How much revenue restaurants lose when a delivery app takes them offline
A delivery platform can take a restaurant offline without telling anyone, and for restaurant operators the loss is easier to state in hours than in percentages. Across more than 12,000 monitored listings in the UAE in July 2026, the average listing lost 9.5 trading hours in a single month. That is a full trading day per listing, per month, on one market. Multiply it by the number of platforms a chain sells on and the figure stops being a rounding error.
Those hours are counted only inside the hours each location told the platform it would trade. A branch closed on schedule is not downtime. A branch that told the platform it would be open, and was not orderable, is.
How many trading hours does a restaurant lose to delivery downtime in a month?
9.5 hours per listing per month is the UAE figure for July 2026, and a listing means one restaurant location on one delivery platform. That is the unit to build everything else on, because it is the unit a brand can multiply.
A single location selling on three platforms is three listings, so roughly 28 hours a month. Twenty locations on three platforms is sixty listings, so around 570 hours a month, which is the equivalent of one branch being shut for the entire month and then some. Annualised, one listing loses in the region of 114 hours, close to two full working weeks of trading time.
None of that arithmetic requires a new measurement. It is one published average multiplied by a count of listings that any operator already knows. What it does require is the willingness to count listings rather than restaurants, because a brand that thinks in locations will underestimate its own exposure by exactly the number of platforms it sells on.
The averages hide a real spread. In the same UAE month a large share of listings recorded no downtime at all, while a small group lost more than twenty hours each. The distribution is set out in our UAE delivery downtime report. The practical reading is that the average brand does not experience the average. Most of the loss lands on a minority of listings, so the first question for any operator is not what the market average is but which of their own listings are in the tail.
Which delivery platform costs a restaurant the most trading time?
In the UAE in July 2026 the order was Careem Food at 2.94 percent of stated trading hours offline, then Deliveroo at 2.28 percent, Talabat at 1.81 percent, Keeta at 1.02 percent and noon Food at 0.74 percent. In Kuwait the same month the order was Snoonu at 1.88 percent, Deliveroo at 0.55 percent, Jahez at 0.28 percent, Keeta at 0.11 percent and Talabat at 0.09 percent. Panel wide, the UAE ran at 1.68 percent, Saudi Arabia at 2.88 percent and Kuwait at 0.54 percent.
Two things in those lists are worth more than the ranking itself. The first is that the same platform behaves very differently in different markets. Talabat is the third worst performer in the UAE and the best in Kuwait, by a factor of twenty. Deliveroo is four times worse in the UAE than in Kuwait. A brand operating in both countries cannot carry a single expectation about a platform across the border.
The second is that the same restaurants sit on several of these platforms at once. When one platform’s listings are offline four times as often as another’s in the same city, in the same month, for the same kitchens, the difference is not being produced by the kitchens.
Do short frequent outages cost more than rare long ones?
They produce completely different numbers, and averaging them together destroys the information. Kuwait in July 2026 shows both shapes in one market. Snoonu lost 1.88 percent of stated trading hours in interruptions averaging 23 minutes. Jahez lost 0.28 percent in interruptions averaging 10 hours 39 minutes. Two of the platforms in that panel entered the market recently, so their figures cover a shorter trading history than the others and are reported exactly as measured. This is a picture of a market still settling rather than a verdict on any operator.
There is a second reason to use the per platform numbers rather than a blended one. Snoonu accounts for roughly nine in ten of all offline incidents in the Kuwait panel, so any single average incident length for that market describes Snoonu rather than Kuwait.
By total trading time Snoonu costs almost seven times more. By the damage of a single event Jahez costs far more, because ten and a half hours removes an entire day of a branch’s delivery trade rather than a slice of one service. The fixes are different too. Frequent short pauses are a question of who is allowed to pause a branch and whether anyone confirms it came back. Rare long outages are a question of detection, since nearly all of the cost accrues in the hours before anyone noticed.
The same split shows up between markets. The Saudi panel lost 31,357 trading hours in July 2026 at a headline rate of 2.88 percent, but excluding listings that were offline for more than 90 percent of the month, the profile of a location that has left the platform while its listing stays in the catalogue, the rate for the rest of the panel was 0.77 percent. Three quarters of that headline number was dead listings rather than live outages, which is a catalogue problem with a completely different owner and a completely different fix.
How do we turn lost hours into a money number for our own restaurants?
With one multiplication and one number that has to come from your own accounts. Lost trading hours on a listing, multiplied by the delivery revenue that listing normally produces per trading hour, gives the gross revenue that did not happen. The revenue per trading hour is available to every operator from platform settlement reports divided by the hours that location actually traded, and it is different for every location, every platform and every daypart.
We publish lost hours and deliberately do not publish a revenue rate, because we do not measure order values. Any figure we printed here would be an assumption about somebody else’s business dressed up as a measurement, and this page would then be one more source of a number nobody can trace. The general per hour loss figures that circulate online have the same problem in a worse form, since most of them describe total revenue for full service restaurants in the United States rather than the delivery channel of a Gulf quick service brand.
Two adjustments make the multiplication more honest. First, an average hourly rate understates the loss, because outages are not spread evenly across the day. Several of the mechanisms that close a listing fire under load, such as Deliveroo closing a site after three auto rejections inside fifteen consecutive minutes, so lost hours skew towards the hours that earn most. Second, a share of the lost order is not recovered elsewhere. A customer who cannot order at 20:40 does not reliably come back at 21:10, and on most platforms the listing gives them no reason to.
How would we know our own number?
Not from the platform dashboards, and this is the part operators find hardest to accept. A merchant portal reports the current state of a store rather than the history of that state. Nothing in it says how long the branch was unorderable last Tuesday, and several of the mechanisms that produce downtime are recorded by the platform as ordinary closures rather than as faults. Deliveroo files an automatic closure after rejected orders under the same code as a day off the partner set on purpose.
The measurement that produces a defensible number has three properties. It is taken from the customer side, on the public storefront, so it sees what a customer saw. It is checked repeatedly around the clock rather than sampled, so the start and end of each interruption are timestamped. And it is compared against the hours the location itself published, so a branch closed on schedule never counts as downtime. That is the method behind the 9.5 hours above, and it is what Kitchain (kitchain.co) runs across delivery platforms in the UAE, Saudi Arabia, Kuwait and the UK.
The full monthly figures, methodology and limitations are published for each market: UAE, Saudi Arabia and Kuwait. If the goal is a claim rather than a budget line, the evidence has to be built differently again, and we set out what that takes in how to claim for delivery platform downtime.