Delivery App Fees in 2026: What They Take and What Ordering Direct Costs

Most owners can price a case of chicken to the cent and cannot say what a delivery order costs them. The fee is a percentage, it comes out before the money lands, and it never arrives as an invoice anyone has to approve. So it gets treated as weather rather than as a line item, which is how a channel ends up taking a quarter of a sale without anyone deciding that it should.

We build websites and ordering pages, so read this as an interested party showing its arithmetic. Every rate below came off the seller's own pricing page, a regulator's own page or a company's own filing, read on 9 August 2026. Where a figure could only be verified from an older announcement, the date is in the sentence. Where the honest answer is that the apps are worth paying, the article says so.

The 30-second answer

Marketplace delivery costs 15% to 30% of the order. Uber Eats publishes 20% on Lite, 25% on Plus and 30% on Premium, with 7% on pickup when your in-app prices match your in-store prices and 10% when they do not. DoorDash introduced 15%, 25% and 30% tiers with 6% pickup in April 2021. Taking the same order on your own site costs payment processing: 2.5% plus $0.29 through Uber Eats Webshop, or 2.9% plus $0.30 through Stripe. On a $40 order that is $12.00 against $1.29. The marketplace fee is the price of a customer you did not have. Paying it again on their tenth order is the part worth fixing.

What the apps charge in 2026

Start with the numbers each company puts on its own page. Uber Eats publishes three marketplace packages: Lite at a 20% Marketplace Fee, Plus at 25% and Premium at 30%, with Plus and Premium opening at a 0% intro rate for 30 days. Every package carries a 7% Pickup Fee, and Uber is specific about the condition: that rate depends on proof that pricing for in-app pickup matches in-store pricing, and without it the pickup fee is 10%. Running your own drivers through the app costs a 15% self-delivery fee. Borrowing Uber's couriers for orders that came from your own channels is Uber Direct, which starts at $7.99 per delivery. Selling through your own site with Uber's Webshop costs a 2.5% order processing fee plus $0.29 per order.

DoorDash sets up the same way. When it moved to tiered pricing, Restaurant Dive reported on 27 April 2021 that Basic charges 15%, Plus 25% and Premier 30%, that pickup is 6% across all three, and that Premier carries a growth guarantee reimbursing commissions for restaurants receiving fewer than 20 orders a month. The same report noted that the Storefront product for ordering on a restaurant's own site was free apart from payment processing fees. Those are the rates the industry has quoted ever since, and they are five years old, so treat your own contract as the authority.

Two details hide inside the percentages. The fee applies to the order value, not to your profit on it, and the consumer pays their own service and delivery fees on top, which raises the total your customer sees while leaving your side of the transaction unchanged. So the menu price a delivery customer pays is rarely the menu price you set, and the app is the one setting the difference.

The increase that landed in March 2026

Rates on a channel you do not own can move by email. Uber told merchants that new US marketplace fees took effect on 11 March 2026: Lite rose to 20%, Plus stayed at 25% except on Uber One member orders, which now carry 30%, Premium stayed at 30%, and pickup went from 6% to 7%. Custom negotiated delivery rates rose by 3 percentage points, capped at 30%. Restaurant Dive covered the change on 10 March 2026, the day before it started.

Read the Lite line again, because a move from 15% to 20% is a third more fee on every order that package touches. Restaurants that wanted out had until 10 March to say so, and a submission for the alternative had to be in by 11:59pm on 6 March. That window is the real lesson here. A pricing change on a channel you rent gets announced, not negotiated, and the reply-by date is measured in days.

What a 15% cap means in New York now

New York City is where owners think they are protected, and the protection changed in 2025. The Department of Consumer and Worker Protection now lists four separate caps on what an app may charge a restaurant per online order: delivery fees capped at 15%, transaction fees at 3%, basic service fees at 5%, and enhanced service fees at 20%. Add them up and the ceiling in the city famous for its 15% cap is 43%.

The history is short. The City Council voted on 26 August 2021 to make the caps permanent, holding apps to 15% for delivery and 5% for other fees. Then, as Restaurant Business reported on 1 May 2025, the Council passed Int 762-B, which kept 15% for delivery, 3% for credit card processing and 5% for other services, and allowed an additional 20% for enhanced services on top. The bill also lets restaurants charge higher prices on the apps than in the dining room, and put marketing materials in delivery bags.

Two things follow for anyone operating under a cap. The basic package still has to exist, so the 23% version of the deal is always available to you, and buying enhanced visibility is a choice you make rather than a rate you are handed. And the permission to insert your own marketing into a delivery bag is worth more than it sounds, because it is the one moment when a customer the app owns is holding something you printed.

What the platforms actually keep

The commission is not the platform's profit, and knowing that changes how you negotiate. In its quarterly report filed on 5 August 2026, DoorDash reported Net Revenue Margin, its own measure of revenue as a percentage of marketplace order value, at 13.5% for the quarter ended 30 June 2026, on $33.1 billion of Marketplace GOV and 970 million orders. Uber, filing the same day, reported Delivery revenue of $5,245 million on Delivery Gross Bookings of $27,463 million, which is 19.1% of order value.

Those bars are not strictly comparable, and the caption says why: both platforms count consumer fees, advertising and subscriptions, both are net of courier pay, and both cover grocery and retail as well as restaurants. Read them as scale, not as a like-for-like margin. The useful conclusion is that a 30% fee funds a demand machine with couriers, discounts and marketing inside it. You are not being overcharged by a company pocketing 30 cents on the dollar. You are buying access to something expensive to run, at a price set by whoever runs it.

What the fee actually buys

Demand, and it is real. DoorDash's annual report for 2025 counts more than 56 million monthly active users in December 2025 and more than 35 million members across DashPass, Wolt+ and Deliveroo Plus at the end of the year. Marketplace order value grew from $66.8 billion in 2023 to $80.2 billion in 2024 to $102.0 billion in 2025. No independent restaurant is going to reproduce that on its own, and pretending otherwise is how bad advice gets written.

What the fee does not buy is the customer. The app holds the account, the payment method, the notification permission and the reason to open something at 7pm on a Friday. That is the asset, and it is being rented back to you at 15% to 30% every time the same person orders again. The platforms are candid about it in their own filings. DoorDash's annual report lists among its competitors "merchants that have their own online ordering platforms".

What a direct ordering page costs

Payment processing, and not much else. Uber's own Webshop, which puts ordering on your site while your listing stays in the marketplace, is 2.5% plus $0.29 an order. Stripe publishes 2.9% plus $0.30 per successful domestic card transaction with no setup fee and no monthly fee. DoorDash's Storefront was free apart from processing when it launched. On a $40 order, all three land between $1.29 and $1.46, against $8.00 on a 20% marketplace fee.

Then be honest about the rest of the bill, because a free ordering page is not a free channel. Somebody has to build the site and keep it working, which is a real recurring cost we broke into four separate bills in a previous piece. If you want delivery on your own orders you still need couriers, and Uber Direct starts at $7.99 a delivery, which on a $40 order is close to the 20% marketplace fee anyway. Direct ordering wins clearly on pickup and on delivery you drive yourself. On courier-fulfilled delivery it wins by a smaller margin than the chart suggests.

The other real cost is demand. Traffic that used to belong to the app has to come from somewhere, which usually means your search listing, your signage, your receipts and your regulars. That is slower than switching on a marketplace tier, and it compounds instead of repricing itself in March.

Why a percentage hurts more than it reads

A commission is charged on revenue and paid out of margin, and margin is thin right now. The National Restaurant Association's 2026 State of the Restaurant Industry, published 12 February 2026, projects $1.55 trillion in sales and 15.8 million jobs, and reports that 42 percent of operators said their restaurant was not profitable last year. Sixty percent reported softer customer traffic, and more than nine in ten named food, labor, insurance, energy and swipe fees as significant challenges.

We are not going to invent a margin figure for your restaurant, because the real one depends on your rent, your labor model and your menu. Do the arithmetic on your own numbers instead. Take last month's marketplace sales, apply your tier, and put the result next to your rent. Owners who have never done this are routinely surprised to find the apps are one of their largest single suppliers, sitting somewhere between the meat order and the lease.

Why leaving the apps is usually the wrong move

The advice to delete your listing is usually written by somebody selling ordering software. Marketplaces are a customer acquisition channel with the largest audience in the category, and turning one off removes discovery, not just fees. The sane version of this decision splits the two jobs: let the app do acquisition, and stop paying acquisition rates for retention.

In practice that means the first order from a new customer is worth 20% to 30%. The tenth from the same person is not, and every one of those you move to your own page is close to pure margin recovered. The measure worth tracking is not marketplace sales, and not direct sales, but the share of repeat customers who order direct. If that number is flat year over year, nothing you have built is working, no matter how good the site looks.

A sensible order to build in

Do these in sequence, because each one makes the next cheaper. This is roughly how we sequence the work for restaurants and hospitality clients.

One, put real ordering on your own site. Not a PDF menu and a phone number. An ordering page that works on a phone in under three taps, with a card field and a pickup time. This is a build, not a plugin decision, and it is the only piece that has to be right.

Two, decide your pricing position. Menu parity is simplest and honest. Marking up delivery prices is legal in New York under the 2025 bill and is common practice, but note that Uber's 7% pickup rate is conditioned on in-app pickup prices matching in-store prices, so a blanket markup can cost you the cheaper pickup tier.

Three, work the bag. Every delivery order leaves your kitchen with a paper bag stapled shut, and in New York the 2025 bill explicitly allows your marketing inside it. A card with a direct-order code is the cheapest acquisition you will ever run, because the app already paid for the introduction.

Four, win pickup first. Pickup is where direct ordering wins outright: no courier, no marketplace fee, and 7% to 10% saved on every order you convert. It is also the easiest habit to shift, because the customer is already driving to you.

Five, rent couriers instead of a channel. On-demand delivery for orders that came through your own site is a per-delivery cost, which behaves very differently from a percentage as your average order rises.

Six, look at the split every month. Marketplace orders, direct orders, and repeat customers on each. Fifteen minutes with those three numbers will tell you more than any dashboard the apps give you.

What we charge, and who this is not for

Our prices are on the pricing page for the same reason the rates above are public. A focused build is $1,499, paid once. Continuing work, where the site keeps changing and somebody owns the results, is $2,499 a month. The count is past 200 projects shipped, among them Eden Digital, a site we built that went on to grow revenue tenfold, and a redesign for Cornerstone Healing Center that lifted website conversions by 20%. A focused build usually takes about two weeks, and requests get an answer inside 48 hours.

This is not for every restaurant. If you do 15 delivery orders a week, the fee is not your problem and a build will not pay for itself soon enough to matter. If your ordering already runs through a point of sale you like, the honest advice is to switch on its direct channel and spend the money on the reasons your existing traffic is not converting instead. We are worth calling when delivery is a real share of revenue, the site is the weak link, and nobody currently owns the job of moving repeat orders off the marketplace.

Before you sign anything, ask about these

Five short questions, useful against an app, an ordering vendor or an agency including this one.

What is my all-in rate per order, including processing? Get one number for a typical order, not a tier name. Percentages and per-order fixed fees behave differently at $18 and at $80.

How and when can that rate change? Ask for the notice period in writing. March 2026 was a few days of warning for a change of a third on one Uber tier.

Who owns the customer data? Names, emails, order history and the right to market to them. If the answer is complicated, the answer is that you do not own it.

What does the fee include that I would otherwise buy? Couriers, insurance, support, placement, promotions. Some of it is genuinely expensive to replace.

What happens to my orders if I stop paying? On a marketplace, everything stops. On your own domain and your own processor, you keep the page, the list and the history. That difference is the reason to build one, and it does not show up in either price.