Understanding problem-solving strategies and how different tools support problem solving is an important but difficult problem in cognitive science. Cognitive modeling provides one way of understanding and predicting problem solving and the impact of supporting software tools. Modeling typically requires tradeoffs between fidelity of result and difficulty of model building. We used CogTool to explore how well a limited modeling approach can predict performance differences between two applications that support problem solving, specifically, for planning attitude of the International Space Station. We develop a modeling policy for modeling complex behavior using a coarse-level tool with reduced expressive power; then we compare model predictions with experimental data to assess its ability to identify performance differences across systems, tasks, and strategies.