Prior Work

In deciding how to facilitate our users' data analysis tasks, it was useful to look at how other exploratory data analysis tools present vast amounts of information in easily digestible visualizations. Core to such applications is the ability to find patterns in large datasets as well as quickly explore hypotheses and perhaps find unexpected relationships. In order to do this effectively in our application, we looked at several successful tools we felt did a good job in presenting information without being overwhelming.

• TableLens (Inxight)
• Polaris (Stolte and Hanrahan, Stanford University)
• Spotfire (Spotfire, Inc.)

Summary of findings:

Aspects to use:

• Provide logical defaults for graph types.
• Allow users to be flexible in choosing data to visualize.
• Use default colors for variables or types of variables.
• Provide visual feedback in navigation area so user knows where they are.
• Use small multiples that allow for easy pattern recognition without cluttering a single visualization with too many variables.
• Provide feedback regarding data characteristics (ordinal, nominal, quantitative) to help users understand how to compare data, either through color or grouping.
• Provide access to data behind charts in a separate pane, or somewhere logical in the interface.
• Provide intuitive default settings for novice users, and allow advanced users more control.
• Maintain context when focusing on part of the data.
• Provide thumbnails of visual output when possible.
• Provide selection choices over text input when available (drop-down menus, radio buttons, checkboxes).

Aspects to avoid:

• Complicated drill down menus.
• Hiding functionality. All possible commands should be accessible through menus or buttons.
• Inappropriate chart types should not be available for certain types of data.
• Do not force user to do extra work summarizing data.