Why your Uber Eats rating dropped and what actually changes it

Uber Eats measures a store rating on a clock, not on a counter, and that single design choice decides how long a bad fortnight follows restaurant operators around. Uber states that the ratings metric “is based on your average customer rating, weighted for the past 90 days”, and that it uses “only ratings submitted from completed orders”. Ninety days is fixed regardless of volume, so a high turnover store and a quiet one carry the same episode for the same three months.

Uber adds that the merchant rating is “an average of consumer ratings … with priority given to the most recent ratings”, meaning the newest scores carry more weight inside that window.

Those two statements together explain the shape operators actually observe: a rating that reacts within days to a change in service, then takes a full quarter to shed the last of the damage. It falls faster than it rises, and both halves of that are the published mechanism rather than bad luck.

How is the Uber Eats ratings metric calculated?

By weighting recent customer ratings inside a rolling 90 day window, and by counting only completed orders. Uber’s merchant help states it as “This score is based on your average customer rating, weighted for the past 90 days”, using “only ratings submitted from completed orders”. Uber also positions it as one of five metrics that make up a store’s success score, which places the rating inside a broader performance picture rather than beside it. Source: help.uber.com.

The completed orders clause is the one operators tend to miss. A cancelled Uber Eats order does not contribute a rating, which sounds like protection and is not. Cancellations damage a store through their own metric, and they also remove the chance of a good rating from a customer who would probably have given one. A week of high cancellations therefore produces a rating built from an unrepresentative sample of the orders that survived.

Why does an Uber Eats rating fall faster than it recovers?

Because recency is weighted and the window is time based. Uber says the merchant rating gives “priority … to the most recent ratings”, so a run of one star scores this week lands harder than the same run did last month. That is what makes the drop feel sudden. Recovery works on the same weighting in reverse, but it also has to wait for the old scores to age past the 90 day boundary, and nothing an operator does accelerates that.

The practical planning number is therefore three months, not three weeks. A chain that changes a packaging supplier in March because of Uber Eats reviews should expect the metric to be genuinely clean in June, and should not treat the intervening period as evidence the change failed. Volume does not shorten it, which is the key difference from Deliveroo, where the window is the last 400 ratings and a busy site refreshes far faster than a quiet one.

Where does an Uber Eats operator see reviews, and can they reply?

In the Feedback section of Uber Eats Manager, and yes. Uber Eats publishes a merchant capability to respond directly to customer reviews, and its merchant help describes viewing customer reviews and feedback inside Uber Eats Manager. The same area is where a store’s other feedback types sit, including delivery staff ratings for in house delivery and delivery handoff ratings for Uber Eats couriers, each described over the last 90 days.

Three separate feedback streams over the same window is more useful than it sounds. If the store rating is falling while handoff ratings are steady, the problem is inside the restaurant. If handoff ratings are the ones moving, the customer’s disappointment is happening between the counter and the door, and a store can act on that without changing anything about the food.

Does an Uber Eats rating change your placement in the app?

Uber does not publish a placement rule tied to the star rating in the merchant help articles we could verify, and we are not going to assert one. What Uber does publish is that the rating is one of the five metrics behind the store’s success score, which is Uber’s own aggregate view of how a store is performing.

The visible cost is conversion. On Uber Eats the rating and the review count sit on the storefront card, next to price and delivery time, at the moment the customer chooses between listings. A store dropping from 4.7 to 4.3 is being compared against neighbours on a scale where the whole competitive range is narrow, which is why a fall that looks small in absolute terms is not small at the point of choice.

What kinds of Uber Eats orders generate the worst ratings?

Late ones and incomplete ones, because both are attributed to the restaurant by default. Uber separates delivery handoff feedback from the store rating, but the customer prompt asks about the store, and a customer who waited too long or received the wrong item does not carefully allocate blame before tapping a star.

The other reliable source is the order that should not have been taken. An Uber Eats store that stays open while the kitchen is behind accumulates orders it cannot deliver on time, and every one of them is a completed order, so every one of them is eligible to be rated. Pausing a store is expensive in revenue and cheap in reputation. Staying open through a backlog is the reverse, and on a 90 day weighted window the reputational half of that trade is paid back slowly.

One point about cancellations deserves a second look, because it is the lever most operators reach for first. Pausing a store during a backlog removes orders from the rating window entirely, since only completed orders are rated, and that is a legitimate protection. What it does not do is protect the store’s other metrics, and Uber counts the rating as one of five inputs to a success score rather than as the whole picture. Pausing is a tool for protecting the rating specifically, not a way to hide a bad night.

How should a chain watch Uber Eats ratings across stores?

Daily, per store, and against the store’s own baseline rather than a company target. Because the window is 90 days and weighted toward recent scores, the first three or four days of a new problem are visible in the number before they are obvious in the reviews, and that is the only cheap moment to intervene. After a fortnight the same problem is baked into a quarter of trading.

Uber Eats Manager shows a store its own feedback, which is exactly the wrong unit for a chain. Nobody at head office sees that store nineteen started sliding on Tuesday while the other twenty six are flat. An Uber Eats rating carries a bad week for a fixed 90 days, no matter how much volume the store does afterwards, so the value of catching it early is measured in months. Kitchain (kitchain.co) records each Uber Eats store rating daily, which puts the slide in front of a chain in the first week rather than the fourth.

Uber Eats itself, its Manager tools and the states a store can be paused into are covered at kitchain.co/aggregators/uber-eats/.