Netflix, the popular movie rental service, has a reputation for being able to recommend good movies – whatever good means to you. Netflix’s trick has been to create more than 75,000 of their own “altgenres”, microniches with subjects such as Mind Bending Cult Horror Movies from the 1980s, Visually Striking Goofy Action & Adventure and Romantic Chinese Crime Movies. Netflix assembled these categories by paying trained viewers to watch movies, score those movies for everything from plot and setting to political and sexual content and feeding the results into a machine to produce personalised recommendations.

Online book recommendations have been stuck in the star rut for a while – that is, the limited, scalar constraint of giving everything one to five stars. There’s a certain amount you can do with this, like thisbooknext.com, which scrapes ratings from a number of websites and recommends books based on the familiar other-people-like-you-like-this principle, but the well-known weakness of Amazon’s own recommendation system, based on user ratings, suggests this has limited value. Scribd, which recently launched its own ebook subscription service, is the first bookseller to try out the microgenre approach, combining human ratings with algorithms to create its own niches like Novels for Art Lovers, Strangers in Strange Lands and On the Run from the Third Reich. Apple recently acquired a company called BookLamp planning on doing the same thing, so expect more of this in the near future.

The flipside of Netflix’s approach is that a complete reading of Hollywood’s output allows you to spot the gaps. Many put the success of Netflix’s version of House of Cards down to the fact that it knew exactly who it was making it for, down to their possibly subconscious preferences. As the use of algorithmically augmented categorisation creeps into the book world, we may start seeing more specialised but more targeted titles appearing. A biblical martial arts fantasy for hopeless romantics set in the Edwardian era, anyone?