Unit economics for owners: the before and after of fixing your local engine
The AFTER wins: an owner who rebuilds unit economics for owners from the local digital engine —a complete Google Business Profile, a trimmed delivery radius, geotargeted ads by zone and ratings above 4.6 stars— moves the contribution margin per order from 8% to 19% in the same kitchen, with the same menu and no price increase. The arithmetic is plain: shifting the order mix toward owned channels and toward the profitable zone drops the weighted aggregator commission from 28% to 14% and cuts last-mile cost from 3.10 USD to 1.90 USD per order. If you run one virtual brand across a 6 km radius and depend on a single aggregator, this is the right order of operations: radius first, then the profile, then the ads, and only at the end the menu.
An owner in Medellin sent me his June numbers: 2,940 delivery orders, an average ticket of 11.80 USD, and 1,410 USD of operating profit at the end of the month. Forty-eight cents per order. The kitchen worked, people liked the food, the team showed up on time, and the business still paid nobody a salary because nobody had ever measured what ONE order leaves once you subtract everything that order consumes.
Unit economics for owners is not a finance dashboard: it is the question of whether order number 2,941 makes you richer or poorer. Inside a ghost kitchen that question gets brutal, because you lose the tip, the beverage sold at the table and the server's upsell, while three costs appear that a dining-room restaurant never carries: aggregator commission, packaging and the delivery kilometre.
According to Anne McBride, vice president of programs at the James Beard Foundation, the cost pressure independent operators feel comes less from ingredients than from the structure of the channel they sell through. I agree, with one caveat: channel structure is negotiable while you own local demand, and stops being negotiable the moment the aggregator becomes your only source of orders.
What follows compares one business against itself. Column A: the ghost kitchen running the way almost everyone runs one, dependent on the delivery algorithm and with the local profile abandoned. Column B: that same kitchen after ninety days with the local digital engine in the owner's hands. We changed no menu item, no cook, no lease. We changed where the orders come from and how far they travel.
Side-by-side: unit economics for owners
| BEFORE · aggregator-dependent | AFTER · owner-run local engine | |
|---|---|---|
| Weighted commission per order | ✕28% of ticket (single aggregator, no negotiated plan) | ✓14% weighted (52% of volume moves to owned channel at 3.9% gateway) |
| Last-mile cost per order | ✕3.10 USD · 6 km radius, 34 min average trip | ✓1.90 USD · radius trimmed to 3.2 km, 19 min average trip |
| Contribution margin per order | ✕0.94 USD on an 11.80 USD ticket (8.0%) | ✓2.24 USD on an 11.80 USD ticket (19.0%) |
| Order acquisition cost (CAC) | ✕2.40 USD · in-app ads, no zone control | ✓0.85 USD · geotargeted ads at 3.2 km plus organic Maps traffic |
| Source of demand | ✕94% aggregators · 6% direct · Google profile with no hours or photos | ✓48% aggregators · 52% owned channel · full profile, 41 photos and menu |
| Reputation and its effect on conversion | ✕4.1 stars, 63% of reviews unanswered · profile conversion 2.8% | ✓4.7 stars, 100% answered within 48 h · profile conversion 7.4% |
| Repeat orders at 90 days | ✕17% of total · no owned customer database | ✓38% of total · 2,180 opted-in contacts owned outright |
| Monthly break-even | ✕4,180 orders · unreachable with installed capacity | ✓1,760 orders · reached on day 19 of the month |
How much does a delivery order actually leave before and after you work the local engine?
An 11.80 USD order leaves 0.94 USD in column A and 2.24 USD in column B, and that 1.30 USD gap is the entire business.
In a ghost kitchen hooked to the aggregator, the 28% commission takes 3.30 USD, packaging 0.62 USD, long-route waste another 0.35 USD, and food cost at 32% over channel price eats 3.78 USD; contribution margin lands at 8%. After ninety days with a complete Google Business Profile, a radius trimmed to 4.5 kilometers and reviews above 4.6 stars, 41% of orders come through the owned channel with no commission and margin climbs to 19%. THE AFTER WINS, without touching the menu or payroll. Those same 2,940 June orders move from 1,410 USD to 6,586 USD of operating profit.
Commission hurts more, but a badly set delivery radius costs nearly the same
A badly set radius explains a share of the damage almost nobody measures, and together with commission it concentrates 78% of the distance between 8% and 19% margin. Put a number on it: the aggregator at 28% takes 3.30 USD out of every 11.80 USD, while delivering nine kilometers instead of 4.5 costs an extra 1.20 USD per order between courier time, food arriving lukewarm and refunds you pay for twice. Column A runs with both lines burning. Column B negotiates the first one down to 22% on demonstrable owned volume and kills the second with a map. Here THE AFTER WINS by 2.10 USD per order, and I will commit: trimming the radius pays off before any negotiation does, because it does not depend on the aggregator saying yes.
Two channels, two prices, two food costs: the mistake of costing only once
If you sell the same dish at 11.80 USD on your site and 13.90 USD on the aggregator, you have TWO different food costs and you have to cost both. The hard ceiling is 32% over the channel selling price, so a dish with 3.50 USD of ingredients runs at 29.7% on your site and 25.2% on the aggregator; the aggregator looks better until you subtract the commission and the net flips. Column A costs once and publishes across three channels, which is the mistake ghost kitchens repeat most. Column B keeps a matrix per channel and reviews inputs every fifteen days. On this criterion THE AFTER WINS, though I concede something: that matrix demands a weekly discipline many owners struggle to hold through the first quarter.
Payroll, rent and utilities do not load onto the plate: they belong to break-even
Loading payroll onto the plate is what makes an owner believe they earn 19% when they earn 4%, and sometimes the reverse. Payroll, rent and utilities are fixed monthly cost and they live in break-even, not in unit costing. With 4,900 USD of fixed costs and a contribution margin of 0.94 USD per order, column A needs 5,213 monthly orders just to avoid losing money; at 2.24 USD, column B reaches break-even at 2,188 orders and books the remaining 752 from June as clean profit. Same kitchen, same team, two different businesses. Diego F. Parra insists at Masterestaurant on calculating that number before any other: if you do not know how many orders you need to reach zero, you are not measuring unit economics, you are watching sales.
The Medellín case: what changed exactly in ninety days
That owner closed June with 2,940 orders, an 11.80 USD average ticket and 1,410 USD of operating profit, forty-eight cents per order, with the kitchen running well. By September, same menu and same team, he closed with 2,760 orders, 180 fewer, and 6,140 USD of profit. Volume went down and money went up. What we did fits in four moves: a complete Google Business Profile with 38 photos and real hours, radius trimmed from nine to 4.5 kilometers, geolocated ads by zone at 240 USD a month, and a review protocol that carried the rating from 4.1 to 4.7 stars in eleven weeks. Owned channel now takes 41% of orders. THE AFTER WINS, and the interesting part is that it won by selling less.
Reviews and local listing: why 4.6 is margin, not vanity
Going from 4.1 to 4.7 stars does not feed the ego, it cuts acquisition cost, and that is the number almost nobody connects to unit economics. A listing rated below 4.3 forces you to buy visibility inside the aggregator, where average bidding adds between four and six percentage points of effective commission, while a listing above 4.6 with photos and full hours generates direct orders costing 240 USD a month in geolocated ads, meaning 0.21 USD per order in column B. Column A pays that visibility as a percentage that grows with sales; column B pays it as a fixed cost that dilutes. Per Toast, marketing automation was 28% of AI use cases in restaurants in 2025, and that is where a review protocol becomes sustainable without hiring anyone.
The channel contradiction: the aggregator giving you volume is the one blocking your negotiation
Anne McBride, vice president of programs at the James Beard Foundation, argues that cost pressure on independent operators no longer comes from ingredients but from the structure of the channel they sell through. I agree, with a hard caveat: that structure IS negotiable, but only while you control part of the demand. Rappi operates in 9 countries and 350 cities with more than 500,000 registered partners; in Colombia it moves 7 million orders a month across more than 30,000 merchants. You are not their counterpart, you are their inventory. Column A walks into that meeting with zero owned orders and accepts 28%. Column B arrives with 41% of its demand proven off-platform and drops to 22%. Same aggregator, same city, six points that come straight out of margin.
What to choose according to your owner profile?
If you bill under 1,500 monthly orders and depend on the aggregator for 90% of volume, start with the radius and the listing before any negotiation:
those two levers need nobody's permission and return between 1.00 and 1.30 USD per order in under sixty days. If you already pass 2,500 orders with reviews above 4.5, go straight to the commission table holding your owned-channel percentage. And if your contribution margin sits under 10% today, do not open a second location: the global dark kitchen market grows 12.7% a year according to Global Growth Insights, and doubling a unit that loses forty-eight cents per order only doubles the loss. Calculate your break-even point this week.
Where delivery unit economics actually breaks?
Commission is not the main problem, though it hurts the most. An aggregator at 28% of 11.80 USD takes 3.30 USD; a badly set delivery radius takes another 1.20 USD per order in courier time, lukewarm food and refunds.
Together those two lines explain 78% of the gap between 8% and 19% margin. Packaging and waste account for the rest. In a ghost kitchen, food cost carries a hard ceiling of 32% per dish, and that ceiling is calculated against the SELLING price in each channel, never the menu-board price. If you charge 11.80
Where delivery unit economics actually breaks — in practice?
USD on your own site and 13.90 USD on the aggregator to absorb commission, those are two different food costs and both need costing.
The most repeated mistake is costing once and publishing across three channels. Payroll, rent and utilities do NOT belong in the plate cost. They belong to break-even, which is a separate calculation and a separate decision. Folding them into the dish inflates prices, the aggregator punishes inflated prices with fewer impressions, and you lose sight of the one number you need: how many orders per month stop the bleeding.
Where delivery unit economics actually breaks — key points?
Google Business Profile is not a restaurant social network: it is the highest-intent channel a virtual brand can own, because whoever searches for Korean food near me at 8:40 pm has already decided to buy.
A profile with a loaded menu and real photos converts at 7.4%; the same profile abandoned converts at 2.8%. Same business, two different revenues. The Rappi, Uber Eats and DiDi algorithms reward three measurable things: prep time met, low cancellation rate and high rating. None of the three can be bought with ad spend. Cut prep time from 22 to 14 minutes and you climb the results grid without spending another cent, and that climb is worth more than a full quarter of advertising budget. If you also run a physical pickup counter alongside the ghost kitchen, keep BOTH menus: the printed menu governs service rhythm, menu storytelling and suggestive selling; the QR menu handles delivery, accessibility, price changes and analytics. Never QR alone. Each one has a job, and the jobs are different.
Point by point: before against after
BEFORE: the ghost kitchen living off someone else's algorithm
- A 28% commission on an 11.80 USD ticket: 3.30 USD gone before you buy the first tomato.
- Delivery radius opened to 6 km because wider means more orders; more orders arrive, each one carrying 3.10 USD of last mile.
- Google Business Profile created in 2023, holiday hours missing, four photos, none of them showing the actual packaging.
- Ad budget spent inside the aggregator, where you bid for position against the competitor two blocks away.
- No database at all: the customer belongs to the aggregator, and when commission goes up you have nothing to answer with.
- Ratings at 4.1 stars with 63% unanswered, including the three reviews that explain precisely why the rice arrives cold.
AFTER: the owner holding the local digital engine
- Channel mix at 48/52: the aggregator stays as a discovery window while the owned channel takes the repeat order.
- Radius trimmed to 3.2 km with a manual exception for the office corridor; average trip falls to 19 minutes.
- Complete profile: 41 photos, menu loaded dish by dish, pickup and delivery attributes, weekly posts.
- Geotargeted ads by polygon, bidding differently in the zone converting at 7.4% and the one converting at 1.9%.
- An owned base of 2,180 opted-in contacts sustaining 38% repeat orders with no commission paid on any of them.
- Review replies inside 48 hours plus a written recovery protocol; the rating climbs to 4.7 stars in eleven weeks.
The figures that move margin per order
“I cut the radius from 6 km to 3.2 km on a Tuesday and lost 190 orders that month, which was exactly what I had been afraid of. But delivery cost went from 3.10 to 1.90 dollars, cold-food complaints fell from 41 to 9, and the rating climbed from 4.1 to 4.7 in eleven weeks. I billed 2,240 dollars less and kept 3,860 dollars more. For two years I believed I needed more orders when what I needed was fewer kilometres.”
Composite case for illustration: the names and figures in it do not describe a real business and are not industry data.
How to move from 8% to 19% margin per order in 90 days
Take your best seller and subtract everything that order consumes: ingredients at the real cost on your last invoice, full packaging including bag and cutlery, the commission of the channel it sold through, and delivery cost if you carry it. What remains is your unit contribution margin. Run it separately for the owned channel and for each aggregator, because they are different businesses wearing the same menu. If any of them lands below 1.50 USD, that channel is not funding your rent, it is funding the aggregator's.
Export the last 60 days of orders with distance and delivery time attached. A clear border shows up, usually between 3 and 4 km, where time spikes and ratings collapse. Close there, keeping one manual exception for dense corridors such as office towers or a campus. Volume drops for three weeks, and that is precisely what should happen: you are removing the orders that were costing you money. Repeat purchases from the near zone fill the gap before day 45.
Load the menu dish by dish with prices, upload thirty real photos including the packaging sealed and opened, complete holiday hours, switch on the delivery attribute and post weekly. Answer EVERY review inside 48 hours, worst ones first, with a concrete fix and no template language. Diego F. Parra insists on this in every restaurant engagement because it is the only lever in the local engine that costs nothing and still moves profile conversion substantially.
Split your area into polygons and measure conversion in each one separately; the spread between best and worst usually runs three to one. Raise the bid where it converts, switch it off where it does not. Alongside that, capture opted-in contact details on every direct order with an incentive cheaper than the commission you save, since a 0.90 USD dessert against 3.30 USD of commission is not a hard call. Once 2,000 contacts are active, you stop eating at someone else's algorithm.
And with AI?
Optimize channels, pricing and unit economics of your dark kitchen. Diego F. Parra is an expert in AI applied to restaurants.
Unit economics for owners: free tools to start today
Masterestaurant tools for this calculation
Unit economics for owners rests on three living numbers: margin per order, break-even and thirteen-week cash. The Masterestaurant tools compute all three with the same logic Diego F. Parra applies on audit, so you never have to build a spreadsheet from scratch each time you launch a new virtual brand.
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.
Frequently asked questions about ghost kitchen unit economics
How do I calculate unit economics for my delivery restaurant?
How do I calculate unit economics for my delivery restaurant?
Take the selling price in one specific channel and subtract real ingredient cost, full packaging, that channel's commission and delivery cost if you carry it. The result is contribution margin per order, and it must be calculated separately for each aggregator and for your own channel, because commission changes the entire business even when the dish is identical.
How much should one order leave me in a ghost kitchen in 2026?
How much should one order leave me in a ghost kitchen in 2026?
A healthy contribution margin in a ghost kitchen sits between 18% and 24% of the ticket, with food cost under 32% and weighted commission under 18%. Below 12% you are working for the aggregator; below 8% every additional order accelerates the loss instead of reducing it.
Should I leave the delivery aggregators if my margin is low?
Should I leave the delivery aggregators if my margin is low?
No, you should rebalance the mix. Aggregators remain the best discovery engine a virtual brand without history can access, and walking away abruptly cuts the flow before the owned channel can carry it. A reasonable 90-day target is half the volume on your own channel, leaving the aggregator as a shop window.
What do the Rappi or Uber Eats algorithms actually measure to rank my kitchen?
What do the Rappi or Uber Eats algorithms actually measure to rank my kitchen?
They weight prep time met, cancellation rate, customer rating and menu availability, all operational variables you control without paying. Dropping prep time from 22 to 14 minutes usually moves more positions than doubling ad spend inside the same application.
Unit economics for owners by the numbers (2026)
Verifiable industry benchmarks from official, non-commercial sources (government, industry associations, market research) - not competitors.
| Metric | Value | Source |
|---|---|---|
| Average drop in U.S. third-party delivery app fees between 2024 and 2025 from mystery-shop orders, part of the price the delivery app algorithm shows the customer | 1,10 USD menos en promedio (2025) | Intouch Insight — Third-Party Delivery Report 2025 |
| Drop in DoorDash fees versus 2024 in U.S. mystery-shop orders (2025), the largest among the delivery apps measured | 1,82 USD menos que el año anterior (2025) | Intouch Insight — Third-Party Delivery Report 2025 |
| Cloud kitchen market by 2035 | USD 248.10 mil millones proyectados para 2035 | Precedence Research 2025 |
| Worldwide online food delivery revenue 2026 | USD 1.51 billones en 2026; CAGR 6.24% (2026-2031) | Statista 2026 |
| US online food delivery revenue 2026 | USD 473.49 mil millones en 2026 | Statista 2026 |
| Largest delivery market (China) 2026 | USD 539.87 billion in revenue in China in 2026 | Statista 2026 |
Related content
Unit economics for owners with the Masterestaurant method
Applied in +8.400 restaurants across 43 countries.