In this article
Your venue has been on Uber Eats, Deliveroo or Just Eat for months, and orders barely trickle in. The problem is rarely that you're not 'on' the platform — it's that the platform ranks you somewhere near the bottom of the search results for your postcode or category, a place almost nobody scrolls to. Each of the three big delivery platforms ranks restaurants with its own algorithm, and according to the platforms' own merchant documentation, that algorithm watches a handful of things most venues have never deliberately set up. This article counts up seven of those numbers.
This isn't an article about cutting your platform commission — that already exists on this site, and it's about what you pay the delivery platform. It isn't about your own website ordering channel either. This article is about something almost no venue has ever looked up: how the platform decides which restaurants a guest even gets to see inside their postcode or category, and in what order.
That isn't guesswork. Unlike a Google ranking, which has been speculated about for years, delivery platforms publish parts of their ranking mechanics themselves — in their merchant help centres, in their own 'success score' policies, and occasionally even in their own engineering team's blog posts. The problem isn't that the information is secret. The problem is that almost no venue has ever gone looking for it.
The numbers in this article come from that official documentation. Uber Eats and Deliveroo are the platforms actually used across most of the EU; DoorDash doesn't operate in most European markets, but it's the one platform of the three that publishes exact percentages for what a photo, a description or a cancellation does to your revenue. Those figures illustrate a mechanism that, according to Uber Eats' and Deliveroo's own — more loosely worded — documentation, applies to them just as much: which factors count is broadly the same across all three, even where not every platform makes every exact percentage public.
Seven numbers, in this order: what a photo per dish does to your revenue; the rating threshold that puts you in a separate, more visible category; how long you have to accept an order before the platform auto-rejects it; the cancellation ceiling that drops you out of the featured list; how many separate signals Uber Eats itself names in its 'success score' — without saying how heavily each one is weighted; what a plain menu description adds, on top of the photo; and what sponsored placement costs, on top of the commission you're already paying.
Why almost no venue knows this algorithm
Most owners treat a delivery platform like a phone book: your venue is listed, so guests find you. That stopped being true once these apps moved from an alphabetical or chronological list to a personalised ranking — every guest sees a different order, recalculated on the fly around whatever the platform currently judges most likely to lead to an order.
Uber's own engineering team has publicly explained that the Uber Eats ranking system optimises for exactly one thing: the probability that a guest who sees your listing actually orders — your conversion rate. The more people who see your listing and go on to order, the more often the platform shows you to more people. That isn't a secret formula, it's the platform's own stated, published objective — and yet almost no venue has ever looked at its own conversion rate, let alone tried to improve it.
The rest of this article counts up what most venues have never looked up: what a photo does to revenue, which rating puts you in a separate, more visible category, how long you have to respond before the platform auto-rejects you, which cancellation ceiling drops you out of the featured list, what Uber Eats itself says it measures, what a description adds on top of a photo — and what sponsored placement actually costs.
Free guide Everything about digitalisation in your restaurant, in one guide From point-of-sale to delivery apps — the complete guide to the technology in your venue. Read the guide7 numbers that decide whether a guest ever sees you
Every number below is traced, by name and by source, to the named platform's own merchant documentation at the time of writing. Platforms occasionally change these programmes and thresholds without a major announcement — check the platform's current help page if in doubt. Where a platform doesn't make an exact percentage public, this article says so plainly rather than inventing one.
1. 44%: what a plain photo per dish does to your revenue
In 2022 DoorDash published the results of its own study across more than 15,000 local merchants: menus where every item has its own photo book up to 44% more revenue per month than menus without. A header image at the top of the storefront adds up to another 50% on top of that, and a logo up to 23% — both separate from the item photos themselves.
That figure isn't disconnected from ranking. Because Uber Eats' algorithm optimises for conversion rate (see above), and a photo demonstrably pushes more people over the line to order, a missing photo works against you twice: you sell less to the people who do see you, and the platform shows you to fewer people in the first place as a result. Deliveroo confirms the same mechanism in its own partner communications: menu photos can lift orders by up to 25%.
This isn't about one nice atmosphere shot at the top of your menu. It's about coverage — what share of your individual dishes has its own recognisable photo. A venue with ten photos across a forty-dish menu has 25% coverage, and loses the revenue of every unphotographed dish at exactly the moment a hesitant guest is scrolling past it.
Same venue, same postcode, same category — the only difference is whether the five free numbers (1 through 6, minus the paid option) are set up.
Illustrative — the exact order a guest actually sees is recalculated per guest and also depends on distance, availability and that guest's own ordering history. What doesn't change: every number above pushes in the same direction.
2. 4.6 stars: the rating threshold that puts you in a separate, more visible list
Uber Eats' own 'Top Eats' programme requires a customer rating of at least 4.6 stars (alongside a handful of operational minimums) to qualify. Venues that clear it get a separate Top Eats badge next to their name and a more prominent spot in results — not just 'a bit higher', but a whole separate, more visible display category.
DoorDash runs a comparable but separate programme, 'Most Loved': a lifetime average rating of at least 4.5 stars, combined with a low cancellation rate (see number 4), unlocks a dedicated 'Most Loved' carousel on the app's home screen — a slot venues without the badge simply never appear in, whatever their ordinary ranking is.
The point isn't that 4.2 stars is 'bad' — that's a perfectly good rating. The point is that it sits just under a hard threshold that unlocks a completely different display tier. A restaurant on 4.7 stars and one on 4.4 stars might differ by a few places in the ordinary list; in the premium category, one appears and the other simply doesn't.
3. 10 minutes: how long you have to accept an order
Deliveroo's own help centre is unambiguous here: if you don't accept or reject an incoming order within 10 minutes, the system auto-rejects it, so a guest isn't left waiting indefinitely for confirmation. An auto-rejection counts, in practice, exactly the same as a cancellation you make yourself (see number 4).
Uber Eats goes a step further than Deliveroo here: according to Uber's own documentation, a venue can be paused entirely — invisible to every nearby guest, shown as 'currently unavailable' — once several orders in a row go unanswered. That isn't 'ranked a bit lower': that's off the list entirely, until you manually unpause or it reopens automatically (by default at 6am the next day).
The practical fix is written into the same documentation: turn on auto-accept. That shows the guest their order as 'confirmed' immediately, skipping the 'confirming your order with {venue name}' window that makes a slow response so visible in the first place.
4. 1.1%: the cancellation ceiling that drops you out of the featured list
DoorDash's 'Most Loved' programme (see number 2) sets an explicit ceiling: your avoidable cancellation rate — orders you cancel for reasons within your control, like wrong opening hours or a dish you'd simply run out of — has to stay under 1.1% to appear in the carousel. Above that, you drop out of that featured slot even if your rating is otherwise fine.
Separately, DoorDash runs its own operational standard for 'downtime' — the time your store is fully unavailable during your own listed opening hours: the target for a well-performing store sits at 1% or lower over a rolling 90-day window, with a hard ceiling around 20% for so-called 'virtual brands' before automatic restrictions kick in.
The distinction between 'avoidable' and 'unavoidable' is the whole point here. A cancellation because a courier never showed up typically doesn't count against you. A cancellation because your opening hours are wrong in the system, you never marked a dish as sold out, or you were simply too busy to pick up — does count, and is exactly the kind of cancellation a few minutes of setup avoids.
5. 5 signals, no published weight: what Uber's 'Success Score' actually measures
Uber Eats rolls part of its ranking input into its own per-store 'Success Score', documented in Uber's official merchant policy. Under 'operational excellence' it names four concrete signals: the share of orders you fail to fulfil, delay in prep time that keeps the courier waiting, the share of orders with an accuracy issue (wrong or missing item), and the share of orders with a taste-or-quality complaint attributed to your store.
On top of that, the Success Score counts two separate components that have nothing to do with 'delivering on time': how many of your menu items have a photo and a description (see numbers 1 and 6), and how customers rate your store (see number 2). Named explicitly, that's six separate signals — four operational, plus menu coverage, plus rating.
What Uber doesn't publish is how heavily each of those six signals is weighted relative to the others. That isn't an omission in this article — it's literally what the platform itself does and doesn't disclose. The honest conclusion is: you now know WHAT is measured, not EXACTLY how much each part counts. That's precisely why the calculator further down this article is explicitly an educational tool, not a copy of Uber's undisclosed formula.
Two of the seven numbers are a measured revenue percentage. The rest are factors the platform names explicitly, without attaching a weight to them.
Sources: DoorDash's own 2022 study of 15,000+ merchants for the two measured percentages; Uber Eats' 'Success Score' policy and Top Eats terms, and DoorDash's 'Most Loved' terms for the rest. Where no percentage is shown, the platform itself doesn't publish one either.
6. 18%: what a plain description adds, on top of the photo
The same DoorDash study behind the 44% figure in number 1 also measures the effect of a good menu description, separate from the photo: items with a clear description — what's in it, how it's prepared, what the portion includes — book up to 18% more revenue per month than items without. That's a separate effect on top of the photo, not double-counted with number 1.
The reason is straightforward and stated plainly in DoorDash's own merchant documentation: a photo shows shape, colour and portion size faster than text can, but a name alone ('The Special') tells a guest who doesn't know your venue nothing about what's actually in it. A description removes exactly that uncertainty, at the moment a guest is deciding between your dish and the one next to it.
Categorisation counts here too: a menu split into clear sections (starters, mains, drinks) rather than one long list is exactly the kind of 'menu coverage' Uber's Success Score counts as a separate signal (see number 5) — independent of what any individual photo or description earns on its own.
7. Up to 30%: what sponsored placement costs, on top of the commission you already pay
Everything above is organic ranking — free to improve, just by getting your data right. There's also paid placement: Uber Eats' 'Sponsored Listings' lets you pay separately, on top of your standard commission, for a prominent spot at the top of the app's home feed, through a cost-per-click model where you set your own weekly budget and bid per click.
That paid placement sits on top of your ordinary commission, not instead of it. Uber Eats' own commission tiers run, depending on the plan you choose, up to roughly 30% of order value on the most extensive package — sponsoring is a fully separate budget stacked on top of that, not an alternative to it.
In early tests, Uber itself reported roughly €5 in revenue for every euro of ad spend — a figure that comes from the platform itself, from an early testing phase, and should be read with the caution that implies. The thing worth remembering: sponsoring buys visibility, it doesn't buy the six free factors above — a venue with a weak rating and few photos that also starts sponsoring is paying to show a weak listing to more people, not to improve it.
Calculate your own visibility score
Fill in the five numbers below as they stand for your venue today, and get a score out of 100 — plus the one change that would win back the most points right now.
This score is an educational tool built on the seven numbers above — not a copy of Uber Eats', Deliveroo's or DoorDash's real, undisclosed formula. No platform publishes its full ranking formula (see number 5); this is the best estimate that can be built from the figures that ARE published.
Your visibility score
Five numbers, one score, one action.
Biggest opportunity:
Educational estimate based on published figures — not an official score from Uber Eats, Deliveroo, Just Eat or DoorDash.
This calculator covers the five numbers you can act on today — not the other two (acceptance speed as a hard threshold, and paid placement), which behave more like an on/off gate or a separate budget than a scale you improve gradually.
What the score doesn't measure: how consistently you hold these five points over time. One good month after a bad rating doesn't immediately shift your average — most platforms work off a rolling average across several months, precisely so a temporary spike doesn't count as a structural improvement.
Where to start tomorrow
Three steps, in the order that earns the most for the time they cost.
1. Check your menu against the two free, measured numbers
- Go through your full menu on every platform you're on and count how many dishes have no photo — that's number 1, the biggest measured lever.
- Add one sentence describing what's in it to every item without a description — that's number 6, a separate effect on top of the photo.
- Calculate your own visibility score with the tool above to see which of the two would earn you the most.
2. Fix the two hard thresholds
- Turn on auto-accept where available, so you never hit the 10-minute limit from number 3.
- Review why you cancelled orders over the past month, and fix the two most common, avoidable causes (wrong opening hours, un-marked sold-out items) — that's number 4.
- Check whether your rating sits above 4.6; if not, ask specifically for a rating from guests who just left happy (number 2).
3. Consider sponsored placement only after that
- Don't spend on ads before the five free numbers are in order — sponsoring only shows a weak listing to more people (number 7).
- Start with a small, separate budget and measure yourself how many extra orders it earns, separate from your regular commission.
- Repeat this whole check every quarter — platforms occasionally change their thresholds and programmes without a major announcement.
The short answer
Delivery apps rank restaurants with an algorithm that, according to the platforms themselves, mainly watches conversion rate — driven most strongly by photos per dish, your rating, how fast you accept orders, your cancellation rate, and whether you deliver on time and correctly. Menu descriptions and sponsored placement sit on top of that.
Two of those seven numbers are a hard-measured revenue percentage (44% for photos, 18% for descriptions); the rest are factors the platforms themselves name without attaching a weight to them. That distinction is exactly why this article never invents a figure the platforms themselves don't publish.
The fix usually costs no money — just the time to go through your menu for photos and descriptions, turn on auto-accept, and track your cancellations for a quarter. Sponsored placement is the one of the seven that does cost money, and it's best used only after the other six are in order.