• by Steve Smith , Staff Writer @popeyesm, January 15, 2010

Personalization has always been a very tough nut to crack. Generally, most of us like the idea of a highly customized feed of information, but we would rather not have to engage in all of the form-filling and box-ticking it requires at most content sites. Some providers like Claria and others have tried to throw behavioral technologies at the problem and craft personalized content off past usage. The problem with these approaches is that it takes time for them to get a handle on your patterns and to push back to you recognizable results.

 

On the other hand, it is also possible for automated or manual approaches to overpersonalize an experience so that the user loses any sense of discovery or serendipity. I like my iGoogle home page of subscribed RSS feeds, but in total it is a very limited set of sources and items from only a swath of my life. I could craft something a bit more diverse, but how much of a managing editor do any of us want to become for our content consumption?

The process is evolving, however. When I mouse over a headline on my iCurrent (www.icurrent.com) personalized home page, a pop-up prompts asks me if I am "Interested?" in the story. If I click the up arrow, I get a larger box of choices inviting me to "Tune Your Channel." The iCurrent engine extracts from that story more granular topics, keywords, people, the source of the story and even the type of story source (mainstream media, etc.). By checking off the boxes, I am refining the filter and telling the personalization engine to favor specific kinds of content. This is all being done on-the-fly and only when I choose to tune the personalization functionality.

"We basically come from the point of view that there is only one way to get at personalization, and that is getting the right degree of user participation," says Ramana Rao, co-founder and CEO of the small 7-person company in San Francisco. iCurrent is trying to be that personalized newspaper front page that MyYahoo, iGoogle, DailyMe, and scores of others have tried.

Rather than confront the user with a long form to fill with content choices, iCurrent is hoping to balance iterative interaction from the user with immediate results on the page so that the user becomes a lean-back consumer of content who can easily lean in at any point to tune or add a channel. And unlike the feed-based MyYahoo or iGoogle pages, the user doesn't have to move their boxes around the layout. The engine determines the page order based on the level of new content and your own recent level of interest in the content. "You aren't in the panel and plumbing business," says Rao. The system is giving you something very concrete, a specific story, to respond to and to use as a foundation for tuning.

The two main places where iCurrent feels it is innovating technologically are in "entity extraction" and in classification. The engine is spidering content from 27,000 news sources. Then it finds personal names, companies, products and other keywords that are best able to leverage its unique kind of user-driven personalization. These entities become the objects put in front of the user to prompt further tuning of a channel. The system also classifies content into groups that sound a lot like behavioral segments rather than content taxonomies (i.e., RV users or runners).

Offering users easy ways to tune a channel also allows for evolving interests. Rao theorizes that most of us may start following a topic at a very broad level, but as our own acquaintance with a topic matures, certain types of stories or sources may become redundant. Having an interface that allows users to tune topics up or down in granularity or even discover that they really are only interested in select sub-segments of a topic, allows for the system to breathe more than an RSS feed page. One of the areas of concentration for Rao and his small team is figuring out which entities and tuning options to put in front of users around each article so they can dial their interests up or down most effectively.

Business model issues aside (such as how iCurrent attracts an audience among all the other aggregators), the pursuit of personalization needs to stay alive. For reasons that are beyond me, we seem to have accepted a certain amount of content and advertising clutter on our desktops. We tolerate Web sites that have more ads than a NASCAR racer and we ourselves keep more windows and Web pages open at a time than seems reasonable.

But as content moves off  the desktop and onto mobile devices and even apps for set-top boxes and the like, the need for more sophisticated personalization will become critical, I think. Algorithms alone really aren't enough, and we certainly can do better than the rough cut of topic selection that most mobile news apps give us. iCurrent is an interesting step towards finding that balance because it recognizes that the ultimate answer involves a kind of ongoing conversation between publisher and reader. The unexpected payoff for both content provider and marketer is what interactivity really should have been about all along: a deeper understanding of your audience's real needs and preferences.