A feedback score
Everyone is rated somewhere between 97 and 100, so the number has run out of room to separate anybody. That is not a flaw in this market, it is what happens to every rating system where leaving a bad score costs the person leaving it something.
- What it is
- An average of ratings left by buyers
- Where you meet it
- On listings and vendor pages
- What it settles
- Very little. It is compressed near the top
- Cannot tell you
- How this order will go
The specimen
vendor B rating 98.9% orders 1,377 disputes 6.4%
the ratings are basically the same number. the last column is not.
Two sellers who look almost identical on the field people read. The rating separates them by three tenths of a percent. The dispute rate separates them by a factor of eight.
Why the number compresses
Leaving a poor rating is not free. It can affect how the seller treats a pending order, it invites an argument, and it takes effort at a moment when the buyer mostly wants to stop thinking about it. Leaving a good one costs nothing and ends the interaction cleanly.
Multiply that by every transaction and the scale collapses upward. What is left is not a measure of quality, it is a measure of how many people wanted to be done.
What to read instead
| Field | Why it survives the pressure |
|---|---|
| Dispute rate | Opening a dispute takes real effort, so the number is not inflated by politeness |
| Account age | Cannot be bought or rushed |
| Recent activity | A long dormant account that reappeared is a different thing from a busy one |
| The text of poor reviews | The few that exist are worth more than the thousands that do not |
What your own feedback is for
It is one of the few things you can put back. A short factual review of an order that went badly, written after the process finished rather than during it, is worth more to the next reader than another line at the top of the pile. Written during a dispute it becomes leverage, which is visible to everyone and works against you.
The thing no score can do
Predict a decision. A record describes how somebody has behaved and stops being predictive the moment they choose otherwise, and that choice is not visible from outside. Which brings this room back to where it started: size the order so being wrong about any single seller is survivable, because no number on the page can carry that weight for you.
What a very new account with a perfect score means
Almost nothing, in either direction. Twelve orders at a hundred percent is a small sample and a small sample sitting at the ceiling is exactly what you would expect from any account, honest or not. Age is the field that cannot be manufactured quickly, which is why it survives this argument while the rating does not.
The related pattern worth noticing is a long dormant account that has suddenly become busy again. It has all the history and none of the recent behaviour, and accounts change hands quietly precisely because history is the valuable part.