I have always thought that open card sorting is a great thing. With this technique, you give end users a set of content written on index cards, and ask them to group it in ways that make sense for them in context of the application.
I’ve had great successes in the past, and I use the technique in training courses.
However, I have been running a card sort over the last few days, and am not as confident that the outcomes will be useful.
I realised today what has happened. In the past, and on training courses, I have done the exercise on either smallish groups of content (where up to 80 cards represent all content), or on large groups of content that already have some type of pattern (for example, large content but of a limited range of content types).
In my current exercise, I’m trying to sort a large amount of content, and it is incredibly varied. Because it is so varied (and I’m not a user of the information), I didn’t want to pre-group much of it. My participants have a very daunting task of trying to group individual pieces of very varied content. No wonder they are telling me it is hard (luckily they understand much of the content fairly well).
I suspect that open card sorting is most useful when:
1. The content set is reasonably small; and/or
2. The content already has patterns that allow some pre-grouping
3. The participants understand the content well.
I will get some useful information out of the process, but nowhere near enough to create a reliable navigation structure…