Profitable delivery: re-engineering last-mile economics for restaurants
Verdict: delivery is not profitable by default — it is profitable by design. In 2026 the question is no longer «sell on an aggregator or not», but running the channel with two distinct engines: third-party aggregator (maximum reach, 15-30% commissions that devour contribution margin) versus dark kitchen with an owned channel (low CapEx, control of food cost variance and the last mile). For the 1-3 location operator losing money on every app order today, the profitable route is hybrid: use the aggregator as paid acquisition measured by CAC, then migrate repeat orders to the direct channel, where contribution margin per order climbs from negative into the 12-22 point range. Every order that fails to cover its theoretical cost + fee is a subsidy drawn straight from your EBITDA.
Delivery stopped being a pandemic add-on and became permanent restaurant infrastructure. The global food delivery market reached USD 1.22 trillion in 2024 (Statista Market Insights, Online Food Delivery 2024), and cloud/ghost kitchens are projected at USD 88.7 billion in 2026 with a 12.6% CAGR through 2033 (Grand View Research 2026). The problem is not demand: it is that most restaurants sell plenty of delivery and earn nothing from it.
The cause is structural. The average operator adopted aggregators without recomputing unit economics: kept the dine-in menu price, absorbed 15-30% commissions, added packaging and loaded the channel with labor its break-even could not support. The result is a channel that bills revenue but subtracts EBITDA — a silent subsidy to the aggregator.
This white paper treats the last mile as a financial engineering problem, not a marketing one. Diego F. Parra and the Masterestaurant framework break the delivery order into its cost components, quantify the commission leak, model stress scenarios for input inflation, and deliver a 90-day roadmap to move the channel from loss center to margin center.
Side-by-side comparison
| Third-party aggregator | Dark kitchen + owned channel | |
|---|---|---|
| Commission / fee per order | ✕15-30% of ticket (leading aggregators) | ✓2-4% (owned-channel payment gateway) |
| Startup CapEx | ✕≈ USD 0 (existing kitchen only) | ✓USD 15,000-60,000 per dark-kitchen cell (Research and Markets 2024) |
| Last-mile control | ✕None (aggregator courier) | ✓High (owned fleet, courier or hybrid) |
| Customer data ownership | ✕None (aggregator keeps the CRM) | ✓Full (owned base, remarketing, LTV) |
| Typical contribution margin per order | ✕−8% to +6% (by category) | ✓+12% to +22% (optimized direct channel) |
| Acquisition cost (CAC) | ✕High, implicit in the commission | ✓Measurable and amortizable on repeat |
| Geographic scalability | ✕Immediate (aggregator coverage) | ✓Per cell / managed territory risk |
Chapter 1 — Why does delivery bill big yet leave no margin?
Delivery erodes EBITDA not because of weak demand but because almost no one recalculates unit economics before entering the channel.
Global delivery hit USD 1.22 trillion in 2024 (Statista Market Insights, Online Food Delivery 2024) and Mexico alone moved US$9.22 billion that year, growing at a 14.66% CAGR (Statista 2024): demand is abundant. The problem is structural. The average operator kept the dine-in menu price, absorbed aggregator commissions of 15-30%, added packaging, and loaded the channel with labor its break-even never supported. Diego F. Parra repeats it in every engagement: an order that leaves 68 cents on the dollar in the dining room may leave 12 or less through an aggregator. The result is a channel that bills hard but operates as a silent subsidy to the aggregator, draining the margin that keeps the house alive. The aggregator optimizes ITS reach; the dark kitchen optimizes YOUR margin — two engines with different objective functions, and confusing them is what breaks the channel's unit economics.
Chapter 2 — Aggregator and dark kitchen solve different problems
The ghost kitchen market reached USD 70.4 billion in 2024 (Research and Markets, Ghost Kitchen Market 2024) and the global cloud/ghost segment is projected at USD 88.7 billion in 2026 with a 12.6% CAGR through 2033 (Grand View Research 2026): that expansion is no fad, it is the sector's answer to a broken margin. The aggregator buys you maximum reach in exchange for a perpetual 15-30% commission per transaction. The dark kitchen or own channel cuts that toll by eliminating floor, server, and premium rent. Under the Masterestaurant framework, the right call is not «either/or» but assigning each order to the engine that leaves the most margin: reach where volume pays the commission, margin where volume already exists. With the aggregator you pay per transaction forever; with your own channel you pay CapEx once and lower the marginal cost per order as volume rises.
Chapter 3 — CapEx once versus commission forever
The gap compounds fast: on a Latin American online delivery market of USD 12,917.3 million in 2024, growing at 8.6% CAGR through 2030 (Grand View Research 2025), a 25% commission applied perpetually transfers one in four revenue dollars to the aggregator — forever. The own channel invests once in integration, ordering kiosk, and contracted logistics, and past a certain volume threshold cost per order falls while the aggregator's stays fixed as a percentage. Diego F. Parra models that crossover point per restaurant: where the aggregator already brings a stable 40-60 daily orders, moving half to the own channel usually recovers CapEx in under six months and frees 8 to 14 margin points previously lost to commission. The aggregator gives you reach but hides the customer; the own channel gives you the data to build LTV, remarketing, and menu by territory. That information asymmetry is the invisible cost of third-party delivery: you never see who bought, how often, or what they stopped ordering.
Chapter 4 — Customer data is worth more than the order
The virtual restaurant and delivery market reached US$66.3 billion in 2024 and is projected at US$140.4 billion in 2033 (Verified Market Reports 2024), and much of that value is captured by knowing the diner, not just dispatching them. With first-party data, a restaurant segments by zone, tunes the menu by territory, and reactivates dormant customers at near-zero acquisition cost. Diego F. Parra puts it bluntly: on the aggregator you rent traffic; on the own channel you build an asset. The first channel dies the day it raises fees; the second capitalizes every order into measurable LTV. Delivery break-even is not the dining room's: packaging, commission, and the last mile change the prime cost per order, and that number —not the physical menu's— decides whether you win. A dish costed at 30% food cost in the dining room can climb to 42-48% effective prime cost on the aggregator once you add packaging (USD 0.40-1.20 per order), a 15-30% commission, and assembly labor.
Chapter 5 — Delivery break-even is not dine-in break-even
On a Spanish delivery market near USD 5 billion (Ken Research 2025), operating with dine-in costing is a guaranteed loss. The Masterestaurant method rebuilds prime cost per channel: each delivery order is costed with its own packaging, commission, and share of logistics before pricing. Diego F. Parra insists the delivery menu must carry different prices and even different dishes than the dining room; matching the two is the costliest and most common mistake. The delivery channel withstands input inflation only if it is modeled in advance, not once margin has already vanished. Automation marks the defensive path: the food robotics market reached USD 1.81 billion in 2023 (Grand View Research, Food Robotics Market 2023) and delivery robots hit USD 795.6 million in 2025 (MarketsandMarkets 2025), signaling an industry compressing cost per unit to survive. In a channel where the aggregator takes a fixed 15-30%, every point of input increase hits an already thin margin directly.
Chapter 6 — Modeling stress from input inflation
Diego F. Parra runs stress scenarios in every engagement: what happens to channel EBITDA if protein rises 12%, packaging rises 8%, and the aggregator adjusts its commission. The dark kitchen cushions better because its fixed cost per order is lower and scales with volume, while the aggregator transfers all inflation straight to the operator. Turning delivery into a margin center is a 90-day project with measurable milestones, not a one-day price tweak. The Masterestaurant framework splits the quarter into three stretches: first, break the order into its cost components and measure the real leakage from commission and packaging; second, redesign the delivery menu with its own prices and prime cost, migrating 30-50% of stable volume to the direct channel; third, activate customer data for LTV and remarketing by territory. Context backs the urgency: the global cloud kitchen market reached USD 80.3 billion in 2025 (Grand View Research, Cloud Kitchen Market 2025) and dark kitchens USD 58.1 billion in 2024 (Global Growth Insights 2024).
Chapter 7 — 90-day roadmap: from loss center to margin center
Diego F. Parra sets the goal with a number: by day 90, the channel must deliver positive EBITDA or shrink to only the orders that truly pay their commission. Without that design, delivery stays a subsidy to the aggregator. The aggregator optimizes ITS reach; the dark kitchen optimizes YOUR margin. They are engines with different objective functions — confusing them is what breaks the channel's unit economics. On the aggregator you pay for every transaction forever; on the owned channel you pay CapEx once and lower the marginal cost per order as volume grows. The aggregator gives you reach but hides the customer; the owned channel gives you the data to build LTV, remarketing and territory-level menus. The delivery break-even is not the dine-in break-even: packaging, commission and last mile change the per-order prime cost, and that number — not the physical menu's — decides whether you profit.
Aggregator vs. dark kitchen: criterion-by-criterion analysis
When the aggregator DOES make sensePaid acquisition
- Launching a virtual brand with no customer base: the aggregator is instant traffic.
- High-ticket categories (>USD 25) where 20-30% commission still leaves positive contribution margin.
- Covering hours/zones where your owned fleet cannot reach at marginal efficiency.
- Treating the commission as measured CAC: if the customer re-orders through your direct channel, the aggregator paid for itself.
When dark kitchen + owned channel winsMasterestaurant
- Proven recurring volume: repeat justifies the cell CapEx and cuts the fee from 25% to 3%.
- Low-to-mid ticket categories where the aggregator commission erases all margin.
- You need the customer data to build LTV, remarketing and zone-level menus.
- Multi-brand virtual from a single kitchen: you leverage fixed prime cost across 3-6 concepts.
Side-by-side comparison
| Third-party aggregator | Dark kitchen + owned channel | |
|---|---|---|
| Commission / fee per order | ✕15-30% of ticket (leading aggregators) | ✓2-4% (owned-channel payment gateway) |
| Startup CapEx | ✕≈ USD 0 (existing kitchen only) | ✓USD 15,000-60,000 per dark-kitchen cell (Research and Markets 2024) |
| Last-mile control | ✕None (aggregator courier) | ✓High (owned fleet, courier or hybrid) |
| Customer data ownership | ✕None (aggregator keeps the CRM) | ✓Full (owned base, remarketing, LTV) |
| Typical contribution margin per order | ✕−8% to +6% (by category) | ✓+12% to +22% (optimized direct channel) |
| Acquisition cost (CAC) | ✕High, implicit in the commission | ✓Measurable and amortizable on repeat |
| Geographic scalability | ✕Immediate (aggregator coverage) | ✓Per cell / managed territory risk |
The size and economics of the last mile in numbers
“We were doing 3,400 app orders a month and thought we were winning. When Diego broke the order down, each one lost USD 0.80 after commission, packaging and waste: we lost USD 2,720 a month by «selling well». We spun off a delivery-only virtual brand, migrated repeat orders to WhatsApp with our own gateway, and cut the fee from 27% to 3.2%. Contribution margin per order went from −3% to +17% in 74 days, without raising the ticket.”
90-day roadmap: from loss center to margin center
Break an average delivery order into its components: real food cost, packaging, aggregator commission, imputable labor cost and waste. Compute contribution margin per order and per channel. Most operators discover here that 20-40% of their delivery SKUs sell below theoretical cost + fee. This number, not the dine-in menu's, is your starting point.
Apply menu engineering to the channel: raise price or redesign dishes that cannot absorb the commission, retire those that bleed margin, and build a delivery menu with food cost ≤ 32% per dish and optimized packaging. Design the virtual brand: a delivery-only concept that leverages your kitchen without cannibalizing dine-in. The goal is that every app SKU carries positive contribution margin AFTER the fee.
Stand up the owned channel: WhatsApp/web ordering with a payment gateway (2-4% fee vs. 25%), courier fleet or hybrid for the last mile, and customer data capture. Use the aggregator as paid acquisition and migrate repeat orders to the direct channel with a measured incentive. Every customer who re-orders through your channel amortizes the aggregator's CAC.
With unit economics proven, decide the CapEx: a dark-kitchen cell for multi-brand, or keep leveraging the current kitchen? Install the KPI dashboard — contribution margin per order, aggregator/direct mix, CAC, LTV and food cost variance — with reviews at 3, 6 and 12 months. Scale only where per-order margin is positive and territory risk is managed.
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
Ecosystem tools to operate the channel
The Masterestaurant framework does not stop at diagnosis: every roadmap component has a concrete ecosystem tool that operationalizes it. These three cover model design, channel growth and last-mile cash control.
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 the book «From Slave to Owner» (Amazon) · International keynote speaker for the HORECA sector.
Frequently asked questions about delivery economics
Why do I lose money if I sell a lot through the app?
Why do I lose money if I sell a lot through the app?
Because selling is not earning. Between aggregator commission (15-30%, leading aggregators), packaging, waste and imputable labor, many delivery SKUs sell below theoretical cost + fee. Contribution margin per order, not volume, decides whether the channel adds or subtracts EBITDA.
Should I open a dark kitchen or stay with the aggregator?
Should I open a dark kitchen or stay with the aggregator?
It depends on recurring volume. The aggregator is ideal for acquisition and high ticket (>USD 25). The dark kitchen with an owned channel wins when repeat justifies the CapEx (USD 15,000-60,000 per cell, Research and Markets 2024) and cuts the fee from 25% to 3%. The profitable route is usually hybrid.
What food cost should a delivery dish have?
What food cost should a delivery dish have?
Food cost per dish should stay at ≤ 32% maximum, same as dine-in, remembering that payroll, rent and packaging are NOT loaded onto the dish: they belong to the channel's break-even. The delivery menu is redesigned with menu engineering so it can absorb the commission.
How do I recover the customer the aggregator gives me?
How do I recover the customer the aggregator gives me?
Treat the commission as CAC. Include a measured incentive (QR, coupon, note) inviting a re-order through your direct channel — WhatsApp or web with your own gateway (2-4% fee). Every migrated customer amortizes the aggregator cost and gives you the data to build LTV and remarketing.
Sector data 2026 (official sources)
Verifiable industry benchmarks from official, non-commercial sources (government, industry associations, market research) - not competitors.
| Metric | Benchmark 2026 | Source |
|---|---|---|
| Mercado de entrega de paquetes por dron en 2023 | USD 585,9 millones | Grand View Research — Drone Package Delivery Market 2023 |
| Proyección de entrega de paquetes por dron a 2030 | USD 5.238,8 millones (CAGR 38,7%) | Grand View Research — Drone Package Delivery Market 2030 |
| Entregas comerciales por dron de Zipline (abril 2024) | 1 millón (primera empresa en lograrlo) | Grand View Research — Drone Package Delivery Market |
| Unidades de drones de reparto proyectadas 2024 a 2030 | de 32.456 a 275.703 unidades | Grand View Research — Drone Package Delivery Market |
| Cuota del delivery de comida en el mercado de drones 2024 | 36,87% | Grand View Research — Drone Package Delivery Market 2024 |
| Pedidos de DoorDash en el cuarto trimestre de 2024 | 685 millones (+19% interanual) | DoorDash — Q4 y Full Year 2024 Financial Results |
Download this document as PDF
The full text is free to read on this page. To take the corporate PDF with you, leave your details — we'll also email you the direct link.
Related content
Turn your delivery into a margin center
If you sell a lot through the app and never see the money in the till, the problem is the channel's unit economics, not demand. Diego F. Parra and the Masterestaurant framework break down your delivery order, quantify the commission leak and design the hybrid aggregator-owned route that raises your contribution margin per order. Start by modeling the business with the ecosystem tools.