Own channel vs apps: what changes in EBITDA when you stop renting your own customer

1. Which channel leaves more money per order served: your own or the aggregator?

Your own channel leaves more money per order served, always, because the aggregator takes its cut off gross sales while the direct channel only pays the payment gateway and the courier.

Run it with cash numbers: on a 20-dollar ticket, with food cost at 30% and kitchen labor at 18%, the direct order leaves roughly 8.20 dollars after gateway and outsourced delivery; that same order through an app, with platform commission at the high end, drops to 3 or 4 dollars. That gap is no local anomaly: it is arithmetic that repeats in every market with heavy concentration, and concentrated markets are the rule — iFood holds 87% of e-food bookings in Brazil according to Statista (2024), and Grab holds 53.9% of Southeast Asian food delivery according to Momentum Works (2024). Below half a million in annual revenue, there is only one call to make: WhatsApp Business plus a Google Business Profile with an order button, and not a single dollar spent on custom development.

2. Under 500 thousand dollars a year: your channel starts on WhatsApp, not on a custom app

An operator in this band billing 400 thousand a year with 35% through delivery hands roughly 140 thousand gross dollars to the aggregators; at mid-band commission that means between 25 thousand and 40 thousand dollars a year, a figure usually buried inside «other operating expenses» in the P&L. The decision threshold here stays simple: until the direct channel clears 200 orders a month, neither a paid ordering engine nor an in-house fleet is justified. The owner in this band does not fight the app, uses it as a storefront and works repeat purchase through direct messaging, which is the only asset left standing when the platform rewrites its rules. In the 500 thousand to one million band, a direct ordering engine pays for itself and delivery menu engineering stops being optional. The rule I take into board meetings: once the direct channel clears 300 monthly orders, a fixed-subscription ordering platform costs less than variable commission from month one.

3. From 500 thousand to 1 million: pay for the ordering engine and cut the delivery menu

And the second move carries as much weight as the first, because most operators publish the dining-room menu on the app, with dishes that fall apart across twenty minutes of transit, then blame the platform for the one-star review. Cut down to the six or eight dishes that travel, put packaging into the spec sheet — between 4% and 7% of selling price — and recalculate food cost with that cost inside. Independent kitchens hold 61.7% of cloud kitchen revenue according to Grand View Research (2025): this band is the heart of the market, not the exception. Past a million a year, stop thinking about switching the app off and start thinking about the MIX. The operating target is to bring the direct channel to 30% of digital orders within twelve months and hold there, because past that point you carry enough volume to renegotiate commission with the platform without bluffing.

4. Above 1 million: split the volume and negotiate commission with your own data

An operator billing 1.4 million with 45% digital moves some 630 thousand dollars through channels; shifting 30% of that to the direct channel recovers between 55 thousand and 90 thousand dollars a year now lost to commission. In Spain, where Glovo sits near 31% and Just Eat near 26% of the market according to Ken Research (2025), a split between two players gives real negotiating room; in Brazil, with iFood at 87% according to Statista (2024), pricing power belongs to the platform and the direct channel works as insurance rather than as leverage. Above five million a different profile shows up — the celebrity-chef restaurant or the large-format themed venue — and it usually errs through excess confidence in its own brand. That profile sells reservations and dining-room experience, so it treats delivery as incremental revenue and accepts commissions it would never sign into a lease.

5. Above 5 million: the high-end profile buys traffic, not relationships

My reading is that the accounting error gets worse there: the operator compares commission against a 22% dining-room margin and concludes delivery «is not worth it», when what is not worth it is delivery WITHOUT a proprietary database. At 5 million in revenue and barely 12% digital, we are talking about 600 thousand dollars flowing through the app and some twenty thousand yearly contacts that never reach the CRM. The threshold for this band is one dedicated person on direct channel and CRM, plus a repeat-purchase campaign budgeted against the commission no longer paid. In a group or chain above ten million, the direct channel stops being a marketing project and becomes infrastructure with an owner, a budget and a monthly committee. Figures change scale here: ten million at 40% digital means four million through channels, and every percentage point migrated to the direct channel is worth between 3,500 and 6,000 dollars a year in avoided commission, so twenty points of migration pay for a full team.

6. Groups and chains above 10 million: the direct channel is infrastructure, not a project

The useful comparison is not aggregator versus own website, it is customer acquisition cost in each one; and as a reference for ecosystem scale, iFood reports over 380,000 partner establishments across more than 1,500 Brazilian cities (2024): you are one row in that table, your brand name printed small. Diego F. Parra and the Masterestaurant team build that architecture by revenue band, pulling commission out of «other expenses» and giving it its own P&L line. Switching the aggregators off overnight costs between 25% and 40% of digital volume in the first quarter, and the second quarter does not bring it back. The scenario deserves to be followed all the way through, since almost nobody does it: the drop in orders drags purchasing down, the supplier pulls the volume discount, food cost climbs two or three points, and with the kitchen at half load labor per order shoots up.

7. What happens if a group switches the aggregators off overnight?

By month four the owner reopens the app on worse terms, without ranking and without recent reviews. The correct sequence runs the other way:

first build a proprietary base while app volume pays for the traffic, then migrate through incentive — a discount equal to half the commission, priority preparation, or free shipping on the second order — and only at the end trim exposure. Online meal delivery in Latin America will pass 39 billion dollars by 2027 according to Statista (2024); that market gets contested, never abandoned. Pull aggregator commission out of «other operating expenses» and open its own P&L line before you close the month: that move alone changes decisions. What stays invisible in the decision architecture never gets managed, and a commission treated as an administrative expense never reaches the pricing committee or the menu engineering session, so the operator prices delivery like the dining room and gives away margin without noticing.

8. Commission leaves «other expenses» and enters the pricing committee this month

With the line visible, the conversation shifts by itself: differentiated pricing per channel appears, the dish that sells heavily through the app and leaves 1.80 dollars appears, and the question of what a contact is worth appears. An honest comparison between channels is not about percentages, it is about accounting nature: on the app you pay a recurring EXPENSE for every order, while on your own channel you build an ASSET that can be used again without paying twice. The difference is not the commission, it is WHO owns the relationship. An aggregator order and a direct order leave the same kitchen, with the same food cost and the same labour; the only thing that changes is that one leaves a reusable trace and the other leaves none. Comparing channels by commission percentage alone is therefore an accounting mistake: you are weighing an expense against an asset. The second break point is the menu.

9. Where delivery unit economics actually breaks

Plenty of operators publish the dining-room menu inside the app, with dishes that arrive cold or fall apart after twenty minutes in a bag, and then blame the platform for the one-star review. Delivery menu engineering is a separate exercise: trim to items that travel, put packaging inside the recipe card, recalculate food cost with that packaging included and keep it under 32%. The third is territory risk. Concentrating 87% of volume in one platform, as happens in markets where a single player dominates —Statista (2024) reports that share for iFood in Brazil—, means a unilateral change in commission, ranking algorithm or promotional policy moves your EBITDA without your vote. Corporate governance has a name for that: single-counterparty exposure. And there is a paradox worth resolving before deciding: the own channel earns more per order and still, almost never, should the apps be switched off. The aggregator is where the customer who has not met you yet still lives, and the AI recommendation shortlists now mediating «food near me» searches feed on public signals —listings, reviews, availability— that platforms amplify.

10. Where delivery unit economics actually breaks — in practice

The answer is not picking one channel; it is setting an exposure ceiling per platform and using aggregator commission to buy customers you afterwards move into your own base.