First Interactive Prototype

Our users liked the look of the tabs, and that the controls were on the column and row headers, close to the chart grid, but there were some design issues:

• The main problem was that data was lost, or appeared to be lost, when using the tabs to navigate down the hierarchy.
• The tab component really belongs in a dialog (e.g., for selecting preferences), and it was problematic that in our interface the act of switching tabs affected the display.
• Some testers did not understand the symmetrical relationship between the columns and the rows, and the hierarchical relationship between the tabs. We realized that the hierarchy might not have been apparent because the tabs were all at the same level.

So we changed the row and column headers to have levels denoting Survey, Question and Response levels. The legal hierarchy comparisons still held, and since one of the illegal comparisons was survey vs. survey, this did not need to be in both headers, so now it is only in the column header. We also moved the controls back into a control panel, with buttons determining whether the current selection would be applied to the column or row. There was too much duplication having the controls on both the column and the row headers. Instead of using controls as labels, static text labels are now used. In addition to the category, question and response selection, the control panel contained an Export List panel, containing the charts that had been marked for later export or display in the chart grid.

We determined how the reporting and data mining tools would work together, removing features that duplicated reporting tool functionality, such as the display of all charts in a particular category at once, and the display of respondent comments. Also, a good deal of shared functionality (dataset management, filter/query creation) would need to be accessed by reporting tool users as well as datamining users. Because the reporting tool is an HTML interface, it makes sense to have that work done in the reporting tool (but have the results available to both tools), so that the datamining tool users would not have the jarring experience of two different types of dialogs. For example, datasets (named sets of one or more surveys), will still be available for comparison, but will now be created in the existing reporting tool. We changed our terminology from "Dataset" to "Surveys" and from "Answer" to "Response" based on user feedback.

We also designed (but did not have time to implement) a "Group" dialog that would be used to combine desired responses together, or break combined responses apart.