should both know from memory. Every other conversation about delivery gets easier once that number is on the table.