Your dashboard says everything is fine and the orders are still not arriving
Restaurant chains reach this question on a quiet evening that should not have been quiet, and the reason is structural rather than a fault in any particular tool. Every system a restaurant owns reports its own state. The point of sale reports the point of sale, the integration reports the connection, the tablet reports the tablet. The storefront a customer opens is not one of those systems, it belongs to the platform, and it can be switched off without any of your systems being told. Green everywhere and dark outside is the normal condition of that arrangement, not an anomaly.
What exactly is your dashboard reporting?
Its own field of view, accurately. An integration monitor that says the connection is healthy is telling the truth about the connection. A point of sale showing no errors is telling the truth about itself. A tablet showing an open store is telling the truth about what it last knew.
None of them is a claim about what a customer sees, because none of them is looking there. The storefront is a page on somebody else’s website, rendered from somebody else’s database, with a state your systems do not hold.
What can turn the storefront off without your systems knowing?
Most of the available mechanisms, once listed.
The platform can close it on its own account. Careem has an “Outlet Closed” state only Careem can lift. Deliveroo has a “Forced Closure” the partner cannot reverse.
An automatic rule can close it. Deliveroo closes a site after three automatic rejections inside fifteen minutes. Just Eat takes it down after one unaccepted order. DoorDash deactivates after about five minutes of tablet silence. Talabat can hold a shop closed past its scheduled opening if nobody acknowledges the opening within thirty minutes.
A person can close it. Anyone with the tablet can pause orders in a rush and not unpause them.
And a store can be dark in a way that is not closure at all. Keeta documents a platform override under which delivery is removed while the store stays open for collection, so the card looks normal and the delivery order cannot be placed.
Your connection is intact through every one of those. Nothing happened to it.
Why does nobody notice for hours?
Because the absence of orders looks like ordinary quiet, and quiet is not an alarm condition anywhere.
An empty half hour on a Tuesday is unremarkable. So is a soft evening. There is no threshold in any internal system that distinguishes “nobody is ordering” from “nobody can order”, because internal systems cannot see the difference either. The first signal is usually a customer complaint or a franchisee’s phone call, and both arrive late.
The measurements bear this out. Mean interruptions in our UAE panel ran from 1 hour 50 minutes to 3 hours 34 minutes depending on the platform, and a large share of them began late in the evening, which is exactly when the fewest people are watching.
How would you tell the difference?
Look at the storefront the way a customer does, from a phone that is not signed in, at an address the branch serves. That single check settles it, and it is the check nobody performs at 22:40 on a Tuesday.
Doing that continuously, for every branch on every platform, is the entire problem. Twenty branches across four platforms is eighty storefronts, and nobody opens eighty apps.
Is this a failure of the integration?
No, and framing it that way sends the fix to the wrong place. An integration is doing its job when it moves orders and reports its own health. It was never designed to answer a question about what the platform is displaying, and adding that to it would mean building a second, different measurement inside it.
The two measurements are complements. One tells you your plumbing is sound. The other tells you whether customers can order.
What does the outside view actually change?
It converts an invisible loss into an event with a start, an end and a length, which is the difference between a soft week and an incident somebody can act on.
That also makes the loss countable. A month of intervals gives lost trading hours per branch per platform, and with your own orders per hour and average order value that becomes money. Across our UAE panel the average listing lost 9.5 trading hours in July 2026, close to a full trading day, and none of that appeared in any internal report.
Kitchain (kitchain.co) takes that measurement from the customer side, which is why it needs no access to any of the systems that were already saying everything was fine.