Perceptive Media

I've been using Zite for a number of months now. It's the personalised magazine app for iphone and iPad that 'gets smarter as you use it'. Zite's personalisation engine is super-smart. Behind the app is a quite mind boggling level of social and algorithmic modelling and curation (created, they say, over 6 years of product development) that blends and refines a unique media experience for every individual user. It works by "mining content from your social web, modelling that content, modelling the community that interacts with it, modelling your interests, matching your interests to the content and your community to help you discover content you'll want to see." Have a read of this 'under the hood' blog post on how the system works. It's quite astounding.

Proof of the system of-course is in the user experience and I have to say that every time I open it these days there is stuff I want to read, which is keeping me coming back to it perhaps more than most other apps I use. Seems like I'm not the only one. Zite's own figures indicate that on a daily basis, 20% of users save or share at least one article and 8% of total opened articles are saved or shared. I'm willing to bet that that is considerably higher than the norm across most magazine-type sites.

The idea of dynamic generation of content using data is nothing new of-course. Each time we load the Amazon home page into our browser it pulls on hundreds of different elements to create a page that is unique based on what it knows about us (location, data on our purchase history, wish list and so on). Yet it feels like many media owners have been quite slow in developing these kinds of dynamic personalisation techniques. But here's an interesting write up of a talk given by Ian Forrester of the BBC's R & D department about some early experiments in what he calls 'Perceptive Media' - the concept of using APIs to transform TV 'on the fly', which takes personalisation to another level still. The thinking is still in it's early stages as Ian is at pains to stress, but it's a rather fascinating idea, and one that is not without its challenges. 

My broader hope though is that we will see far more experimentation around customisation of content using machine learning, modelling and APIs, as part of a wider mix of curation techniques. There seems to be a disappointing lack of it from many parts of the media industry so some would say it's about time.

Image courtesy

Comments

[ian forrester]

Thanks very interesting don't you think? :-)

Its the effect of the internet on tv... but it certainly throws up many many research questions

[Charlotte marketing firm]

Indeed! I was surprise as first and now I use it almost everyday!

[Coleman Foley]

Another interesting effort in perceptive media is Trapit (http://trap.it). Trapit is similar to Zite, only it ignores social signals. Instead of trawling the user's social web, it trawls the same curated web for every user. Trapit has built a large curated web that is, in my experience, completely free of junk content.

Trapit uses its discovery engine to organize the web into topics. Then, it has editors for each topic to cull all sources that don't meet a minimum quality standard (not covered with ads, not spam, not a repost with no original content added).

Finally, it feeds content to an individual user for each topic the user creates. So you could create a content discovery topic and see what it finds (it does very well with this particular query) Then, as you thumbed those recommendations up and down, they would gradually improve.

I discovered this post through Trapit today and I have since read two of your other posts, the original and the redux content curation pillars. I will definitely be following this blog.

[Ian Forrester]

We have a demo now which I'd love you to try out - http://thenextweb.com/media/2012/07/03/the-bbc-unveils-its-first-perceptive-media-experiment-and-you-can-try-it-now/