I’ve been trying to find the time to do a bit more work on the library search terms I was looking at earlier https://libwebrarian.wordpress.com/2011/04/01/library-search/ So here a bit later than I hoped are a few more thoughts based on what search terms people have been using.  The data comes from search terms used by people searching via the tabbed search box on the library website homepage, so they have the option to search the library catalogue, library website, and our e-resources system, Ebsco Discovery Solution.  I’ve also looked at an equivalent set of search terms from the predecessor of EDS, 360 Search.

Don’t put text in the search box
Whenever we’ve put ‘helpful’ text into the search box then those words become the most searched for term, by a considerable margin.  So ‘Search the Library Catalogue’ is ten times more likely to be submitted as a search term than the next highest term.  ‘Search One-Stop’ is five times more likely.  That implies users have just been clicking on the Search button.  One of the explanations that has been suggested is that it offers them a quick way to get to the underlying search system.

Wordle of search terms used to search library systemsWhich slightly begs that question that you might just as well give users a link to the search system rather than a search box.

At the moment we’re testing stopping users being able to click search until they’ve entered some search terms and not putting any text in the box to gauge user reactions.

Cross-over of terms between different search tabs
I’ve done some more work on looking at the top 100 search terms to compare what users type into the catalogue, website, federated search and discovery search boxes.  When I looked at the top 20 then approximately 40% of the search terms were identical across the search platforms.

Unsurprisingly that changes when you look at the top 100.  Only about 22% of the search terms are the same across all the search boxes, with 45% used in at least three search boxes.  The graph below refers.Graph of search word percentages

There are a couple of figures that seem to stand out.  41% of website search terms have only been used in the website search, and that’s understandable.  But at the opposite extreme only 16% of search terms used in the catalogue search are unique to the catalogue.  That may be a peculiarity of a distance learning institution that there’s less unique print content maybe.

As well as the search terms we’ve also got the number of times each of the terms have been used.  This gives a slightly different picture

Graph of numbers of similar searchesIn terms of quantities of searches carried out then an average of 40% are common to all search boxes.  That matches the 40% of the top 20 terms that are common to all search boxes.

There is quite a big drop off after the top 10 searches in terms of numbers of searches.  The top 10 are in the 100s and there’s quite a long tail.  So because a high proportion of the top 20 results are used across all search tabs then when you add up the total search numbers it pushes the percentage up much higher than when you just count the search words.

Again the catalogue has fewest unique searches, but federated and discovery search all have more unique searches than the website, which is a little curious.  I’d have expected there to be considerable common ground between federated and discovery search and for them to have many similar terms.  That bears some further investigation.

Direct comparisons of the different types of search
I’ve included graphs below of each of the four search types comparing the percentages of search words with search numbers.   Catalogue searchIn each case the graph looks at the list of search terms seen by that search system and then compares them with the other search terms to show whether they are common to 1, 2, 3, or 4 search lists.

Discovery search

Federated search

Website search

Final thoughts
Using the numbers of times the search terms are used is to my mind more representative of what users are searching for.  That gives a pattern that says that more than half of the time users are using the same terms in each search box.  Although you’d expect that cross-over between federated and discovery search, and to maybe a lesser extent, the catalogue, you wouldn’t expect it with the library website as it contains mostly help and support materials or information about services.

That seems to me to imply that:

  1. Users aren’t all that clear what the different tabs are searching.  Website in particular is an ambiguous term.  You couldn’t really argue with a user that insisted that they are all websites anyway.
  2. There’s a lot of common terms that could equally (and legitimately) be used against most (if not all) of the search tabs and users might well expect (and want) useful results
  3. That if users are going to type the same search term into every box then it would save their time (isn’t that one of Ranganathan’s laws?) just to let them type it once.

It’s been an interesting exercise and some useful evidence to compare alongside our search evaluation and focus group work to feed into our website redesign project.