I conducted an initial test run of a card sort today for my Twitter client Bitter. A card sort is a usability technique in which you hold sessions with potential users of your application and you, well, have them sort cards. Each card will contain a label used in your application. In an open card sort, the users will create groupings of cards themselves and name each grouping. In a closed card sort, the users will group the cards into the group names you provide them.

If your application is a web site, a card sort could help you organize the pages and links that belong under each section for your site’s navigation. If your application is a traditional desktop app, a card sort could help you organize the menu items under a top level menu. For my card sort, I used the labels from my options dialog (and even some potential future labels I plan to add) and had the user sort them.

I expected to end the card sort session with several groupings and have a better idea of how others may view the labels. I did indeed end up with that, but discovered several other things in the process that I didn’t expect. There were a couple assumptions I had made about how the options were to be used with the currently logged in user and the card sort showed me that those assumptions may not necessarily be made by other users.

Another thing I found was that you need to carefully think about and decide what to put on your cards. You have the choice of simply putting the label on the card, or you can put the label and a description of the label on each card. If you put just the label, then you’ll likely have the same problem as me of context-sensitive labels. For example, I have some labels named “New Tweets”, “New Replies” and “Sent Direct Message” among others. By themselves, you aren’t quite sure what they refer to, but once you see them grouped under the “Sounds” section it becomes more obvious. So should the cards stay pure and say precisely what the label will say in the application or should you muck with your independent variables and change the cards to include context like “New Tweets Sound“? That’s something you need to weigh the pros and cons of and determine for yourself and also probably consider when analyzing the results of the card sort.

Also, let your participant talk during the session. Let them ask questions. All this is valuable information. You can then ask yourself why the participant said this or that or why they asked what a certain card meant. You may think the answer is obvious, but maybe it’s not. But if they do ask you what a specific card means, ask them what they think it means. That’s yet more valuable information.

It’s also a good idea to video record the whole thing. It’s impossible to take notes and get everything during the session. Also sometimes you’ll see things on the recording that you didn’t see during the session. Just make sure you get permission first.

Overall, even though this was just an initial test run of a card sort, I can already see the benefits. I did this run through to see what works and what doesn’t work and now have a pretty good idea. If you are writing an application whether in your own free time or professionally, a card sort is a good idea. It doesn’t even have to be formal. Asking a coworker to take a few minutes and come over to your office for a quick card sort would work fine. Often times we just get so close to our applications, that we lost perspective. We get tunnel vision on how it’ll be used and how things will be used and interpreted. A card sort is simply a tool to get other perspectives and to rid yourself of tunnel vision.