This research from Duolingo explores how to accurately measure language learning achievement within a self-directed platform where users control their study path. To overcome the limitations of standard exercises, the authors developed two specialized assessments: the Checkpoint Quiz and Review Exercises, which integrate testing into the learning process. By applying Educational Data Mining (EDM) techniques to the data collected from these systems, a case study established a causal link between engaging with higher difficulty skill levels ("leveling up") and improved performance on assessments. Ultimately, the study demonstrates how these controlled assessment methods and data analysis can provide actionable insights for improving Duolingo's product development and curriculum design.