Comparing Two Methods for Predicting Navigation Problems in Information Hierarchies

When locating target items in information hierarchies, performance depends much on selecting the right nodes along the path. Two methods for evaluating category membership, card sorting and a simple statistical approach, are compared with regard to their predictive capacity of navigation difficulties. We developed a category structure for a web site and obtained metrics from both card sorting and a statistical approach based on the word frequencies of category content items. We then conducted a user study to compare actual user performance with predictions from the two approaches. Despite the simplicity of the statistical approach, it produced marginally better predictions. In particular, it predicted problematic target items that took nearly twice as much time on average to find than the remaining targets in the web navigation study. We review these results and discuss usefulness and implications of predicting navigation problems in information hierarchies for the information architect