Magnet Relevancy System for AOL
Mountain View, 2006

Following AOL's acquisition of Weblogs, Inc in 2005, the portal became interested in solving it's next big problem. With the immediate flood of traffic and growth of new content properties on it's newly acquired blog platform, the new problem was how to get users to relevant content in a sea of blogs. I was part of small design and development team in Mountain View, California that quickly developed relevancy technology and an experimental front end called Magnet.


Engadget, TMZ and Moviefone are a few of many AOL content properties.


The landscape of content touchpoints

The idea was for the system to pay attention to all of the user's activity around content, ie, all of the bookmarking, searching, accessing, sharing and liking on content. A score or weight was given to each action and then the content's characteristics were applied to the user's interest graph in the form of an 'image cloud.'



Process
AOL, considered a large, slow-moving company by most, is not generally known for our innovation, either in process or in products. This particular project was a major departure from our normal process of ideation through delivery. When a new project is considered, the information usually comes from a Product Manager through a lengthy PRD (Product Requirements Document) in which he or she creates requirements for the new product, generally in a vacuum. That is to say, that User Experience has historically no input at the beginning. The PRD then serves as the team’s “bible” in designing the product.



The idea for this project, however, stemmed from a conversation with an executive who wondered if his novel concept had any validity or potential. This poster describes the process the team went through, the artifacts derived from that process, the environment in which it was created, some interesting emotional fallout and the final outcome.



Review
"Clicking on the Magnet tab, you are presented with a page of images. You click on the images that interest you, and then click "Next." A customized news page then displays. At first I laughed, thinking how typical of AOL to make its users click pictures since they're probably not savvy enough to comprehend RSS feeds. (Do you like gossip? Click the picture of Paris Hilton! Do you like travel? Click the picture of the airplane!). Picture book internet!, I thought. However, it turns out that there's more to this feature than this nifty little parlor trick. Apparently, the page continues to learn about your interests over time using "collaborative filtering and active inputs" (or so says TechCrunch). I'm not exactly sure what that all means, but somehow the page had me pegged - automatically subscribing me to both GigaOM and Cute Overload. Oh, so Magnet is like Tivo for internet homepages! Well, that could be handy."



Patents
1. 20080209351 USER PROFILE SNAPSHOTS - method for personalizing content for a particular user using profile snapshots in a computing system comprising a user 08-28-2008
2. 20080209350 ACTIVE AND PASSIVE PERSONALIZATION TECHNIQUES 08-28-2008
3. 20080209349 PERSONALIZATION TECHNIQUES USING IMAGE CLOUDS 08-28-2008
4. 20080209343 CONTENT RECOMMENDATION USING THIRD PARTY PROFILES 08-28-2008
5. 20080209340 PEER-TO-PEER ACCESS OF PERSONALIZED PROFILES USING CONTENT INTERMEDIARY 08-28-2008
6. 20080209339 PERSONALIZATION TECHNIQUES USING IMAGE CLOUDS 08-28-2008