Delivery app ranking: before vs after with the Masterestaurant framework

Chapter 1 — Either the listing converts or the algorithm buries it

Aggregators do not hand out visibility out of goodwill: they push impressions toward the listing that turns more views into orders, because every impression served to a restaurant converting at 11% earns them twice the commission of one served to a restaurant converting at 5%, and that arithmetic outranks any promotional budget you care to push. With online food delivery in Mexico at US$ 9.22 billion during 2024 and a 14.66% CAGR according to Statista (2024), the fight for that impression gets harder every quarter. Your first six menu items carry nearly all the weight: your own photograph rather than a stock image, a description stating portion weight and side, a readable price with no discount stamped over it. When I review a stalled listing, the problem is almost never ad spend; it is that the hero dish photo was uploaded in 2022 and nobody looked at it again.

Chapter 2 — Your declared prep time is a contract with the algorithm

Declaring 18 minutes of prep and delivering in 29 punishes you twice, since it drags down the order rating and pushes your listing lower in the estimated-time sort, which is the favourite filter of the hurried 1:15 p.m. customer. It sounds backwards, yet capping capacity at peak —closing order intake for twenty minutes once the line is already flooded— trades volume for position, and position pays you back over the following weeks. Turn it around: if you accept twelve simultaneous orders your kitchen serves in twenty-nine minutes, you win those twelve and lose the ranking on the next six hundred. In a market like Latin America, which moved USD 12,917.3 million in 2024 with a projected 8.6% CAGR through 2030 according to Grand View Research (2025), giving away position to save one night is a bad trade. A three-unit metropolitan operator showed me his dashboard with 41% of orders tied to co-funded promotions and asked why sales climbed while cash fell; the answer lived in channel arithmetic, not in the algorithm.

Chapter 3 — Permanent discounting destroys your price signal

Aggregator commission, half the discount and packaging all bite into contribution margin before the first plate leaves the line, and once 41% of demand gets used to a marked-down price, that price becomes the real one and the menu price turns decorative. The fix was not killing promotions overnight: we built two channel-exclusive items at 27% food cost, portioned for the container, priced on their own, and cut co-funding below 15% of orders across eleven weeks. Sales gave up 6%. Contribution margin gained 9 points. Holding 4.7 stars has little to do with replying to reviews and everything to do with three decisions made in the kitchen: packaging that arrives hot and without spillage, a complete order, and a missing item resolved before the customer writes. Every percentage point of orders with an incident drags the rating down slowly and recovery takes months, because an average is defended with fresh volume rather than apologies.

Chapter 4 — A 4.7 rating is an operating floor, not a target

The Masterestaurant framework I apply sorts the channel into four components —listing, operations, pricing and reputation— and ties each one to Prime Cost and break-even, which is the language a board actually understands. In a sector where cloud kitchens moved USD 80.3 billion in 2025 according to Grand View Research, with USD 88.7 billion projected for 2026, reputation is the only channel asset you cannot rent. Below US$ 500 thousand a year the channel has to be profitable from the first order because there is no cushion: two exclusive items, zero co-funding, commission negotiated against a minimum volume, and delivery kept under 25% of total sales. Between 500 thousand and one million the first channel-dedicated person appears, and the classic mistake there is adding platforms before mastering one. From one to five million a separate virtual brand with its own menu starts to pay, because diluting the parent brand in discounts costs more than opening a fresh listing.

Chapter 5 — Every revenue band plays a different game

Above five million the problem changes nature: commission negotiation moves from 28-30% into the 18-22% range and those points fund an in-house data cell. Past ten million, the group must measure contribution by platform, by unit and by daypart, or it will be subsidising entire cities without knowing it. A celebrity-chef restaurant or a large-format themed venue above five million a year faces a cost the small band never sees: the reputational risk of serving badly through a channel it does not control. When the chef's name is the asset, one cold delivery photographed on social media outweighs the three thousand orders that produced it, which is why these groups tend to run their own fleet for the short radius and leave customer acquisition to the aggregator, knowingly paying for double logistics. Add packaging: moving a high-ticket plate without it collapsing demands materials costing three to five times the standard container, plus an assembly step that occupies one person at peak.

Chapter 6 — The high end pays costs nobody budgets for

I was wrong for years telling these houses to simply stay out of delivery; what changed my mind was watching the channel hold 20% of sales with margin intact once the menu was redesigned instead of trimmed. Launching a virtual brand from the same kitchen makes sense when idle capacity is measurable in hours and when the new menu draws on 70% of the inventory you already buy; outside those two conditions, what looks like a second revenue stream is a second waste stream. The global virtual restaurant market reached US$ 66.3 billion in 2024 and is projected at US$ 140.4 billion by 2033 according to Verified Market Reports (2024), which explains the enthusiasm, though enthusiasm does not pay break-even. The honest test is to shut the virtual brand down for three weeks and watch whether parent-brand sales rise: if they rise, you did not create demand, you moved it between pockets.

Chapter 7 — Virtual brands: when they add and when they cannibalise

When it genuinely adds, the signal is clean, because the virtual brand captures dayparts the main one never touched. Ask for four numbers per platform and per unit, weekly, no exceptions: view-to-order conversion, real prep minutes against declared minutes, share of orders with an incident, and contribution margin after commission, co-funding and packaging. That last one is the figure almost nobody calculates and the only one that answers whether the channel funds your operation or bleeds it. With ghost kitchens in Asia-Pacific projected from US$ 21.73 billion in 2024 to US$ 60.59 billion in 2032 at a 12.8% CAGR according to Coherent Market Insights (2024), the channel will keep growing with or without you, and growing inside it without that dashboard means gambling with the house cash. Start this week with one thing: replace the photographs of your six most-viewed items and measure conversion fourteen days later.

Chapter 8 — What separates an operation that climbs the ranking from one that pays to be there

The aggregator does not sell visibility, it sells conversion probability. When your listing turns 11% of views into orders, the algorithm shows it more because each served impression earns more commission; at 5%, no ad budget fixes that arithmetic. So the work starts with the photo, the description and the price of your first six items, not with the promotion budget. Prep time is a contractual promise to the algorithm. Declaring 18 minutes and delivering in 29 punishes twice: order rating drops and the listing slides down the estimated-time sort, which is the filter a hurried customer relies on most. Capping peak capacity hurts volume and pays in position. Permanent discounting destroys the price signal. A sustained 25% cut across four months re-anchors willingness to pay: when you switch it off, the order drop exceeds the lift it created, and contribution margin never fully recovers. I prefer a tactical 48-hour offer attached to a low food cost dish.

Chapter 9 — What separates an operation that climbs the ranking from one that pays to be there — in practice

Reputation is an asset with a half-life. A 4.8 rating built over ninety days absorbs an isolated incident; a 4.4 with unanswered reviews turns any incident into a three-position drop inside the category, and clawing it back costs six to ten weeks of clean operation. The digital channel has its own break-even and almost nobody calculates it. With 28% effective commission, packaging at 9% of the ticket and 30% food cost, what remains for payroll and rent is thin; at 21% effective commission and channel food cost under 27%, the same volume pays structure. The difference is not the algorithm: it is channel menu engineering. Depending on one app is pure territory risk. A commission change, a listing suspension over incidents, or a competitor entering your polygon with discount capital can move 80% of your digital sales in a week if you have no direct channel and no worked Google Business Profile.