Own delivery vs apps in 2026: what is a measurable trend and what is hype
Verdict: in own delivery vs apps the right 2026 answer is not picking one, it is splitting the traffic by margin logic: aggregators buy DISCOVERY at a cost of 15% to 30% commission, while the owned channel buys REPEAT at a marginal cost near 6-9% between payment gateway and courier. The rule I have applied for years: the aggregator keeps a new customer's first order, the owned channel keeps the second onward. If your delivery average ticket sits below 12 USD, no app leaves you margin and the real conversation is a different one.
A client with two locations in Medellín showed me his June 2026 close with a number he read as a win: 41% of sales came from aggregators. We measured the contribution margin of that 41% and it was 8.4%. The remaining 59%, between dining room and his own WhatsApp and web ordering, delivered 34.1%. He was selling almost half to earn a tenth.
That arithmetic governs the own delivery vs apps conversation in 2026, and almost nobody runs it before signing. Aggregator commissions across Latin America move in a 15% to 30% band depending on country, plan and paid visibility, on top of co-funded promotions, service fees and packaging costs the restaurant absorbs in full.
This year's shift is not that apps got more expensive. It is that the owned channel stopped being a technology project and became a three-piece setup any serious operator can assemble in weeks: a worked Google Business Profile, an ordering engine with a local payment gateway, and a messaging layer for repeat orders. The technical barrier fell; the operational discipline barrier did not.
And here sits the tension almost nobody resolves: the owned channel is more profitable PER ORDER, yet it generates no new demand by itself, while the aggregator generates new demand and destroys margin. Whoever picks only one loses on one side or the other. The way out is architectural, not philosophical: each channel plays a different role in the customer lifecycle, and each gets measured differently.
Side-by-side comparison
| Aggregators only (traditional method) | Governed mix (Masterestaurant method) | |
|---|---|---|
| Effective commission on gross sales | ✕15% to 30% depending on plan and country | ✓6% to 9% on owned channel; 15-30% on acquisition only |
| Channel contribution margin | ✕8% to 14% at a 30% food cost | ✓26% to 34% at the same 30% food cost |
| Ownership of customer data | ✕0% of the list; the aggregator keeps phone and habit | ✓100% of owned orders: name, frequency, ticket |
| Cost to acquire a new customer | ✕3 to 6 USD equivalent per first order | ✓3 to 6 USD once, near 0.4 USD on every repeat |
| Time to first operating order | ✕7 to 21 days of onboarding and verification | ✓10 to 30 days to build profile, engine and gateway |
| Dependence on a third party to sell | ✕Total: one account suspension kills 100% of the channel | ✓Partial: the owned channel holds 40-60% of sales |
| Available growth lever | ✕In-app advertising and co-funded promotions | ✓Local SEO, Maps, reviews, geo ads and an owned list |
How much real margin does an app order leave versus your own channel?
An aggregator order leaves between 8% and 12% contribution margin when commission sits near 25%, while that same ticket through your own channel leaves 30% to 35%, because there you only pay the payment gateway (2.9% to 3.9%) plus a fixed subscription.
That twenty-something point gap governs the decision, and the 2026 trend makes it urgent: online delivery moved roughly USD 1.4 trillion last year (Statista, Online food delivery statistics & facts 2025), so the channel is no longer marginal for anyone. A client with two locations in Medellín closed June with 41% of sales through apps and an 8.4% margin on that slice; the remaining 59%, between dining room and his own website, left 34.1%. He was selling almost half to earn a tenth. Before negotiating any plan or paid visibility, run that dual margin per channel on your own closed month.
Commission is not the only cost: count the co-funded promotions
Latin American commissions range from 15% to 30% depending on country, plan and contracted visibility, yet the REAL cost of selling through an aggregator usually lands three to six points higher once you add co-funded promotions, service fees and packaging you absorb entirely. Operator numbers explain why the pressure will not ease: Delivery Hero booked €12.8 billion in segment revenue during 2024, up 22% year over year (Delivery Hero, Q4 and FY 2024 Results), and Rappi hovered near USD 800 million in net revenue in 2023 (Statista 2024). Those platforms need every transaction to pay, and the adjustment comes out of the restaurant. What to do by size: if you bill under USD 20,000 monthly, refuse co-funded discounts and stay on the base plan; above USD 60,000, demand volume pricing and measure each promotion's return on a fourteen-day cut. Your own channel starts paying for itself somewhere between 18 and 25 daily orders, and below that threshold it costs more than it saves.
The crossover point: between 18 and 25 owned orders per day
The reason is how the cost behaves: an aggregator charges per transaction, so ten orders mean ten commissions, while an owned channel charges for infrastructure —domain, ordering engine, gateway, messaging— and that block runs USD 90 to USD 250 monthly whether you sell five or five hundred. With a USD 12 average ticket and 20 daily orders, the owned channel saves around USD 1,400 a month against a 25% commission. At six daily orders, the savings do not even cover the subscription. Work out your threshold with plain division: monthly fixed cost divided by savings per order, and do not build a store until your current app volume clears it. Google Business Profile is today the most profitable acquisition channel for delivery, because a well-tended listing shows in the Maps local pack at no cost per click and that traffic converts better: it arrives with direct purchase intent, not browsing a carousel of twenty restaurants.
The local engine: visibility you do not pay for by the click
A location with 4.6 stars and 300 reviews plays a different league than one with 4.1 and 40, and the difference is built by asking for a review at handoff, not by buying visibility. A warning belongs here: a badly chosen primary category is the error I find most often when auditing listings, and fixing it moves Maps positions within weeks. Diego F. Parra insists within the Masterestaurant method that the listing gets worked like a menu gets worked —monthly review of photos, hours and replies— because it is the one digital asset the restaurant rents from nobody. Whoever orders through an app is not your customer, they are the app's customer, and that is the silent loss no P&L shows. With no phone, no email and no purchase history, you cannot build repeat business and you keep paying commission forever on the same diner. Your own channel flips that logic: the marginal cost of bringing back someone who already bought runs 6% to 9% between gateway and messaging, against 15% to 30% to buy them again on the platform.
Repeat business is where you win: the customer record beats the order
If a restaurant with 600 monthly app orders recovered just 30% of those diners into its own channel over a year, the margin difference clears USD 6,000 at the region's average ticket. What to do: slip a discount code into every package, redeemable only on your site, and measure how many come back. Automating order-taking with AI is still early and deserves watching rather than becoming your backbone: only 6% of restaurants in the United States use it to take customer orders (National Restaurant Association 2026), and that market usually runs two or three years ahead of Latin America in technology adoption. What does work today is the light layer: automated order-status replies over WhatsApp, cross-sell prompts in the cart, and route prioritization. The full version —a conversational agent taking the whole order by voice— breaks on accents, modifications and dishes with local names, and every mistaken order costs the plate plus the reputation.
AI order-taking: the trend you should not adopt in full yet
Running one location, stay in the light layer. Running five or more with a stable menu, pilot voice at a single site for one quarter before deciding. A ghost kitchen does not fix a margin problem, it hides it behind fresh capital expenditure, and it is the trend most operators are adopting for the wrong reason. China holds more than 3,200 ghost kitchen facilities and is the largest national market (Coherent Market Insights 2024), but that ecosystem runs on urban density and logistics costs the region does not reproduce. In Latin America the capital to test it is drying up: agrifoodtech investment fell 24% during 2024 to USD 249 million (AgFunder 2025), while India climbed 215% to USD 2.5 billion. Translated into cash: if your food cost sits at 34% and your app channel leaves 8% contribution, opening a kitchen without a dining room multiplies bad-margin orders.
The overrated trend: the dark kitchen as a margin lifeboat
Fix the recipe costing and the channel mix first; a dark kitchen earns its place once proven demand of your own outgrows your production capacity. Split by margin logic and set a numeric target: 60% of sales through owned channels —dining room, website and WhatsApp— and 40% through aggregators, treating the app as paid discovery rather than your main kitchen. Adopt three things NOW: a Google Business Profile listing with fresh reviews and the right category, an ordering engine with a local gateway, and a messaging routine for repeat orders. Watch without investing yet: AI voice, in-app social commerce, and weekly meal subscriptions. The market gives scale context —the United States moved close to USD 353 billion in online delivery during 2024 and China close to USD 450 billion (Statista)—, so the channel will not shrink; what changes is who keeps the margin. Open your month-end close, separate contribution by channel, and set your split target before Friday.
The four differences that decide the outcome
Aggregators charge per transaction; the owned channel charges for infrastructure. Ten daily app orders cost ten commissions, while that same volume on your own site costs a gateway fee plus a domain and a software subscription that does not scale with sales. The typical crossover point sits between 18 and 25 owned orders per day, and below that threshold the owned channel does not pay for itself. Inside apps visibility is bought; inside the local engine it is earned. A Google Business Profile with fresh reviews, the right primary category and real photos surfaces in the Maps local pack without paying per click, and that traffic converts better because it arrives with direct purchase intent. A restaurant at 4.6 stars with 300 reviews plays a different league than one at 4.1 with 40. Customer data changes the entire economic game. With phone, frequency and ticket you can calculate lifetime value and decide what to pay for acquisition; without it, every order is an island and marketing turns into a bet.
The four differences that decide the outcome — in practice
Aggregators do not hand that data over, and that is the real reason the model squeezes, more than the commission itself. A dark kitchen inverts the two heaviest costs. With no dining room, rent drops from a typical 8-12% of sales to 3-5% and front-of-house payroll disappears, which makes a 25% commission hurt far less. That is why a ghost kitchen can live inside an app and a brick and mortar restaurant with a full room almost never can.
Head to head: aggregators against the owned channel
Depending on apps aloneTraditional method
- Fast onboarding and volume from week two, with no technical team and no meaningful upfront investment.
- The aggregator supplies demand: you never learn to generate traffic, you rent it month by month.
- Your margin gets decided on a screen you do not control, with unilateral plan and commission changes.
- Co-funded promotions move your in-app ranking, yet they eat another 4 to 8 margin points.
- With no owned database there is no targeted repeat: every order pays full commission again.
Governed mix with an owned channelMasterestaurant
- Google Business Profile worked properly: category, delivery attributes, dish photography and weekly posts.
- Owned ordering engine with local gateway, differentiated menu pricing and packaging costed separately.
- Repeat orders through messaging segmented by frequency: the fortnightly customer gets a different message than the quarterly one.
- Apps stay switched on with a narrow job: capture strangers and fill off-peak hours.
- A weekly dashboard showing margin per channel rather than sales per channel, because that is where the real decision lives.
Side-by-side comparison
| Aggregators only (traditional method) | Governed mix (Masterestaurant method) | |
|---|---|---|
| Effective commission on gross sales | ✕15% to 30% depending on plan and country | ✓6% to 9% on owned channel; 15-30% on acquisition only |
| Channel contribution margin | ✕8% to 14% at a 30% food cost | ✓26% to 34% at the same 30% food cost |
| Ownership of customer data | ✕0% of the list; the aggregator keeps phone and habit | ✓100% of owned orders: name, frequency, ticket |
| Cost to acquire a new customer | ✕3 to 6 USD equivalent per first order | ✓3 to 6 USD once, near 0.4 USD on every repeat |
| Time to first operating order | ✕7 to 21 days of onboarding and verification | ✓10 to 30 days to build profile, engine and gateway |
| Dependence on a third party to sell | ✕Total: one account suspension kills 100% of the channel | ✓Partial: the owned channel holds 40-60% of sales |
| Available growth lever | ✕In-app advertising and co-funded promotions | ✓Local SEO, Maps, reviews, geo ads and an owned list |
Signals you can actually measure
“We hit 41% of sales through apps and thought we were winning. Diego made us split margin by channel and the aggregator left 8.4% against 34.1% from direct orders. We built the Maps profile, an ordering engine with a gateway and WhatsApp repeat flows, and raised prices 12% on the app only. In five months the owned channel went from 6% to 38% of delivery sales, total revenue rose 9% and operating margin moved from 11.2% to 19.6%. We are still on Rappi, but now apps bring strangers and we keep the people who already know us.”
How to build it in 90 days without switching the apps off
Open a sheet with four columns per channel: gross sales, commission and fees, true food cost including packaging, and delivery cost. What remains is your contribution margin per channel. The surprise shows up almost every time: the channel with the most sales leaves the least. If your delivery food cost exceeds 32% per dish, fix that BEFORE touching anything else, because no channel strategy rescues a broken recipe cost. Mark your average ticket per channel too; if the app ticket sits under 12 USD, the problem is the menu you uploaded, not the commission.
Claim and complete your Google Business Profile: correct primary category, delivery and pickup attributes, real hours, 20 dish photos shot in natural light and one post per week. Ask for reviews with a QR code on every package, not with a generic message three days later. Answer EVERY review, bad ones included, within 48 hours. This layer is what puts you in the Maps local pack when somebody searches for food nearby, and that traffic charges you no commission at all.
Install an ordering engine with a local gateway and your own domain, load the menu with photos, and raise prices 10% to 15% ON THE APP, never on your own site. That gap becomes your conversion argument and it is standard practice among operators who survive the model. Define delivery zones by drive time rather than kilometres, and hire couriers by the hour during peak windows instead of per order. By the end of month two you should already see the first direct orders from customers who met you on Rappi.
Every owned order leaves a phone number and a habit behind, so segment by frequency and write three different messages: one for the customer who ordered once and vanished, one for the fortnightly buyer, one for the weekly regular. Run geolocated ads within a three-kilometre radius on a small, sustained budget pointing at your own menu rather than the app. Review every Monday what share of delivery sales came from your own channel; the 90-day target is 25%, and from there you govern.
And with AI?
Optimize channels, pricing and unit economics of your dark kitchen. Diego F. Parra is an expert in AI applied to restaurants.
Free tools to apply this now
Method tools for deciding with numbers
The own delivery vs apps decision comes down to three numbers rather than opinions: contribution margin per channel, customer acquisition cost, and a twelve-week cash projection. These three Masterestaurant ecosystem tools cover exactly that trio, and I use them in this order with every operator who arrives with the commission problem on their back.
Diego F. Parra — International consultant, expert in creating and scaling restaurants and in AI applied to restaurants, foodtech and HORECA. Methodology applied in 8.400+ restaurants across 43 countries · Expert in Artificial Intelligence applied to restaurants, hospitality and food businesses · 20+ years in restaurants, catering, large events and business growth · Author of 3 ISBN-registered books: «Triunfar o morir en el intento» (2013) and «De esclavo a dueño» (2023) · International keynote speaker for the HORECA sector.
Questions I get every week
Should restaurants leave delivery apps in 2026?
Should restaurants leave delivery apps in 2026?
Do not leave them, narrow them. Apps remain the cheapest channel for a stranger to try you the first time, and their equivalent acquisition cost of 3 to 6 USD per new customer is competitive. What you must stop is letting them own the repeat, because that is where a 15% to 30% commission gets paid on a customer who was already yours.
How long until an owned delivery channel turns profitable?
How long until an owned delivery channel turns profitable?
Between 60 and 120 days when the Maps profile is worked and there is a customer base to activate. Typical break-even sits between 18 and 25 owned orders per day, since infrastructure cost is fixed while aggregator commission is variable. Below that volume, keep the owned channel as a repeat engine and do not expect it to replace the app.
Is it fair to charge more on apps than on my own site?
Is it fair to charge more on apps than on my own site?
Yes, and it is what operators who survive the model do. A 10% to 15% price gap on the app covers part of the commission and hands you a real conversion argument toward direct ordering. Check the price parity clauses in your contract on each platform, because they vary by country and by contracted plan.
Does a dark kitchen solve the commission problem?
Does a dark kitchen solve the commission problem?
It eases it, it does not solve it. A ghost kitchen cuts rent from 8-12% of sales to 3-5% and removes front-of-house payroll, so it can absorb a 25% commission that a restaurant with a dining room cannot. Yet it still lacks customer data and an owned brand, so a dark kitchen that builds no direct channel has merely changed creditors.
Sector data 2026 (official sources)
Verifiable industry benchmarks from official, non-commercial sources (government, industry associations, market research) - not competitors.
| Metric | Benchmark 2026 | Source |
|---|---|---|
| Penetración de usuarios en restaurant delivery 2024 | 15,7% (proyectada a 18,1% en 2029) | Statista — Restaurant Delivery Worldwide |
| Marca virtual líder en EE.UU. por ubicaciones (Brooklyn Calzones) | 1.474 ubicaciones (12% de cuota) | Locmatic — State of Virtual Restaurant Brands 2024 |
| CAGR del mercado de ghost kitchens 2022-2032 | 11.65% anual | Statista/Toast (vía OysterLink) |
| Inversión inicial de una ghost kitchen | USD 75.000–200.000 | OysterLink 2025 |
| Ghost kitchens activas en EE. UU. | ≈7.606 operaciones | OysterLink 2025 |
| Margen de las ghost kitchens de alto desempeño | 10–30% (vs 3–5% del restaurante tradicional) | OysterLink 2025 |
Related content
Grow your restaurant with the Masterestaurant method
Applied in +8.400 restaurants across 43 countries.