Using Visualization in Websites

by Dr. Bob Bailey

June, 2003

In some websites the amount of information is so large that the context may be lost whenever it is displayed on a single computer screen. If we try to have the entire structure visible all at once (so we did not lose the context), the details may be too small to read. For example, the HypViewer product on the right below is showing over one million IP addresses. If you think that adding visualization methods to a website automatically will improve user performance in this situation, you may want to consider the results of a couple of recent studies.

Korin Werner (2002) at the University of South Dakota compared the performance of three visualization techniques that were intended to facilitate retrieving information on the Web. The methods were Star Tree, HypViewer and The Brain.

 

 

The more familiar Internet Explorer also was used for comparison. Korin had 27 undergraduate students first become familiar with the methods, and then use each one of them to find six specific target pages on the ‘porsche.com’ website. The target pages were either 2, 3 or 4 clicks away from the homepage.

The researcher reported that Internet Explorer elicited both the fastest retrieval times (p<.05) and the highest percentage correct (p<.05). This was true for retrieving targets requiring 2 or 3 or 4 clicks. The three visualization methods all were reliably slower than Internet Explorer, and differed little among themselves on this task.

Chaomei Chen and Yue Yu (2000) of Brunel University in the United Kingdom noted that beginning in the early 1990s there was an increased interest in information visualization. However, most of the focus has been on developing many alternative visualization techniques, with much less effort made to validate the different methods. They conducted a meta-analysis of available studies that had presented empirical findings on the validity of visualization techniques.

These researchers attempted to more clearly understand the effectiveness of information visualization as it is related to information-retrieval tasks. One of their hypotheses was that users would perform more efficiently and accurately with interfaces that had visualization components. They focused on information-retrieval visualization methods that reported variables related to efficiency (performance time) and accuracy (error rate, precision, etc.).

• Appear in a journal or conference proceedings,
• Include an experimental design,
• Include at least one experimental condition with a visualization component,
• Include at least one dependent variable on efficiency or accuracy,
• Report its results in sufficient detail to be replicated, and Analyze the results using an appropriate statistic.