Improving faceted search

By Jan Srutek

In a previous life, I redesigned the product finding experience for, which was pretty much just a good implementation of faceted search. I’d love to be able to show you what that looked like, but unfortunately, these days, the Sun product line is mostly an html spreadsheet in a modal overlay on the Oracle acquisitions web site.

I did write about it at the time, with reference to the importance of sorting out the data architecture if you have any hope of manipulating that data in a way that enables a simple, usable interface.

Scroll forward a few years, and I’m directed to another implementation of faceted search on the patientslikeme web site (via Jan Srutek, via Jared Spool), which takes an inordinate amount of data, provides some interesting methods for interrogating that data and, in my case, rather quickly diminishes your will to live when you actually find some’ patients like me’.

Of course, I say ‘patients like me’, but what I’m really looking at when I see a list of 44-year-old, white, not Hispanic or Latin males, who are not yet dead, is just a view of the data that I’ve filtered to be as ‘like me’ as the faceted search allows me to make it, based on the patients who have registered and who are willing to let anybody know everything about them. What interests me more about this are the filters I have available, and the interface I’m using to generate that view and whether that view is useful , meaningful or even interesting.

To be honest, I don’t find that it’s any of those things, but it does highlight that there seems to be very little movement in the design of faceted search. Between my own experiences of designing for this experience, and coming across it every day on retail, travel, ecommerce and any number of other sites, there appear to be limited ways in which you can move the same objects around the interface (sliders, checkboxes, drop-downs, free text inputs) and a similarly limited number of ways in which you can display multiple results and allow the user to retain control, without resorting to chevron blindness or the curse of the ‘remove’.

It is true that you can use appropriate technology to improve the interaction, so that instant feedback to user inputs make the whole ‘fiddling about’ experience more obvious and enjoyable. On patientslikeme, there’s a familiar ‘click and wait’ model, which lets you know something is happening, even if it’s not always obvious what it is. As ‘finding’ things increasingly involves a cross-channel experience, there are even more opportunities/challenges for improving faceted search, maybe with a radically different models, supported by different interfaces and user interactions. But maybe it’s never really improved over the last few years because, really, it doesn’t have to. If it does what it does, and you’re pretty familiar with it, it probably doesn’t need changing for the sake of it, but I’d love to see good examples of where doing it differently does actually mean doing it better. From a user’s point of view, of course.

What do you think?