There are millions of choices on the web. Search engines narrow the choices down to thousands. Does that help? Recommendation engines attempt to narrow the choices down to a handful based on your previous choices, and the choices of people deemed to be like you.
The New Your Times has a story today "Like this? You'll Hate That. Not All Web Recommendations Are Welcome" which details some of the problems with recommendation engines, and mistakes that were made by Wal-Mart and others.
Netflix and Amazon have recommendation services for movies, books, and music. They keep a profile of your purchases and ratings of previously purchased/rented items. They do a reasonable job of recommending new titles, but they tend to offer titles by the same artist, author, director, or something in the same genre. Pretty simple and very narrow. The variety and serendipity is lost. These services don't recommend the truly outstanding, yet unknown stuff.
Circuit City and other retailers use recommendation engines to suggest an extra battery or tripod when you buy a camera, or a DVD player when you buy a new TV. These are helpful suggestions because they suggest the correct battery among hundreds of choices, and remind us of what we might need with this specific purchase.
C/Net reviews, ePinions, and others offer consumer reviews of products to help us decide what we want. Consumers like you and me spend a lot of time researching and reviewing products to help other with choices. I have found these recommendations to be excellent and unbiased.
In the blog world we have Tech Memeorandum, Digg, Reddit, and other recommendation and ranking services to help us find the most popular content. I use these services and find them very helpful. However, for me, the best sources for new interesting content are the blogs of respected opinion leaders like Om Malik, Robert Scoble, Richard McManus, Brad Feld, Fred Wilson, Dare Obsanjo, Michael Parekh, and others. For me these bloggers are like editors or filters for the best information.
Search engines and recommendation engines also utilize opinion leaders and authority figures to rank results. Inside their ranking algorithms they give extra weight to certain sources, and penalize other sources known for spam. I worked at AltaVista for several years and was one of the few people who had access to the ranking algorithm. You might be surprised at the elements of ranking and the weights given to each. It is part rocket science and part intuition. In essence Google gives preference to "authority" figures too through their PageRank system. Pages that have lots of inbound links are thought be be authoritative, and are given more weight than pages with few links.
Recommendation engines are still an area ripe with opportunity. There is more innovation to be done, and big money to be made by entrepreneurs. Finding the right mix of ranking factors, authority ratings, personal feedback, and collaborative filtering will produce a new powerful recommendation engine that will boost sales for any site, and make a ton of money for its inventor.
Seen any good recommendation engines lately?
