Multifaceted search is a passion of mine—I have seen this technique improve experiences in both e-Commerce and other information retrieval interfaces. But, as useful as multifaceted search can be, it still suffers from one of the bugaboos of database search—exactitude. Is the future of multifaceted search a little fuzzy?
Steve Lavine says yes. Steve heads Transparensee, a start-up that claims to have built a fuzzy version of multifaceted search. He demonstrated it for me recently and it looked as interesting as its demonstration at the 2006 Demo conference.
The best thing about multifaceted search is the way it can eliminate “no results” queries. By showing searchers only valid choices to select, they never get a “not found.” Transparensee goes that one better—besides showing the exact matches for the facet values selected, it also shows items that are “close.”
Steve showed an example of a commerce search for digital cameras. If the searcher specifies a set of features and price and narrows the list down to one camera, Steve asked “Wouldn’t you want to know that another camera met all the criteria but was $1 more than your price limit?” Similarly, if you were searching for a restaurant in a particular neighborhood, Steve went on, “Wouldn’t you want to know about another restaurant that met all your criteria but was two blocks away?” That is what fuzzy multifaceted search can do.
Steve also showed off a different kind of multifaceted search user interface that uses slider bars—searchers use the sliders to make certain facets more or less important than others. In effect, the searchers decide the relevance ranking algorithm on the fly. This technique avoids the biggest user interface problem in multifaceted search: searchers don’t know that they need to select facet values in order of importance.
With traditional multifaceted search (if you can call anything this new traditional), searchers select their facet values one at a time and see which ones remain after each search. If the searcher selects the desired price and sees a key feature disappear, backtracking to select that needed feature (and pay a higher price) is the only action available. Backtracking frustrates searchers, which is bad because frustrated people abandon at higher rates. Worse, some searchers won’t even know how to backtrack or know that they must backtrack—abandonment is even higher for them. The solution to this dilemma is for searchers to gradually realize that they need to select facet values in priority order—if that feature is the most important, select that first (followed by selecting from the prices available).
But many searchers don’t think that way. They’d like a more fluid experience. Fuzzy multifaceted search is one way to provide that experience.
I have seen the future of multifaceted search, and it’s kind of fuzzy.
