When a solution is available but no one uses it, a certain kind of frustration develops. If you were to walk down any school hallway right now, you would probably see students hunched over worksheets, the class moving at the same pace, and the same problems, regardless of who has already figured it out and who is quietly lost. Ironically, schools have had software for years that could solve this exact issue, and the most recent Stanford data makes it more difficult than ever to ignore it.
Stanford researchers and Khan Academy examined data from over 200,000 students and discovered that even a small amount of time spent annually using mastery-based math software was linked to quantifiable learning improvements. several hours. Not for a semester. Not a comprehensive curriculum revision. several hours. Until you sit with that detail long enough to understand what it actually suggests—that the barrier to improvement for struggling students may be far lower than anyone assumed—it seems nearly unbelievable.
Traditional instruction is not the same as mastery-based programs. Instead of moving forward just because the calendar indicates it’s time, students advance only after proving they have understood a concept. While John, seated two desks away, is still working on second-grade material, Jane may be working through third-grade material. On paper, it seems doable. Teachers claim it’s a completely different matter in a classroom with thirty students. The main reason these programs haven’t taken over classrooms as the research may indicate is due to logistical challenges rather than a lack of evidence.
Stanford associate professor of computer science Emma Brunskill has been forthright about this. Students in the fourth and eighth grades are doing noticeably worse on math tests, “particularly since the pandemic,” she said. The issue is not one of abstract policy. It has an impact on children’s opportunities, careers they perceive as viable, and the financial foundation they begin adulthood on. Brunskill’s lab appears to sense the weight of that.
What early data can tell us was further revealed by a different but related strand of the Stanford research. Researchers discovered that just two to five hours of student activity within an edtech tool could predict whether a student would place in the top or bottom performance tier on an assessment months later, using datasets from various educational platforms, such as a literacy game used by children in Uganda and two math tutoring systems used by middle school students in the United States. Not an exact score. but a significant signal. A teacher or the software itself could respond to this kind of early warning before a child falls too far behind.
Perhaps the most undervalued discovery here is not the prediction accuracy per se, but rather what it suggests about customization. A system could theoretically adjust in real time, providing more support, slowing down, or encouraging a teacher to check in before the gap widens, if it can identify within a few hours which students are likely to struggle at year’s end.
Whether the educational system is prepared for that level of responsiveness is still up for debate. Schools are massive, sluggish establishments. Teachers are overworked. It takes training, trust, and a willingness to reorganize what a classroom actually looks like on a Tuesday morning to adopt new software.
Nevertheless, the data continues to mount. The evidence consistently points in the same direction. It turns out that a few hours of appropriate online math practice can reveal a lot about a student and potentially alter their future.
