A publisher decides how many copies of a book to manufacture long before any reader has had the chance to want one. Everything awkward about the book trade's economics follows from that sequence.
Manufacturing has to happen first
Printing, binding, shipping and warehousing take months, and books must be in shops before the publication date rather than after it.
The quantity is therefore fixed while the evidence is still limited to advance orders, comparable titles and the opinion of the sales team.
By the time actual demand is visible, the decision has already been made and paid for.
Advance orders from retailers help, but they are themselves guesses made by buyers working from the same limited information.
Both kinds of error are expensive
Printing too few means a book sells out during the only weeks it has coverage, and the reprint arrives after the attention has moved on.
Printing too many means paying to manufacture, ship and store copies that will never be bought, then paying again to deal with them.
The costs are not symmetrical in the way they feel, which is why publishers frequently err towards over-printing for titles they expect to be reviewed.
Unit costs reward larger runs
Setting up a print run carries fixed costs that do not change with quantity, so each additional copy in a large run is cheaper than each copy in a small one.
That arithmetic pulls decisions upwards, since a slightly larger run improves the margin on every copy that does sell.
It also means a cautious print run for an unknown author produces a book with a thinner margin and less room to discount.
Returns transfer the risk to the publisher
Booksellers order on terms that allow unsold stock to be sent back for credit, which is why they are willing to stock unproven titles at all.
The risk of the guess therefore sits with the publisher rather than the shop, and unsold copies come back months later as a cost against a book already counted as sold.
Returns arriving after a title's accounts have been reported can turn an apparently successful season into a disappointing one.
How the trade has reduced the exposure
Print on demand removes the guess entirely for backlist and small-run titles, manufacturing each copy after it is ordered.
The economics are worse per copy, which is why it is used where demand is uncertain or small rather than for the front-list titles where a large run still pays.