Jared Spool is one of the few user experience pros who focuses on Web site search. And when he says something, I pay attention. His post yesterday crams into one article more about site search behavior than most people will ever know. But I have one disagreement with him. It’s rare, but it happens.
Jared says that his research shows that people using site search want just one answer rather than a list of great search results. I’m certain that he is interpreting his research correctly, but I wonder if there might be some selection bias at work. Here’s why I say that.
My friend Andrei Broder worked with a few other luminaries years back to identify several different types of searches: navigational, informational, and transactional. This was a great insight, because it was generally thought that all searches were informational—the searcher desired a screen-full of useful search results to choose from. Andrei and colleagues proved that some searches are attempts to do something (transactional) or go somewhere (navigational) and they typically need the right result at the top.
And many Web site searches do require just one correct answer, as Jared points out. But not all of them. In my work at ibm.com, I noticed that the most preliminary searches often were informational ones. Someone might search for “e-mail archiving case studies”—they don’t want to get just one. Now, sure, if you have a page on your site that lists every blessed e-mail archiving case study, that would be a great #1 result, but you usually don’t have that kind of aggregation page for every possible query.
Searchers would not want your “Content Management Case Studies” page as #1, even if that list included every e-mail archiving case study, because it also includes too many other irrelevant case studies. Instead, searchers would love a list of case studies that match the query. They could scan through that list and click several results, drinking in that practical information they crave.
If I am right, then why did Jared’s study show otherwise? It might have to do with sample bias. When search tests are defined, they often focus on popular queries. Those queries are far more likely to have a single page that aggregates the information in one place, simply because so many people are interested in the subject. I suspect that if unpopular, unusual search queries were studied—and those queries are the bulk of the volume—you might see a different result. I can’t prove it however, so I’d love to see more information about what kinds of queries were tested.
Jared is smart to tell folks that not all queries are informational. He might even be on solid ground to say that most are not. I just think that a good portion of them are.
So, do listen to Jared. About almost everything. But just understand that some searches are informational searches, where a rich list of choices is exactly what your customer needs.
