Card sorting is a user-centered design method for defining the information architecture of a Web site. It is a method used by IA professionals and Web designers everywhere. The basics involve bringing in a small group of target users of the site and giving them index cards with the names of content pages written on them. The participants are then asked to divide the cards into appropriate groups of related content that make sense to them. In an open card sort there are no predefined headings for the groups and the participants are asked to name the groups. In a closed sort, the participants are given group names and asked to place the content under those predefined headings. That is the basic structure of a card sort. If you want more information this site has a very extensive description. After you have collected the data from all the users then comes the process of compiling and analyzing the data to arrive at a IA that agrees with what most users would expect, based on the results of the card sorts.

There are a number of advantages to card sorts:

• They are simple and easy to do
• Cheap
• Provide a lot of data

There are also some disadvantages:

• Results can vary widely
• It can be difficult to make the data useful in complex sites
• They don’t take user tasks into account
• They put users in an unnatural scenario

While card sorts are an industry standard and very well accepted I question their usefulness for the same reason that all focus group testing should be questioned. It puts users in an unnatural situation. You aren’t learning what a real user would do, or if a real user would find what they are looking for, you are learning how a person who is given a pile of cards would organize them. These are two very different things and certainly calls into question the usefulness of the data.

I remember a talk I once attended where a former Sapient team member that had worked on the United Airlines Web site back during the dot-com boom had spent a ton of time and a ton of United’s money doing lots of focus group tests to assure that users could find and book flights as easily as possible. The result was a design that lasted only a few months before they had to redesign it based on real user data.

While I believe in some circumstances cards sorts and focus groups can be useful, they are not a silver bullet to perfect usability. There is much more to be gained from careful analysis of real world user data and making adjustments in IA based on this data. While card sorts can provide a starting point, it has to be seen as just that, not the final.