DreamBox Learning — Harvard CEPR Finds Modest Gains for Students Who Reach Recommended Usage, but Real-World Usage Fell Short

DreamBox is an adaptive elementary maths platform that adjusts lessons to student responses. Harvard's Center for Education Policy Research studied its use in two sites, Howard County (Maryland) public schools and the Rocketship charter network, using NWEA MAP scores.
The study found that the average student improved nearly 4 percentile points after only 14 hours of use, with larger gains for more intensive users. But the design was quasi-experimental and the authors stated it was not built to establish causality.
Usage was the practical weak point: the programme recommends about an hour a week, while students averaged roughly 35 to 44 minutes per week across the sites. School- and teacher-level decisions, rather than student preference, had the greatest influence on time spent.
The case is useful as a moderate-evidence example: promising associations, no randomised confirmation, and implementation fidelity as the binding constraint.

Read the full analysis: https://cepr.harvard.edu/news/harvard-researchers-probe-student-time-spent-using-dreambox-math-software