How do they estimate it without an early indication of demand?
Publisher's experience & they would probably reach out to bookshop chains to see what the possible demand will be. If you under-estimate, you can do a second run quite cheaply and even market the first as 'sold out'. If you over-estimate, you are stuck with a lot of 'remainders'.
I saw a figure the other day that Aston's book sold 60,000, which is very high (he was initially expecting 20,000).
But of course there are exceptions (probably many). There was a landmark book
'Birds of Tasmania', advertised with pre-orders about 8 months before launch, so well before printing. But they kept the number fixed, as it was a premium product. They sold the entire print run in pre-orders before it was launched!!
Three 'significant' books I've been involved with (ie not self published!) -
1) Tasmanian heritage topic: Print run of 2,000 softcover and 100 hardcover collector's edition' based on the most we could get for our budget & type of book we wanted. It sold out and we are on second run of 1,000 soft cover. The hardcover was priced at strong premium and paid for the entire print run!
2) Tasmanian colonial biography: Print run of 500 softcover and 50 hardcover based on estimated demand - after 3 years, still about 200 softs and 25 hardcovers left. Not selling.
3) Tasmanian heritage topic: Print run of 1,000 softcover and 200 hardcover. We thought they would run off the shelves - popular topic (Tasmanian Heritage), nice coffee table book, strong authors. Its struggling, but selling consistently, if slowly.