upa - home page JUS - Journal of usability studies
An international peer-reviewed journal

Card Sort Analysis Best Practices

Carol Righi, Janice James, Michael Beasley, Donald L. Day, Jean E. Fox, Jennifer Gieber, Chris Howe, and Laconya Ruby

Journal of Usability Studies, Volume 8, Issue 3, May 2013, pp. 69 - 89

Article Contents

Label the Top-Level Categories

In addition to the number and contents of categories, it is of course critically important to determine the label names of those categories. Category labels can make the difference between users’ immediate recognition of the nature of the content residing there or cause delay and confusion. An excellently grouped set of content cannot overcome a label that does not reflect its content accurately or that is simply not intuitive or immediately recognizable.

To get a first, general idea of the labels of categories generated by participants, make an initial review of the category labels. The goal of this step is not to identify the final set of category labels, but to get a sense of the general types of category labels the participants provided.

To do so, first create an initial draft of category labels that are based on the labels study participants provided. A useful tool for this activity is the item-by-category matrix, which will show how participants grouped items and the category labels they used. See Figure 4 for an example. In this example, cells contain the percentage of participants who placed each item into a category, and the labels they created for each category. For example, participants placed Contact Us in the About Us category 30% of the time.

Figure 4

Figure 4. Item-by-category matrix

Because participants have the opportunity to create labels with no restrictions, the odds of exact matches across participants may be low. So, your role as analyst is to identify and group labels that are similar, if not exact. Examples from the sample card sort are the three labels “About,” “About Aldo’s,” and “About Aldo’s Coffee,” as well as the other “About” variations. In short, you will search for “themes” of labels. To find themes in card sort data, look for labels that

For each of these themes, you will have to eventually choose one label to use. Your choice may be based on criteria such as the more common options you see on websites, specific terms preferred by the organization for whom you are creating the IA or even a preference for using short vs. verbose labels.

At this point, to help you manage your data, you may wish to create a table containing two columns. The first column should include a list of all of the category labels provided by all the participants. Once you create this list of category labels, go through the list and standardize the labels based on a theme as discussed above. In other words, reconcile any differences, whether they are due to plural disagreement (Chair vs. Chairs) or very close synonyms (Couch vs. Sofa). Place these standardized labels in the second column. Many online tools also allow you the ability to standardize the category labels.

For the moment, consider these labels placeholders. Remember, you still aren’t making a final decision about categories or labels. You’re just standardizing the data so you can determine the level of agreement you have at this point of your analysis.


Previous | Next