05/10/06
The Master Plan
Abebooks has announced a project to revise its search facilities. No big news, right?
Wrong. Abe will (and indeed, reading beyond the headline, is) be using the data collected by Librarything to create better recommendations. What does this mean? Librarything allows users to catalogue their books, and to create virtual libraries. As the data in these libraries is centralised, LibraryThing then has fun running comparisons across the data sets, and drawing conclusions about one user’s taste based on all other user’s taste. So, if I’ve read All’s Fair and Big If, then LibraryThing knows that OK, I may be a security risk to the US President, but that I may also enjoy The Testament of Yves Gundrun (which, actually, I acquired for Canongate years ago). And that’s pretty cool.
What makes it different to Amazon’s ‘people who shopped for’ or ‘people who bought’ is that it is based on book reading rather than shopping. It’s based on passion and taste. And that’s what makes it valuable.
LibraryThing is the purest example of ‘collaborative filtering’, i.e. using other people to (jargon alert) recommend stuff they like to you from the ‘long tail’ or millions of products. It doesn’t make these decisions based on newness, press coverage or bestseller status: it is probably more likely to recommend a book that was published years ago than last week. Which means it is meritocratic, personal and extremely valuable.
Abe are doing a very smart thing here and it will give them an edge. It is exactly how they should leverage their stake in LibraryThing, and what is cool about it is that Amazon can’t compete with it, and whilst publishers should be able to they can’t either.
This is very exciting and I can’t wait to see how it pans out. The only odd thing is the marriage with Abe: it is surely a better fit with Amazon. Who knows, maybe where this is going will lead to Abe moving more into focus on Amazon’s acquisition radar. If they can nail how to get people to buy more books that they enjoy more, Amazon doesn’t care if they are new or second hand. On paper, it could be the holy grail…
RSS feed for comments on this post. TrackBack URL
# Comment by Tim @ 3:14 pm, October 5, 2006:
Hey. First, LibraryThing is very public about what we do, so there’s no between the lines yet. (And, note that Abe does not and never will receive user-specific data.) As for fit, LibraryThing’s ability to “drive users down the tail” is part of the charm for Abe. Thus, one of the standard recommendations for a Harry Potter book is Madeline L’Engle’s _A Wrinkle in Time_, which isn’t selling so well this week but is on a lot of “Thingamabrarian” bookshelves, and on those of Abe’s sellers.
Thanks for the mention!
# Comment by David Kessler @ 7:57 pm, August 3, 2008:
Similar taste data based on readers’ express opinions is bound to be stronger than “customer’s who bought this…” type data for two reasons. First of all purchasing data indicates only what customers expect - their expectations may well be disappointed. Secondly many people have multiple interests and there is no basis for the premise that those interests follow a pattern So looking for a pattern may be an exercise in self-deception.
This new approach is very promising.