Ratings on retail sites and reader communities rarely average anywhere near the middle of the scale. They bunch towards the upper end, and the reason is selection rather than generosity.
Ratings come from a filtered group
Only people who bought and read a book can rate it, and people mostly buy books they expect to like. Genre, cover, blurb and recommendation all filter the audience before page one.
A reader who dislikes a book's premise never becomes a rater of it. The dissatisfied population largely stays outside the sample.
What is measured is therefore satisfaction among the self-selected, not quality assessed across all readers.
The stronger a book's signalling, the tighter that filter becomes, which is why tightly branded genre titles carry the highest averages of all.
Abandonment removes the low scores
Readers who stop after forty pages usually rate nothing at all. There is no obligation to score a book and little social reward for scoring one you did not finish.
Those abandonments are exactly the experiences that would produce one and two star entries, so the scores least likely to be recorded are the lowest ones.
A completed book carries a built-in bias: finishing it is already a mild endorsement.
The scale is used as a recommendation
In practice most raters do not use the full range. Five is treated as approval, four as approval with reservations, and anything at the bottom as a deliberate statement.
Because low scores read as hostile rather than merely descriptive, many readers avoid them or skip rating altogether.
The scale effectively compresses into its top half, which is why differences of a tenth of a point carry weight readers would not expect.
Awareness of the author compresses it further still, since rating a debut novelist harshly feels personal in a way that rating a film does not.
Early ratings behave differently
The first ratings a book receives often come from advance readers, review programmes and existing followers of the author, all groups predisposed towards it.
General readers arrive later and pull the average down, which is why a score frequently drifts downwards in the months after publication.
Reading a score usefully
The average matters less than the shape and volume behind it. A book with thousands of ratings has a broad audience; one with a few dozen may only have been read by people close to it.
The written reviews at the lower end are usually more informative than the number, because they say what the book actually is rather than whether a favourable audience liked it.