Significance Understanding and thinking critically about scientific evidence is a crucial skill in the modern world. We present a simple learning framework that employs cycles of decisions about making and acting on quantitative comparisons between datasets or data and models. With opportunities to improve the data or models, this structure is appropriate for use in any data-driven science-learning setting. This structure led to significant and sustained improvement in students’ critical thinking behaviors, compared with a control group, with effects far beyond that of statistical significance.