Nicholas Wagner and Prof. Rondinelli lay out their concerns about the proper application of data science methods in their article, Theory-Guided Machine Learning in Materials Science appearing in the open-access journal Frontiers in Materials. Drawing upon both the materials science and statistical literature, they illustrate why the unique nature of materials datasets presents obstacles to simply applying data science methods commonly used elsewhere. They address universal issues in materials informatics such as overfitting, descriptor selection, and constructing interpretable models and provide recommendations for each. Finally, they provide their own general workflow built around the principle of “as simple as possible, but no simpler.”