« go back

Is 'Amazon recommends' really the future of the media experience? I hope not. Metadata deserves a richer future

I was listening happily to our metadata and convergence CIKTN panel at Innovate 10 this week, until my colleague Simon, the panel’s metadata champion, said he actually looked forward to his ‘amazon recommends’ suggestions.

I thought, how can I call myself a champion for ‘experience led innovation’ if I don’t push back on this! 

I admit I personally consider Amazon to be one of the great wonders of our time and I have been a frequent customer for a decade. But, like many, I use it because it gets the basics right: fewest clicks to order and fast reliable delivery. I’ll even admit that I have, once or twice even bought an ‘amazon’ recommendation. (usually a business book when researching a specific topic).

What I have never ever done, is actually look forward with delight to Amazon’s prompt that “if I like this book, (it’s metadata based algorithms compute that) I might like this other book”. 

In fact my only bad Amazon experiences happen when I'm trying to discover new content. The benchmark ‘traditional’ experiences of discovering new content are rich and varied, they include reviews, recommendations from friends & family and discovering new titles from browsing nice graphics on carefully merchandised displays. 

One of the key findings of the CIKTN beacon report on digital content consumption was that customer value will be found in the richness of the new media discovery and consumption experiences. So It seems to me that to ‘get real’ the metadata conversation needs be a bit more inquisitive about the cultural richness within these traditional media worlds. If we can get this ‘old / new’ balance right then we will have a very British take on new media that can have a massive global impact. 

So my challenge to you Simon is this: show me and the world how metadata can be harnessed to create really enjoyable, highly valued content discovery experiences that will awe and inspire lovers of the traditional media, or at the very least offer something a lot richer than ‘Amazon recommends’!


Posted by Gus Desbarats – theme champion experience led innovation



8 people have had something to say so far

Hi Gus,

Very interesting post - do you have examples of the types of richer experience you are talking about?


Mark McNally
Posted on 15/10/10 09:05.
Examples: Any good book store, the Times Literary Supplement, a book club, the NME, John Peel, the Sunday Times Culture supplement, a friend or relative who guesses right, etc.. I guess my point is there is huge potential value in recognising that content discovery is in itself, a form of entertainment with emotional value. Don't get me wrong, I'm not being reactionary, I'm just a techhie who believes in the link between experiential quality and mass adoption. I'm looking for metadata allied to, say computer gaming skills, to create an interactive on screen 'magic bookstore' that pushes all the right emotional buttons, I'm looking for the new better multimedia equivalent of 'book covers' . I believe that profit from Metadata exploitation would get bigger faster if it was inspired by richer discovery narratives, like say the proposition 'we become what we read', or even just a micro mind mapping of which cognitive buttons are pushed by which form of stimulus at each step in the content buying experience. Sort books then move on to TV and Web where we chose our content from 'word lists' for goodness sakes
Posted on 15/10/10 09:32 in reply to Mark McNally.
Yes Gus, I am very interested in how the process of online discovering can become much more interactive and experiential using richer interactive tools and intelligent matching. "Recommended" is a bit flat. I will have to get engaged with the Metadata discussion.
Posted on 15/10/10 10:31 in reply to Gus Desbarats.
I'm not sure I follow the argument here. Sure, there's a very different experience to be had from data-driven and single-author-driven recommendation, and a strong argument that the former has yet to prove a viable replacement for the latter (at least at its best). But what's the metadata angle? Are you suggesting that it should be possible to run sentiment analysis across datasets coming from more authored cultural filters (TLS, John Peel etc)? That sounds like an interesting project (though there would obviously be some significant technical hurdles to overcome).

On a more mundane level (and as a close colleague of Simon's) I share a sense of dissatisfaction with Amazon's recommendations, but I also think a lot of problems would be solved without changing the fundamental framework of collaborative filtering based on purchase decision, supplemented by rating. The missing step for my money (at least on the book front) would be if Amazon could implement a simple "don't show me books by authors some of whose books I already own" option. That way it might be worth my while to try to get fiction recommendations from them. As things stand, all I get is an impenetrable wall of Greene, Bellow, Calvino, Chandler, Krazy Kat and other favourites - and I don't need Amazon to tell me about them.

At the very least, it would help if they could not recommend alternative editions of actual books I already have! Both of these feel like easy wins and I remain baffled why they haven't been implemented.
Posted on 18/10/10 11:59.
On the Amazon recommends theme - it would be really helpful if you could at least identify whether or not you were buying the item for yourself or for someone else. I buy stuff for family members or friends and Amazon then seems to assume that I am suddenly into trains or craft - which I am not!
The Metadata dilemma is an incredibly complex one. How many linkages do you need to get something which more closely matches the complexity of human decision making? Does anyone link psychological theories of choosing new content with the data?
Posted on 25/10/10 09:42 in reply to Matthew Shorter.
Good post.

Amazon Recommends provides an easy touch-point for describing content discovery systems. As an AI developer, I've found it to be helpful - it lets you explain a recommendation system to anyone very quickly - and a hindrance - it makes all recommendation systems sound a bit prosaic.

One of Amazon's strengths is that they making shopping efficient. But there's the potential for more engaging recommendations systems. Just as there are "functional" shops and engaging shops.
Posted on 25/10/10 10:42.

Well, I've started to address the challenge you've set me. Oof! My response has turned into something of a setting-out-of-my stall on the whole issue of stats-based recommendations, so it's, er, long. Given that, I've broken it down into 5 parts (!) which I'll publish over the coming days. Here's the intro:­irect=%2Fweb%2Fmetadata%2Farticles

Posted on 26/10/10 14:51.
Nice reply! I think you get my point exactly. The question is how do we accelerate the creation of these more engaging new ways of connecting to content? Who leads? The publishers etc. in the creative industries or the technologists? or is there a need to get a punch of people passionate about content in a room with some metadata/search pros to brainstorm a richer future?
Posted on 01/11/10 10:55 in reply to Daniel Winterstein.

Most read articles

Detroit in transition - guest blog from Elizabeth Anderson-Kempe

Since the collapse of the U.S. housing bubble, resulting financial crisis, and the onset of...

Is 'Amazon recommends' really the future of the media experience? I hope not. Metadata deserves a richer future

I was listening happily to our metadata and convergence CIKTN panel at Innovate 10 this week,...