In science and engineering education, the use of heuristics has been introduced as a way of understanding the world, and as a way to approach problem-solving and design. However, important consequences for the use of heuristics are that they do not always guarantee a correct solution. Learning by Design has been identified as a pedagogical strategy that can guide individuals to properly connect science learning via design challenges. Specifically, we focus on the effect of simulation-enabled Learning by Design learning experiences on student-generated heuristics that can lead to solutions to problems. A total of 318 middle school students were exposed to a lesson that integrated design practices in the context of energy consumption and energy conservation considerations when designing buildings using an educational CAD tool. The students were pre- and posttested before and after the 2-week long intervention. The data analysis procedures combined qualitative with quantitative methods along with machine learning approaches. Our analysis revealed two distinct groups of students based on their learning achievement: the naive developing heuristic group and semi-knowledgeable fixated heuristic group. Differences between the groups are discussed in terms of performance, as well as implications for the use of computer simulations to improve student learning.