NYPL Labs

The Stat: At the top of the NYPL.org web site, there is a fairly traditional search box that is set to search “everything” (meaning the NYPL web site, the Digital Gallery and both catalogs). Over the last couple of months, we have begun to track all of the search terms that users are entering into this box, since it is the one place where patrons are not prompted to search for one particular type of information.

We are evaluating potential designs for a new search results page, and we were curious about what sort of terms people were entering in the general search box. We manually sorted the most popular 1,500 terms into five categories: Person, Keyword, Title, NYPL-specific, and Domain. For Person, we didn’t distinguish between searches for authors and searches for people, only whether a name was entered. The breakdown turned out as follows:

Keyword searches make up almost half of all searches, with Person searches at 15% and Title searches at 10%. Those numbers didn’t seem too surprising, but the other two categories gave us a lot of food for thought. NYPL-specific searches–terms like “dvd”, “hours”, “lost card”, “late fees”, “classes”, etc.–made up 24% of all searches, a surprisingly high number. And 5% of all searches were patrons entered domain names into the search box, which might seem unusual until you realize that we have many “catalog-only” terminals in our library branches that cannot access the Internet outside the NYPL sites. Despite all the signs posted, lots of people still try typing in domain names into any box they can find, desperately hoping to reach “google.com” or “yahoo.com”or “facebook.com”.

The Lesson: That 24% of NYPL-related search terms is a formidable number. Clearly, if we treat all search queries as catalog searches by default, we’re doing a disservice to more than 1 in 4 patrons. Fortunately, it seems like many of the most common internal library queries occur over and over again (the single most popular search term is “tumblebooks”, our interactive children’s storybooks, and we see many variant spellings of it). These general queries tend to occur away from the “long tail“, in the leftmost high-frequency part of the graph. We can probably count on “renew books” and “late fees” being evergreen search queries that warrant special treatment and a bit of extra coding.

We’re testing a number of designs for the new NYPL search interface (launching in 2009) that will produce special responses for the most common non-catalog-related searches. These responses will be highlighted in a way to make them obvious and clearly demarcated from catalog searches. With 91 locations and a century of history inside our walls, it would be dangerous to assume that people are always simply looking for a book.