Recommendations and Revelations

Search engines are indisputably a potent way to generate value, but it may be recommendation engines, which encourage users to keep coming back for similar content, that actually pay the biggest dividends for content companies and publishers in the long term.

In its basic form, recommendation technology—modeled on the approach of online bookseller Amazon—suggests content on the basis of what like-minded customers consume, connecting users with relevant content that their peers recommend. But the paradigm goes far beyond that to link consumers not only with content they like, but ultimately with users who share their interests and passions. There are also hybrid models emerging in the spaces in between—combining filtering, artificial intelligence, and human-computer interaction—that take this all to a new level.

I was privileged to see some of this exciting work-in-progress at RecSys 2007 at the University of Minnesota, an inaugural event that attracted a select crowd of 120-plus participants from 16 countries across five continents. In addition to introducing the industry’s best and brightest researchers and students, it also showcased some radical new thinking from vendors including AOL, eBay, Google, and Unilever.

A particularly interesting twist on recommenders comes from MyStrands, a U.S.-based provider of content discovery and recommendation technology that has successfully harnessed the dynamics of social networks to develop what it calls a “social recommendation engine.” Atakan Cetinsoy, MyStrands’ VP of business products, told me the company’s patented technology relies on the wisdom of crowds to recommend the right music content to the right users.

Essentially, it analyzes how people listen to and organize their music and learns from these patterns to suggest the right content to the right users in real time. “The user listens to the music and we listen to the user,” he said. When enough users listen to a song A followed by a song B, then the system concludes with certainty that the two songs are similar. Using this insight, the system can further suggest tracks. The system also promotes community by allowing users to see the listening patterns of friends as well as peers the recommender system has identified are a perfect match based on their music tastes and preferences.

Another company that rocked at RecSys was Aggregate Knowledge. This must-watch newcomer drives highly targeted and relevant content and product placements to consumers based on real-time buzz and collective buying behavior, which it monitors across its network of 100-plus media and retail websites. According to Aggregate Knowledge CMO Dave Peterson, “People are the ultimate tiebreaker when it comes to deciding what is relevant. They also demand an easier way to find what they like on the internet.”

With this in mind, Aggregate Knowledge recently took the wraps off the Pique Discovery Network, an offering that is designed to (you guessed it) pique the consumer’s interest in targeted content they will most likely find useful and would not have otherwise known existed. As Peterson says, “It’s all about placing relevant content based on real-time trends versus having to rely on demographic or profile based information.”

To date, more than 60 million people visit Pique Discovery Network sites every month to discover products and information; the network delivers approximately 1 billion discoveries per month. Peterson attributes the popularity to the “hidden gems” his company’s offering can uncover. Aggregate Knowledge’s offering monitors what’s going on across a variety of websites, so it has the inside track on emerging trends and connections between seemingly disparate pieces of content.

It’s also a clever way to avoid the downside of content recommendations, which have a tendency to reinforce the blockbuster nature of the media. It’s a dilemma that Kartik Hosanagar, the Wharton professor of operations and information management who has researched the problem, argues can actually “create rich-get-richer effects for popular products and vice-versa for unpopular ones, which results in less diversity.”

In addition to bubbling up new and relevant content, Aggregate Knowledge’s discovery network allows companies to serve direct product and content placements via websites, email, and affiliate promotions. For example, a consumer opening up an email newsletter from a publication can instantly see related articles and content based on what other consumers are viewing on that site at that moment in time. Fast-paced content changes faster than users can keep up, but companies like Aggregate Knowledge are watching the way users interact with content to enable enlightened discovery.