"Seasons" is a computer-based model of the natural processes that cause the seasons on Earth. The project makes sophisticated visualizations of scientific phenomena, usually reserved for scientists, accessible to the museum visitor. "Seasons" is permanently installed in the Experiment Gallery at the Science Museum of Minnesota. Having nearly completed the development process for this exhibit, we have found that such scientific visualizations may benefit from different evaluative tools and practices than other computer-based interactives. This paper will describe our formative evaluation process to date. The ways in which the formative evaluation has failed are the most instructive, and the paper will look closely at why such failures occurred. Some implications for future development projects and the use of appropriate evaluation methods will be discussed.