Two evaluations of a prototype designed to help expert users visualize key census statistics are conducted. The results yielded are compared in terms of usability issues, task completion (interaction) and ideation facilitated. Ways in which this information may be affected by the use of different data collection techniques, participants and tasks are considered. We report differences in the results of the evaluations in each of the three areas and suggest that flexible and non-disruptive methods be used to investigate whether geovisualization tools can support knowledge construction. We recommend using exploratory tasks, employing 'think aloud' strategies, requiring users to suggest and explain hypotheses and using screen capture to contextualise the data collected.