Own channel vs apps: the numbers that decide how much of each order you keep

The number that decides everything: contribution per order, not revenue

A 45,000-peso order on an app charging 27% commission leaves 32,850 before you touch the product, and if your food cost sits at 32% —the ceiling Masterestaurant accepts per dish— you spend 14,400 on ingredients and keep 18,450 pesos of gross contribution, against the 30,600 the same order yields through your own channel. That gap of 12,150 pesos per ticket separates an operation that breathes from one drowning in revenue, and it explains why sector net margin runs between 3% and 9% according to Statista while owners celebrate record sales. The Medellín virtual brand billing 78 million pesos a month on Rappi lost 1,900 pesos per delivery because its menu had been costed at 34% food cost, designed for the dining room, on top of a commission structure that forgives nothing. High-performing ghost kitchens run margins of 10% to 30%, against the 3%–5% of a traditional restaurant, according to OysterLink 2025 — and that wide range is not statistical noise, it is the distance between operators who blend channels and those depending on the marketplace.

How much real margin does a kitchen living only on platforms keep?

Initial investment for a ghost kitchen runs from USD 75,000 to 200,000 (OysterLink 2025), capital you only recover if contribution per order clears roughly 8,000 pesos across all channels weighted together.

When 100% of volume arrives through commission, the ceiling stays near 10%; when your own channel supplies a third of orders, that same kitchen lands in the 20% band. Before buying equipment, work out how many direct orders per month you need to return the capital within 24 months, then put that figure on the kitchen wall. DoorDash moved USD 21.3 billion of marketplace GOV in the fourth quarter of 2024, up 21% year over year (DoorDash, Q4 and Full Year 2024 results), and China will bill USD 539.87 billion in online delivery during 2026 according to Statista. Against numbers like those, arguing whether platforms are «worth it» amounts to arguing with the weather.

The volume apps do bring and no direct channel replicates

The app does not sell your food: it sells the intent of someone who opened a phone hungry and with no brand in mind, and that discovery traffic carries a value your Instagram will not manufacture alone. The owner who walks away from platforms overnight finds out that 70% of orders came from people searching for food nearby, not searching for him. Treat the app as an acquisition window and measure how many of those customers migrate to your own channel by the second order: if fewer than 15% migrate, your problem is not the commission, it is that nobody is asking for the data. Bringing in a new customer through geotargeted ads costs between 9,000 and 22,000 pesos ONCE, and that customer returns free of charge if you hold their phone number. The same customer on a platform charging 27% costs you 12,150 pesos on a first 45,000-peso order and costs you 12,150 again on the tenth, because commission is not an acquisition expense but an equity stake in the customer's lifetime.

Commission never amortizes: it charges on every future order from the same customer

Run it across ten annual orders: 121,500 pesos handed to the platform against a maximum of 22,000 spent one single time on your own acquisition. What many read as «expensive marketing» is really a perpetual royalty you signed without reading the fine print. With 32% of restaurant expansion strategies leaning on virtual brands during 2025 (Technomic, via Apicbase), the model is normalizing right when its arithmetic deserves scrutiny. Some 86.9% of US virtual brands operate a hybrid model and only 13.1% live exclusively online, according to Locmatic's State of Virtual Restaurant Brands 2024 — a split that came from thousands of daily closeouts rather than from a consulting committee. Brooklyn Calzones leads with 1,474 locations and a 12% share (Locmatic 2024), built on kitchens that already existed and already had customers of their own. The operational reading is blunt: whoever bets 100% on platforms sits in the 13% of the market with the least control over price, data and margin.

Hybrid wins: 86.9% of the market already settled this

Set a quarterly mix target —start at 70% apps and 30% direct, then move five points per quarter toward your own channel— and review it against the contribution report, never against the gross sales screen the app shows you. Sector averages advise you badly when your food cost, your ticket and your neighborhood density look nothing like the average. In a small location under 400 monthly orders, the priority is not leaving the apps but lifting the ticket: at 18,450 pesos of contribution per platform order, every extra 5,000 pesos of ticket is worth more than two negotiated commission points. In a mid-size operation between 400 and 1,500 orders, the game is the mix: moving your own channel from 10% to 30% across 1,000 orders frees 36 million pesos of annual contribution without selling one additional dish. In a multi-brand group, the lever is data: a history of 20,000 customers with phone numbers and frequency lets you launch a new virtual brand on direct acquisition from day one, instead of paying the discovery commission all over again.

How to read these numbers in YOUR operation: small, mid-size and group?

Calculate your figure, not mine. The margin, investment and share figures come from OysterLink 2025, from Locmatic's State of Virtual Restaurant Brands 2024, from Technomic via Apicbase, and from the financial results DoorDash published for the fourth quarter of 2024;

sector net margin comes from Statista and the Chinese delivery market from its 2026 projection. The limits belong in plain sight: most of that evidence is North American and applies to Latin America with caution, where commissions reach 30% and the average ticket is three times smaller. The commission, ticket and acquisition-cost ranges in the tables reflect Latin American market structures in 2026 and shift by city and category. None of these numbers replaces your own daily closeout: they are the reference against which you measure your deviation, and the work Diego F. Parra does with the Masterestaurant teams starts exactly there, at the deviation, never at the average.

The question that decides your price three years from now

What would happen if tomorrow your main platform lifted its commission from 27% to 32%? On 800 monthly orders of 45,000 pesos, those five points mean 1,800,000 pesos less contribution per month, 21,600,000 a year, and you would have no leverage to negotiate because 90% of your customers do not belong to you: no phone number, no frequency data, no way to reactivate them with a campaign of your own. That scenario separates the two kinds of operator we will meet in 2029. Global agrifoodtech investment fell 4% to USD 16 billion in 2024 (AgFunder, Global AgriFoodTech Investment Report 2025), yet in developing markets it grew 63% to USD 3.7 billion (AgFunder News): capital keeps flowing into our region's platforms and capital demands a return. Start this week with the one thing that depends on nobody: ask for the phone number on every direct order and store it with the date.

Four differences that change your P&L

Commission is not a marketing expense, it is a partnership stake. A 27% commission on a business running 30% food cost means the platform takes almost as much as the product does, and it takes it on EVERY future order from the same customer, not just the first. Own-channel acquisition cost is fixed and amortises; app acquisition is variable and perpetual. Bringing a customer in through geotargeted ads costs 9,000 to 22,000 pesos once. That same customer on Rappi costs 12,150 pesos on a 45,000 first order and costs 12,150 again on the tenth. Customer data decides who sets the price three years from now. With no phone and no history you cannot reactivate, cannot segment by frequency and cannot defend yourself when the platform raises its fee. With your own base, one reactivation campaign to 800 lapsed customers recovers 60 to 130 orders.

Four differences that change your P&L — in practice

Speed works in reverse for each channel. Apps deliver traffic in two weeks and charge for it forever; the own channel takes two months and then gets cheaper every month. Anyone reading only the current quarter always chooses wrong.