What Food Delivery Costs a Restaurant: Commission and Menu Price Analysis

Introduction

Ask what food delivery costs a restaurant and you get two very different kinds of answer. One is a commission percentage, quoted confidently in trade press and rarely sourced. The other is the price gap between the delivery app menu and the restaurant's own ordering page, which anyone can observe.

The first is a private contractual term. The second is public data. This post is about measuring the second properly, and about being honest that the first is not measurable by collection.

Two different costs, and only one is observable

Commission is the percentage a platform charges a restaurant on order value, set in a signed agreement. It varies by service tier, negotiated volume, promotional participation and market. It is not published, does not appear in any app response, and cannot be scraped.

Anyone claiming observed commission rates at scale is surveying restaurants, working from aggregated contract data, or inferring. Inference from prices does not work: a restaurant sets its delivery price using commission plus its own pricing judgement, packaging cost and competitive position. You cannot recover one input from the output.

Menu mark-up is the difference between what a restaurant charges for an item on a delivery platform and through its own channel. It is set by the restaurant, visible to any consumer, and measurable at scale.

Actowiz measures mark-up and the consumer-visible fee stack. It does not claim to know commission terms. That boundary is the difference between a dataset an analyst can defend and one they cannot.

What mark-up actually measures

Mark-up is a pricing decision, not a cost pass-through. A restaurant may price delivery above direct to recover platform economics, hold the two equal to protect brand consistency, or price below direct to win volume.

So mark-up answers a different question than commission: how much more does this restaurant charge a delivery customer than a direct customer, and how does that compare across the market? That is the variable operators actually control.

Measuring mark-up: the matched-pair design

The unit of measurement is a matched item pair: the same item, at the same restaurant outlet, observed on a delivery platform and on the restaurant's direct channel, within a short time window.

Every word there is load-bearing.

Same item. Not the same item name. Names differ between channels for one dish, and identical names sometimes denote different sizes. Size, quantity and configuration must agree.

Same restaurant outlet. Not the same brand. Chain prices vary by location. Comparing a delivery price at one branch against a national direct menu introduces geographic variance into a figure meant to isolate channel.

Direct channel. The restaurant's own site or first-party ordering system. An outlet with no direct online ordering has no comparable price and drops out. Say so explicitly: the measurable subset is not a random sample of all restaurants.

Short time window. Both prices captured close together, because promotional pricing moves on either channel. Same-day pairing is the working standard.

Building the matched pair set

Step 5 is where most published comparisons quietly break. If the delivery listing bundles a side into the base price and the direct menu charges separately, the base prices are not comparable. Computing minimum purchasable price on both sides — base plus cheapest option in every mandatory group — puts them on one footing.

The per-pair metric is a ratio, not a difference: delivery minimum purchasable price over direct. Ratios aggregate sensibly across items at different absolute price levels; differences do not.

Report at outlet level as a median across that outlet's matched pairs, then aggregate outlets. The median protects against one mismatched pair dominating an outlet's figure.

The consumer fee stack

Mark-up is only part of what a delivery customer pays above the direct price. The rest sits in fees, and all of it is visible at checkout.

Two design points.

Several of these are basket-dependent, so they cannot be a single number per outlet. Define a reference basket — a fixed subtotal, or a fixed matched item set — compute the fee stack against it, and publish the basket definition alongside the figures.

Surge requires cadence. A once-daily observation cannot characterise a fee that moves with demand. Sample the same outlets at fixed intraday slots across weekdays and weekends. Observed intraday range of busy-area fee, by city is a real deliverable; a spot reading is not.

Total consumer premium is then menu mark-up, plus the fee stack against the reference basket, plus any tax differential. Report the three separately rather than as one blended percentage: a restaurant can act on mark-up and cannot act on the platform's service fee.

How mark-up varies, and why

The interesting output is not the headline average. It is the variation.

By chain size. Large chains generally run more deliberate multi-channel pricing than independents, because they have the pricing function to do it. Whether that means higher or lower observed mark-up is an empirical question per market, and should be reported as median mark-up ratio by outlet-count band rather than asserted.

By cuisine. Categories with thin margins and low ticket sizes face different arithmetic than high-ticket categories. Report by cuisine tag class, using the platform's own taxonomy, mapped to a stable internal one.

By city. Competitive intensity, platform mix and delivery cost structures all differ. City-level medians with sample-size disclosure are the right unit.

By platform. Where an outlet lists on several platforms, mark-up can differ per platform for the same outlet. This within-outlet, across-platform comparison is the cleanest read in the whole dataset, because outlet, item and time are all held constant.

Every one of these cuts needs sample size published next to it. A cuisine-by-city cell with few matched outlets is not a finding.

What you do with the result

A restaurant operator benchmarks its own mark-up against comparable outlets in its city and cuisine, then asks whether its direct channel is priced attractively enough to shift volume. If direct and delivery prices are identical, the operator offers customers no price incentive to order direct.

A multi-outlet brand checks internal consistency. Divergent mark-up across its own branches usually signals menu-management drift rather than deliberate strategy.

An analyst or consultant builds a menu-price index by city and cuisine, tracked over time. The index measures observed consumer price levels across channels, not platform revenue or margins.

A platform or investor reads the distribution of mark-up as a signal of how restaurants are absorbing delivery economics, and the fee stack as a component of consumer-facing price competitiveness.

A diligence team uses mark-up and fee data as one input to unit-economics modelling alongside what the target supplies directly. Observed prices constrain the model; they do not replace the target's own numbers.

The honest framing is the same throughout. You are measuring prices and fees a consumer can see, at defined outlets, over a defined period, with stated coverage. You are not reporting contract terms, and you say so.

Actowiz Solutions builds matched item-pair datasets across delivery platforms and restaurant direct channels, with configuration-aligned minimum purchasable prices and promotion flags.

The fee stack is captured against a documented reference basket, with intraday cadence where surge pricing is in scope.

Frequently Asked Questions

Can commission rates charged by delivery platforms be scraped?

No. Commission is a private contractual term between platform and restaurant and does not appear in any public response. Only the restaurant's menu prices and the consumer-visible fee stack are observable.

What is menu mark-up and how is it measured?

It is the price difference for the same item at the same outlet between a delivery platform and the restaurant's own ordering channel, measured as a ratio of minimum purchasable prices captured within a short time window.

Why exclude promotional prices from a mark-up calculation?

Because a discount on either channel makes the pair non-comparable for a structural pricing measure. Promotions are captured and reported separately rather than netted into the base differential.

You can also reach us for all your mobile app scraping, data collection, web scraping, and instant data scraper service requirements!