We've been doing quite a lot of usability testing work around search recently, which has been a bit of an eye opener, especially when it comes to the dismal performance people have come to accept from search on academic websites. There was as a toe-curling moment in one test where I found myself watching a user manually removing stopwords like 'the' and 'of' which the engine would have actually removed automatically for them. Here's a few other things we've noticed:

  • Users routinely assume that searches for items like authors or chapter titles have to be done from the advanced search page or a specific field, whereas the quick search from the top of the page on our sites actually offers the user the ability to search for concepts, titles, authors, DOIs, ISBNs and so on from one simple field. According to this study, 47% of researchers resort to using the advanced search 'when necessary' which I rather suspect means that they have very little choice in the matter on a lot of sites.

  • Whereas not exposing the backend implementation model to the user is a basic principle of user centered design, a lot of sites routinely force the user to have a knowledge of how the backend search database is structured i.e. between the fulltext, the abstract, title and keywords and just titles, while there is commonly no option for the user to simply search all of these as the default and for the engine to pick the most relevant results. From what we've seen from both testing and search analytics it seems fair to suggest that an awful lot of users have little conception of (or interest in) the difference between these fields.

  • A lot of sites seem to provide a lot more advanced search functionality than people either want, need or can actually use. Whereas Jakob Nielsen pointed out that Boolean search is fundamentally unintuitive back in 1997, many sites continue to incorporate Boolean into their search in such a way that only an advanced user will be able to understand it.

  • Whereas many sites offer a lot of functionality from the search results page that is extrinsic to the search experience they often ignore useful functionality like facets, scoped search or auto-suggest that would be key to the search experience anywhere else.

I should certainly caveat the above by saying that while poor search implementations appear highly common in academic publishing there are certainly some very good examples from suppliers who clearly take search as seriously as we do - and I should also say that I'm very sure we haven't got everything right yet. But it does seem that the best search examples are currently in the minority, which suggests that people should be looking to usability test their search implementation and to scrutinise their analytics data to better understand what people use their search for.