Preference vs desire: how do you predict what people want (to read)?

Posted on 24/09/2015


If holidays are an exercise in escapism, then reading for pleasure while away is doubly so. It’s an extreme attempt to distance ourselves from everyday life: flying to an exciting destination just to escape all over again through a good book.

Irony aside, holidays are when a lot of people buy (and probably also read) books. Each time holiday season approaches, publishers and vendors battle to get their words in the hands of as many as possible. But once your holiday is booked, the difficult question (for me, anyway) is what do I want to read?

Listicles and best of articles are fine for suggesting things that you should read. But assuming that you’re not willing to slog through the classics / rely on the prevailing winds of the best sellers, there aren’t that many different ways to uncover what you want to read.

That’s not to say there aren’t plenty of recommendation websites and forums; it’s just that they work in largely the same way.

This holiday I checked out the following:

They’re all pretty good at generating suggestions (my favourite being TasteKid: seamless user interface), but very similar in terms of how they work. Their suggestions are based around:

  1. Recommending similar, popular books to ones you’ve read
  2. Other books within the same theme/genre grouping
  3. Forums offering peer to peer recommendations

Likewise, Amazon recommendations and reviews provide the same service based on what you’ve purchased previously.

What I’ve not yet seen are different kinds of recommendation engines. It’s one thing being able to predict a novel within the same genre by a similar/ the same author, but after reading a thriller about a missing woman, the last thing I want to read is another thriller about another missing woman.

There have to be more ways to curate, recommend and identify books. A way that merges both the predictive ability of Amazon, with an understanding of human behaviour and culture. A predictive recommendation engine that encourages discovery and variation.

A few sources of inspiration, from other spaces/industries:

Yossarian lives

A visual search engine for uncovering seemingly unconnected concepts. It returns “disparate results with shared attributes”, providing greater variety of visual definitions than a Google Image search.


Rather than waiting for searches, Foursquare’s sensitive geolocation algorithm pre-empts needs, based on where you are, the time of day and the places you’ve liked in the past.

(At least, in it’s former life…). Before it was taken down, Xyo had the ability to find relevant apps, by topic, by function and by device operating sytem. It even found equivalent apps across OS.

Ben Evans has written about the next challenge facing search: not responding to needs, but predicting them. And then being able to fulfil them, instantly, if possible. There’s got to be a commercial model for this. We’ll just have to see what emerges next year…