NPR recently published a blog post on how to judge which book will sell well – editor judgment or data from ebooks. It’s an interesting 21st century problem along with the fact that traditional publishers lose money on 80% of the books they publish according to the article.
There are, of course, a lot of problems with strictly using data to determine whether a book is worth publishing – which I should note no one in the article says is a good idea. It ignores any segment of the reading population that doesn’t read ebooks – perhaps they would buy and enjoy a book that readers of ebooks didn’t like.
My other major objection is that a book that does well in terms of the data might lead publishers to publish a number of books that an algorithm says are similar. This will mean a spate of books that are all the same which history tells us readers do NOT want. Everyone should remember the Twilight publishing phenomenon where suddenly there were eight million YA paranormal romance novels and none of them were the “next Twilight” because readers had already moved on.
There is likely a place for data mining in the publishing realm but I suspect that it is a small piece of a very large puzzle of what readers want and when so that another 500 years from now it will still remain as much a mystery as it is today.