Making media smarter
Neil Perkin, founder of Only Dead Fish, says the best content services of the future will be those that offer smart combinations of algorithmically generated recommendation, with a good measure of content curation, alongside a healthy dose of serendipity...
Last week Google announced that its Prediction API, launched last year, had not only been made simpler and more powerful but become generally available. Meaning that it's now easier than ever for anyone to create apps with predictive capabilities supported by best-in-class technology.
Through using the API apps can, in Google parlance, get smarter over time, learn to continually adapt to changing conditions, and "recommend the useful, extract the essential, and automate the repetitive". At a macro level, it can add a pattern-matching capability to existing datasets, meaning it can be used to predict probable outcomes for current events. At a micro level, it can incorporate machine-learning to power recommendation services.
Some of the potential applications might be to add a layer of automation that would for example allow you to better filter RSS feeds, comments or feedback, filter out spam, analyse sentiment, in order to help prioritise the most relevant or pressing customer comments and queries and route them to the most appropriate person.
But it might equally be applied to learning individual user's consumption, reading and viewing habits in order to identify your most important and engaged customers, customise web pages to populate them with content specific to a user's anticipated interests, and power content or product recommendation based on previous behaviour.
Interestingly, one of the first businesses to make use of the API is a car manufacturer. Ford has partnered with Google in the US to allow them to develop cars that effectively learn from data collected on how and where we drive over time, so that it might learn from our behavior and adapt to it perhaps by suggesting an ideal route or by enabling the car to optimise itself for the upcoming journey in order to improve fuel efficiency.
But the opportunity for media is arguably even more powerful. The practice of accumulating consumption data in order to build user profiles and power content recommendation is not new of-course. LastFM has been Scrobbling (the term it invented to describe this process) for years, allowing users to display their accumulated taste profiles and share them through widgets.
The difference is that now, with as little as a single line of code and some technical knowledge of APIs, this kind of predictive and analytical capability is available to anyone. The API allows you to stream data and tune your predictive models meaning that recommendation services can improve over time. Google is even going to be providing a gallery of pre-built, user-developed models to make it even easier for people to see useful applications and best practice.
So there can be little doubt that predictive capabilities will play a not insignificant role in the future of media delivery. Already we are seeing the development of some interesting content driven applications that capitalise on the capability to aggregate and intelligently interpret data streams. Mombo.com, a movie recommendation service built on top of twitter, uses social monitoring technology to provide a rating for films based on analysis of thousands of tweets about them. If you synch it with your Twitter account it can provide film recommendations based on those that your friends have tweeted that they like.
Smart apps like Peel are capitalising on the trend toward multi-screening. A recent Digital Clarity study of 1,300 British mobile internet users under the age of 25 found that 72% used Twitter, Facebook or other mobile applications to actively comment on shows as they were watching them. Peel turns your phone into a TV remote control, makes it seamless to connect with friends whilst watching, and provides a guide that learns your preferences over time and is able to make content recommendations based on that data.
Similarly, a new generation of iPad magazine apps like Zite ('the personalised magazine app that gets smarter as you use it') are able to scrobble your content consumption, learning not only from what you click on, how long those articles are and how long you spend reading them, but from what you don't click on.
The point is that options that allow users to control content in smarter ways make for a better experience. And better content experiences win. The genius of Flipboard, the iPad magazine app that combines professional with social curation (content that is recommended because your friends, rather than a professional editor, think it is good) is that it gives the user powerful options to control the content they want in a seamless execution.
The potential for content producers of all kinds - for brands, but particularly for media owners - is huge. My only hope is that in the inevitable rush to create better and differentiated content delivery experiences, that we don't forget about the power and the satisfaction that comes from self-discovery of amazing content. The best content services of the future will be those that offer smart combinations of algorithmically generated recommendation, with a good measure of content curation, alongside a healthy dose of serendipity.