This is an excerpt of a training I do when I teach documentation for developers.
Card sorting is a simple and effective way to give users input into what they expect from menu interfaces and pages.
Physical cardsorting
In the simplest implementation, you give a user a stack of index cards that are labeled with the elements of your website and ask them to organize it in the way they would look for information. Variations include letting people write their own menu headers or letting people write and reorganize the elements of a website. The goal is not to have one person determine your menu and website layout, but rather to get an idea of what your users expect, and where they expect it.
That’s relatively straightforward when you have users physically present, but how do you manage it when you are trying to get feedback from users in many locations?
Virtual cardsorting
I’ve repurposed kanban boards to help me. They allow people to easily drag cards around and categorize them, they let you do rankings about where an item should be on a list, and they are less single-purpose than dedicated card-sorting software.
My most successful experiment with this was when I was working on the Gluster documentation. I needed to take a large pile of documentation that had grown over time and break it into navigation categories, but I did not have the depth of technical knowledge it would take to sort it accurately myself. Instead, I turned to the team and the developer community.
First, I created a shared board. I gave the columns general names that would allow for sorting, We also had flag colors that would indicate whether a topic was missing and needed to be created or was present and needed to be removed.
I populated the “unsorted” column with all of the current topics. Then I asked people to drag the cards where they thought they belonged and take a screen capture of their ideal state. This was the trickiest part. I wished there was some kind of merge or consensus feature that would allow me to aggregate the data everyone was giving me, instead of having to work with a bunch of screen captures. However, it turned out that after the first person did the majority of the sorting, people more or less agreed on the sorting and then made only minor modifications. For items where there were a lot of opinions, I could set up a flash meeting where they could explain what they were thinking and hash out the disagreements.
Once I’ve captured the information, it’s easier to make it into menus and refine it down. If items need to be submenus, users will usually tell me about it in comments or when we talk over the cards on the phone. People’s perceptions of menu organization vary based on their job tasks, so you’ll never have complete agreement, but testing with users means that you won’t have as many misunderstandings about what people actually use.
Cardsorting and analytics
Pairing card sorting with analytics lets you understand what it is that people are looking for and hoping to find. Once when I ran analytics for the documentation at a training company, I was surprised to find that the most searched page was the one about how to correctly use title capitalization. When I found that out, I surfaced that page so that it was at the top menu level, even though my “logical” setting had put it far down in a sub-menu.
Conclusion
Cardsorting is a way to gather menu organization information from a group of people. It works best when you consider the input as meaningful but not authoritative. Although it is classically a physical activity, there are some ways to make it work virtually, including kanban board software.
Resources
I often use Trello for card-sorting, but there are several workable open-source options, as outlined in this article: