How Gig App Ratings Actually Work

A rating is an average of a small number of opinions, and averages behave in ways that feel unjust. One low score sits in your number until enough other jobs arrive to outweigh it.

You cannot argue it away. You can dilute it, and dilution is a volume problem.

What the score is built from, and over what window

The score is the average of the ratings customers chose to leave, calculated across a rolling set of your recent jobs. Rolling is the word doing the work: old ratings fall off the back, which is why a number can move on a day you did nothing unusual.

What you need from your own platform is one sentence: is the window counted in jobs or in days, and how many. That sentence is in the app's rating screen or its help center. Read the actual wording rather than taking another worker's version of it, because the two platforms in the same parking lot may count differently.

The second thing to understand is the denominator. Not every customer rates. When only a handful of people have scored you, each new opinion swings the average hard, and the number settles down as the pile grows. Early volatility is arithmetic, not a verdict on you.

Ratings and completion metrics are two different systems

The star rating measures what customers thought. Completion rate, acceptance rate, on-time rate and cancellation rate measure what you did, counted automatically with no human involved.

They live on different screens and they carry different consequences. A night that ends in you canceling a job touches a behavior metric. A night that ends in a customer tapping one star touches the rating. Treating them as one number leads people to fix the wrong thing.

Both feed into account standing, but through separate doors, and which door matters more is set by the platform — what actually triggers a gig app deactivation is where those thresholds get examined.

What customers actually respond to

Customers rate the handover. Whatever happened in the last thirty seconds is what gets scored, and a surprising amount of what they attribute to you was never yours: cold food, a missing drink, an item the shop substituted, a wait caused by a kitchen you stood in helplessly.

The things drivers agonize over barely register. Your route. Your car. Whether you took the ramp or the stairs. Whether you sent a second update message.

What does register is whether you did the specific thing the customer asked for. The delivery note is the whole test. Leave at door means leave it and go. Do not knock means do not knock. Flat 4B means 4B, not the lobby. A photo that clearly shows the door number closes the loop, because it answers the complaint before it is made.

The small things that move it, and the effortful things that do not

Moves it: reading the drop-off note before you set off rather than in the dark on arrival. Checking at pickup that the sealed bag matches the item count. Messaging once when something changes, not three times. Putting the bag down exactly where the note says. Taking a photo that a stranger could use to find the door.

Does not move it: apologizing for the restaurant, driving faster, upgrading your bag, chatting, or waiting at the door hoping to be seen doing a good job.

The other lever is which jobs you accept at all, because the offers you take decide which customers get to rate you. A drop-off into a building you cannot find at night carries a risk that no amount of politeness offsets, and choosing well starts with reading a delivery offer screen before you accept.

Recovering a dropped score

Recovery is dilution, so the only lever is completed jobs with good outcomes. There is no reset button and no way to delete a rating by working harder on the next one.

Which means the practical move after a drop is a run of low-risk work: short trips, addresses you know, well-lit and well-signed buildings, daylight if you can. Complicated jobs pay the same and carry more ways to go wrong while your average is thin.

Some platforms allow you to contest a rating tied to something outside your control, with evidence — a timestamped photo, a support thread about a restaurant delay. Find that form once, use it where you genuinely have proof, and do not fire it off at every low score, because a pattern of disputes is itself a record.

How long it takes depends on your own job volume and on the size of the window, and anybody quoting you a fixed number of days is guessing. If you are early enough that the score barely exists yet, what a first shift on a delivery app is really like is the more useful read.

The metrics you cannot influence, and should stop watching

Some numbers on the stats screen are descriptions of demand, not of you. How many offers you were sent. What mix of jobs arrived. Which zone you were shown. Staring at those tells you about the platform's evening, and you cannot change any of it from the driver's seat.

Add to that list any screenshot another worker shows you. You do not know their zone, their hours, their vehicle, their window length, or whether the image is real.

And hold onto this one: a healthy rating is not the same as a job worth doing. If your score is fine, your metrics are clean, and the work still is not covering what it costs you to be out there, the number was never the problem — deciding when a gig app has stopped being worth it is the harder question underneath.

Stop refreshing the stats page. It updates on the platform's schedule, not on yours, and the minutes you spend watching it are the exact minutes that would have moved it.