Information architecture: Strategies for analysis of card-sorting data for organizing information on the Census Bureau Web site

Previous research showed that one of the major problems for users of the information-rich Census Bureau Web site was in locating or navigating to desired content. Card sorting was chosen as a way to improve the site. Two important aspects of the card-sorting study are reviewed in this paper: preparation of card labels and interpretation of the results. For the preparation of the labels, the focus was on getting and using user-derived terminology during the sorts. For the analysis, the focus was on using additional ways to visualize and analyze the data. Data were analyzed by cluster analysis in the form of a distance matrix in a tree hierarchy as well as a multidimensional scaling view. In addition, spreadsheet analysis was used to identify terms that were overlapping. Results highlight the importance of using the different views/tools to get a better understanding of the data. The results identify clusters of terms that should be grouped together, terms that need to go in more than one location and outlier terminology, which were seen by users as fitting into many different categories. These results will inform the re-organization of site content.