Rappi delivery strategy: the mistake that costs six margin points vs the right method

Which one wins: per-dish costing with the commission loaded in, or the gross sales figure on the dashboard?

Dish-by-dish costing with the commission inside the recipe card wins, and the gap between the two methods is measured in margin points, not opinions.

An owner reading gross sales through the channel confuses it with what actually lands in the register, because a share of it goes to commission depending on the contracted plan, the category and the paid visibility inside the app —the market's comparable aggregator tiers run up to a ceiling near 30%, per DoorDash for Merchants and Restaurant Business Online—. With the commission loaded onto the dish, the same pasta that shows a healthy contribution margin in the dining room shrinks sharply in the channel, and that shrunken figure is the one worth using to set price, portion size and whether the item stays or goes. Averages comfort you. The recipe card decides.

Overhead line at the bottom of the P&L versus commission inside the recipe card

Booking the commission as overhead at the foot of the income statement is the architectural mistake that props up everything else, because a single blended average hides that pasta returns several times what the premium burger returns once commission, packaging and transport shrink are taken out. What you are looking at is not a business: it is a weighted average of whatever sold most last month. Move that same cost down into the recipe card, item by item, and the conversation changes in one working afternoon: a handful of dishes crossing below the margin floor come off the menu, a few prices go up, and channel margin climbs WITHOUT a single extra peso in sales. The P&L never lies, but it does not confess either; you have to go fetch the number at the level where the decision gets made, which is the dish.

An 80-item mirror menu against an 18-item delivery menu

Short menus win the channel, and they win on conversion rather than on operational romance. Publishing the full dining-room menu on the aggregator assumes more options sell more, and user behavior says otherwise: people decide in under a minute, with their thumb, comparing photos, so a long menu breeds paralysis and abandoned carts. Every item that turns once a month also forces you to hold live inventory, and that inventory becomes dated shrink. With 18 items picked for margin after commission and for how well they survive a motorbike, ticket size rises, dispatch time falls, and the kitchen stops fighting two simultaneous menus at peak. Pull whatever does not travel: breaded fish that arrives soggy costs you twice, once in product and once in the rating that drags behind you for weeks.

Owned channel against aggregator: who pays the commission, who keeps the customer

Direct ordering wins on economics and on data ownership, even while it loses on initial reach, which is why the right strategy runs both with clearly separate jobs. Some 58% of customers prefer ordering through the restaurant's own app or website (NCR Voyix, via Restaurant Dive, 2024), a figure that contradicts the belief that the diner is married to the aggregator. On your own channel you keep the phone number, the order history and the frequency; on the aggregator you keep volume and nothing else, because the customer belongs to the platform. Most limited-service operators already offer delivery, per the National Restaurant Association (2025), so nobody is walking away from the channel. Use it to fill dead hours, and work in parallel to move the repeat buyer home.

The case: two locations, one month, two costing methods

For example, a chef-driven restaurant with two locations was billing a large sum every month, a meaningful share of it through the aggregator, and reporting flat profit even as sales grew year over year. Per-dish costing with the commission loaded in exposed what the average had buried: several channel items were running at a bare-minimum contribution margin, and the signature dish sat among the most ordered, pushed there by its photo and by the in-app visibility promotion they were paying for. They pulled those underperforming items, kept the rest of the lineup, set channel prices above dining-room prices and renegotiated packaging down to a lower cost per unit. Some weeks later channel sales were down modestly and channel contribution margin had moved up significantly. The quarter closed ahead on profit, on lower volume.

Mirror pricing against channel pricing: the argument almost nobody wants to have

Differentiated pricing wins, and the standing objection —«the customer will notice»— is true and changes nothing. They notice, they compare, and they keep ordering, because at nine at night they are buying convenience, not price. Holding a mirror price means handing over a meaningful share of your margin to protect a perception the market has already normalized across the region. That said, the adjustment takes craft: you do not raise every dish equally, nor by a large jump in one move, you raise where elasticity allows it, typically a modest amount, and you protect the anchor dish that brings in new users. Diego F. Parra insists on a sequence that is not up for debate at Masterestaurant: clean the menu first, adjust the price after. Do it backwards and you end up making expensive the very items you were supposed to pull.

What happens if the aggregator raises its commission five points next quarter?

If your channel margin sits at 12%, five extra commission points leave you at 7%, and at 7% you are working for free for a platform that also controls the customer and the rating.

That is the question to answer BEFORE signing the visibility plan, not after the pricing-change email lands. Run it with your own numbers: take the five best sellers in the channel, subtract five points from each one's current margin and count how many stay above 25%. Whatever cannot survive that scenario is not a delivery item, it is a dining-room item somebody published without costing. A business that only holds up under the intermediary's current fee has no delivery strategy: it has a dependency, and dependencies get revised when the creditor decides, never when it suits you.

What to pick, based on how you actually operate?

If you run your own kitchen and bill between 20,000 and 120,000 USD a month, cost every item with the commission loaded in and treat delivery as a separate business unit:

its own menu, its own break-even, its own pricing. That is the verdict and it carries no caveats inside that range. If you bill under 20,000 USD and your kitchen sits at half capacity, an aggregator with a short menu is an excellent occupancy engine, and there you can live for a while on 25% channel margins while building repeat business. If you are already at full installed capacity at peak, every aggregator order is stealing a table that pays twice as much, and the call is to shrink the channel or open dedicated production. The cloud kitchen market is projected at 203.72 billion USD by 2033 (Grand View Research), which is precisely the exit for that third profile.

Where the business actually breaks?

The difference is not price, it is the ARCHITECTURE of cost.

A restaurant booking Rappi commission as a general expense at the bottom of the P&L will never know which dish wins and which loses, because the average hides that pasta dishes typically return a far healthier margin than a premium burger, per Sauce (2025). Load it dish by dish inside the spec sheet and the conversation changes in one afternoon: four references come off, three prices go up, and channel margin climbs six points without selling a single peso more. The second break point is the mirror menu. Publishing a bloated menu on a delivery aggregator does not raise conversion; it lowers it, because the user decides in under a minute and a long menu produces paralysis.

Where the business actually breaks — in practice?

Every reference selling once a month also forces inventory, generates shrink and occupies kitchen attention at peak. Fourteen well-chosen references rotate faster, buy better and cook quicker.

The third is time. Rappi penalises missed promises with less algorithmic visibility, so declaring a delivery time and consistently missing it is not a service problem, it is a traffic problem: the storefront falls down the listing and sales drop while the owner has no idea why. Declaring a realistic delivery time and hitting it consistently is worth more than any paid in-app campaign. And there is a tension almost nobody resolves: the channel cannibalises the dining room, yet it also protects it. It cannibalises when the kitchen cannot cope at peak and the seated guest waits an extra twenty minutes; it protects when it fills the 15:30 hole and pays the fixed afternoon payroll.

Where the business actually breaks — key points?

The resolution is operational, not philosophical: switch the storefront off in the slot where the kitchen is near full capacity, switch it on where it has plenty of room to spare.

That decision comes from the tickets-by-slot report, not from intuition. On digital menus Masterestaurant is blunt: the QR is a complement, never a replacement. The PHYSICAL menu stays in the dining room because it controls service pace, carries the menu narrative and enables the server's suggestive selling; the QR contributes in delivery, accessibility, price updates and analytics. Operators who scrapped the printed menu to save on printing lost a share of table average ticket, and recovered it once they printed again.