I come across a lot of websites that have recommendation engines based on your history, or your past purchases or what you favourites are and then they pimp similar stuff back to you. I guess the notion is based on the idea that they can find stuff related to what you ‘like’ and thus not just improve the experience for you, but also increase their sales as you are more likely to make a purchase. Amazon is probably the most shining example of this where the very pristine and fresh home page changes to show what you have browsed before and what you might thus like, with every subsequent visit.
The main area on ‘my’ Amazon.com home page lists 8 main sections. These are ‘Frequently Bought Together’ which shows an image of a book I had looked at earlier today along with two other books, ‘More to Explore’ which also shows a book I looked at recently along with related books, ‘Stay Dry with a New Raincoat’ – which is the one section unrelated to my ‘Amazon Past’, ‘Customer With Similar Searches Purchased’ – which another name for #1 & #2 above. ‘Recommended for you’ which seems to be based on purchase history as it shows an Action Figure I bought for a friend and a Shaver I bought 5 months back. ‘New for You’ which is again a recycled #1. And finally ‘Recommended for you based on your browsing history’ and ‘Inspired by your Wish list’.
A whopping 7 out of 8 sections are ‘personalised’ and pay no attention to stuff which I might want to try out or explore.
I think there are two variants of personalisation: a) Explicit & b) Implicit. The Explicit variant is more common with personalised homepage service such as Netvibes where I specify what news sources I want to get my news from, my widgets and more. E-commerce sites however, cannot use explicit personalisation – they in turn use my explicit choices – such as a my searches and past purchases to make a guess at what else I might like.
I think the biggest problem with this is that it overlooks the exploratory tendencies in people. I like rock music, I like Bollywood – sure I’d like to explore more of the same, but then sometimes I want to try something completely new – recently a friend introduced me to Jazz – which I thought was pretty interesting too – why couldn’t Amazon recommend something like that? I think often people themselves don’t know what they like and would like to sample something new rather than more of the same.
Unfortunately, pretty much all services – from the Pandora’s to the Last.fm of the world empower only incremental exploration by showing related stuff, but haven’t cracked the question of how to encourage people to try out something radically different.
Another issue is the assumption that I want to buy everything I searched for (Personalisation at Amazon appears to based on favourites, past purchases and searches of which the last seems the most unreliable to me). A friend of mine once asked me to buy a Belly Dancing DVD for her, next I know I have 4 other similar titles showing up in ‘You Might Also Like…’ on Amazon! I think search related recommendations are best served on the Search Result pages – which provide related results within that context.
I really wish Amazon would stop deciding my future based on my past, or at least not all of it.
Postscript: Apparently an old report called ‘Beyond the personalisation myth’ published in 2003 (damn they used the same title!) seems to make a similar case based on more extensive research: : Personalising Web sites ‘wastes money’.