Product ratings on most platforms concentrate heavily at the highest value, with a smaller cluster at the lowest and very little between. That shape comes from who submits ratings rather than from product quality.
Rating is voluntary and unevenly motivated
Most buyers who are moderately satisfied never leave feedback, because a product working as expected produces no impulse to act.
Strong reactions in either direction do produce that impulse, which is why the distribution has peaks at the ends and a thin middle.
The average therefore describes the opinions of people motivated enough to submit, which is a different population from purchasers as a whole.
Prompts change who responds
Automated requests for feedback increase submissions from satisfied customers, since dissatisfied ones were already likely to respond without prompting.
Timing matters as well. A request sent shortly after delivery captures first impressions, before the problems that only appear with extended use have emerged.
Platforms that prompt heavily therefore report higher averages than those that do not, independent of what is being sold.
Returns remove unhappy buyers from the sample
A buyer who returns a product usually resolves their dissatisfaction through the refund and never rates it at all, so the worst experiences leave the dataset entirely rather than appearing as low scores.
Where returns are easy, this effect is stronger, which means a generous returns policy raises the visible rating while the underlying complaint is handled privately and never recorded publicly.
Support resolution works the same way. A problem fixed quickly by a manufacturer often converts a would-be low rating into no rating, since the impulse to complain fades once the issue is settled.
Aggregation hides important structure
An average combines ratings of different variants, sizes and colours, and sometimes of successive product generations sharing a listing.
A revised version with fixed problems inherits the old ratings, and a product that has deteriorated retains credit from earlier batches.
Reading the distribution and the recency of reviews reveals more than the headline figure, because the shape shows whether opinion is genuinely consistent or merely averaged.
Compression at the top removes information
When most products sit within a narrow band near the maximum, small differences in average carry weight they cannot really support.
Buyers respond by reading the negative reviews specifically, which is a rational way to recover the detail that averaging destroyed.
That behaviour is why the content of low ratings is more useful than their count: it describes failure modes, which is the information a numeric average cannot convey.