Comments on: Is there really a market for cross-platform recommendation engines?

By: Viaccess

Viaccess — Sun, 29 Jul 2012 13:56:16 +0000

By: Viaccess

Viaccess — Sun, 29 Jul 2012 13:47:19 +0000

The article states what we know for a long time. Every reco approach has it pros and cons. Relying on a single method, be it collaborative filtering, social recommendations, semantic or anything else is bound to create frustration. Why? Since any method has its own lock-in mechanism. For example Youtube related videos will initially please you as it bring more of what you liked but at some point this inward spiral of more of the same becomes displeasing. This has been shown in research by Dr G. Oestricher and S. Riechmann from TAU university who demonstrated that adding user-links (social) to content links (related content) yields in overall satisfaction increase of more than 30% (see research – http://goo.gl/0wQhu)

In short, engine diversification (or as called by VO: Engine blend) promise to improve user satisfaction and overall usage conversion substantially

By: James Doman

James Doman — Thu, 12 Jul 2012 06:45:44 +0000

I disagree. Experiences aren’t commodities, they’re the most important part of modern b2c business and co’s that think otherwise are pretty much destined to fail into obscurity.

The only way these co’s an live is if they are fixing a need. E.g. Chomp, bought by Apple, fixed the need for discovering apps. Yelp helps me decide where to go out for dinner. Zite finds me stuff to read. But like I said, there’s an overarching experience that fulfils a need over just recommendations..

By: Lasse Clausen

Lasse Clausen — Wed, 11 Jul 2012 22:53:16 +0000

For people who are used to linear TV I’d agree that there’s not much need for recommendations. But a large number of American college graduates in recent years don’t get cable subscriptions anymore, only an internet connection that they source their entertainment through. In that case you really need a mechanism to filter out the noise while maintaining the easy, lean back experience that we want for entertainment.

By: Lasse Clausen

Lasse Clausen — Wed, 11 Jul 2012 22:42:17 +0000

I agree that easy trumps everything in entertainment. And in the former, channel oriented environment people first went to the source and had to take the best that was available, getting something better involved too much effort.
Now we’re in the middle of a shift to an object oriented environment and the rise of long tail. Distribution channels (what you call “the experience”) have become commodities and the object is what matters most.

By: Craig Bender

Craig Bender — Wed, 11 Jul 2012 17:10:47 +0000

Seems like all the recommendation engines are trying to solve a problem that doesn’t exist. With over 5 hours of TV viewing each day, the average person has no problem finding something to watch. And maybe TV is like chocolate. If you find a good show you only need to watch one, if you watch a bad show maybe you need to watch another one to be fully satisfied. So a good recommendation engine might result in less viewing.

By: Daniel Bourdeau

Daniel Bourdeau — Wed, 11 Jul 2012 14:58:52 +0000

I would love to see a cross-platform recommendation system that works well. I believe there is plenty of insight to be gained from the types of books, magazines, video games, music, and movies I consume.

I would really like to see a mixed platform that would let me pick/rate some of my favorite books, movies, tv shows, etc. and then select a list of my friends who’s taste I trust. I can honestly say I don’t want my entire Facebook friend’s list or Twitter feed analyzed to get recommendations. As the author accurately stated above, there are a lot of people who I like to hang out with, but wouldn’t take a recommendation from.

Anyway, thanks for the post. Very interesting!

By: James Doman

James Doman — Wed, 11 Jul 2012 14:37:41 +0000

People don’t want to go out of their way to be suggested something – they want it as part of an experience. Netflix is a great service experience: you go there to watch films, not to be suggested films – there’s a higher need.

Foundd et al are going to find it hard to survive by building “discovery” services without going the whole hog and building the service too.

And companies wanting to implement personalised recommendations don’t have to hire 900+ engineers like Netflix do (or, as @xamat says, don’t). Software as a service companies like PredictiveIntent have APIs that anyone can connect to, configure how the recommendations are decided, and request sugestions back to the individual user.

So, to sum up – discovery and recommendations are part of an overriding experience and won’t survive as disconnected services.

By: Lasse Clausen

Lasse Clausen — Wed, 11 Jul 2012 10:46:40 +0000

Good in depth analysis of the space.
I’d like to add that Netflix doesn’t own the best recommendation engine in the world because they never implemented it. http://www.techdirt.com/blog/innovation/articles/20120409/03412518422/why-netflix-never-implemented-algorithm-that-won-netflix-1-million-challenge.shtml