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

Examine the Dendrogram

At this point, assuming you conducted an unmoderated/remote study and used a tool that performed a cluster analysis, you should view the dendrogram (also known as a “dendrite diagram” or “tree diagram”). The dendrogram is a visual representation of item relationships (Figure 2, Sample Dendrogram).

Figure 2

Figure 2. Sample dendrogram

The structure of the dendrogram is similar to that of a tree in that a large branch (represented by the square brackets to the right of item groupings) subdivides into smaller branches, each of which subdivides into still smaller branches, and so on, resulting in a hierarchy of categories and items. In a card sort, the ultimate goal is to derive a set of divisions and subdivisions of items that will translate into a reasonable IA.

Reading a dendrogram requires a basic understanding of how it is depicted.

The dendrogram begins at the left with the complete list of items that were presented to participants for sorting. Then, branching lines (brackets) are drawn between items, based on the degree to which participants grouped the items together. Items that are the most frequently grouped together are placed next to each other vertically, with a line joining them to indicate the association. For example, looking at the top of Figure 2, Ordering and Delivery Info, Secure Orders, and Shopping Cart form a group. That line represents the next level (or branch) in the diagram. It not only joins the items, but also extends to the right, pointing to the next highest level of relationship. If several items have the same correlation, there will be more than one item in a given branch level, as illustrated by each of the branch breakouts shown in Figure 2. At each successive level of the tree, the conjoined lines are represented by fewer and fewer numbers of aggregated units, until at the far right of the diagram only one line remains, representing the complete aggregation of all original items. How quickly the diagram reaches this final state depends on how cohesive the unit clusters are—more differentiation generates smaller clusters and, therefore, more branches. Less differentiation generates larger clusters and, therefore, fewer branches.

The use of online tools reveals the true power of the dendrogram to help visualize the structure of content within a specified number of categories to be represented in the IA. For example, to see where content would reside if all content were grouped into seven categories, you can manipulate the dendrogram to show where content would reside in a seven-category scheme. Online tools typically provide mechanisms, such as a slider and alternate shadings, for specifying the number of categories to depict (for example, Figure 3 shows the same dendrogram as in Figure 2, this time, manipulated to depict eight groupings).

Figure 3

Figure 3. Dendrogram depicting eight groupings of items in alternate green and white shadings


Previous | Next