Different Approaches to the Search UI

Different designers solve the same common problems in different ways. I was interested in looking at how a popular website handles its search interface. I decided (for no very good reason) to look at The Atlantic magazine.

The first surprise that I encountered was the difficulty I had in locating the search interface. I scanned the home page banner pretty intently. No search box. Finally, my eyes rested up on the magazine’s wordmark. Just below it are three menu entries for Subscribe, Search, and Menu. None of these menu options have any visual cues that suggest they’re clickable elements, which makes it hard for a first-time user to determine that they’re affordances. Clicking the Menu option causes a search interface to appear. "Hiding" the search UI behind a drop-down or mouse-over is becoming increasingly common and represents a step backwards in my opinion. Designers who get too clever by eschewing convention risk frustrating their users.

The Atlantic's unusual interface for exposing the search box.

I tested stemming features by typing an incomplete string, “Obam.” The results page informed me that it was presenting me with results for the string “Obama” and gave me the option to search for what I had actually typed instead. If I select the option to search for the exact incomplete string that I entered, it provided me with new results while also helpfully asking did you mean Obama?

When I tried to search for the nonsense word ziggleswam, the results page indicated that no results were found. The Atlantic’s results page failed one of Rosenfeld’s basic commandments for search systems: never leave the user with a dead end. While the system gave me no helpful alternative suggestions, it did at least provide me with another search UI, ready and waiting for another try. It also helpfully kept the search UI populated with the term I had typed so that I could see for myself whether I had typed it correctly or not.

Advantages and Suggestions

A benefit of The Atlantic's search system is that it's based on Google, the search engine and interface that Americans are most familiar with. The system responds to queries, misspellings, and incomplete strings in exactly same way that google.com does. For a magazine site whose archives consist almost entirely of past published articles, this search system works well enough. For this site, we're more interested in ranking than sorting. Sorting is the best approach for comparing many items (as in a catalog website), while the ranking approach used by The Atlantic best serves a user who's looking for relevant hits based on a word or phrase. 

Since The Atlantic's backlog of archived articles is deep, a given search term could return many years' worth of results. The implicit assumption of this search UI is that the user always wants recent articles. That's not necessarily a valid assumption. It's just as likely that the user is looking for an article on a given topic from a particular year. The search UI could be improved by allowing the user to refine their search by specifying a date range for results.