Until very recently, the emphasis in Visualization research has been on methods, their algorithmic underpinnings, and their implementation in systems. Most papers have been of the proof of concept variety: describing new ideas for attacking a visualization problem and demonstrating feasibility and quality by presenting visual and performance results from a prototype implementation. Typical published evaluations might consist of 1) a compelling visual presentation on one or two data sets, 2) a comparison of computational efficiency with known algorithms, and 3) anecdotal visual comparison with other techniques. Such tests have driven creativity and advances within our community, but they do not often lead us to design principles to guide future work nor are they compelling to potential collaborators.