CMU study: Student delay start behavior predicts test scores by 20%

A new study from Carnegie Mellon University, led by Jordan Gutterman and co-authors, validates 'delayed start' as a session-level behavioral detector that generalizes across subjects. Using iReady data from 711 grade 7 students, the team found that delays during Math practice significantly predict standardized test performance in both Math (β = 0.07 SD, p = 0.02) and English Language Arts (β = 0.10 SD, p < 0.001). The researchers went further, applying mixture modeling and sensitivity analyses to identify two distinct student sub-groups: 'early starters' who average less than 5 minutes of delay (20% of students) and 'chronic delayers' averaging over 13 minutes (another 20%).

Compared to students in neither group, early starters experienced positive growth trends (Math: β = 0.11 SD, p = 0.07; ELA: β = 0.15 SD, p = 0.02), while chronic delayers showed negative trends (Math: β = -0.13 SD, p = 0.05; ELA: β = -0.11 SD, p = 0.11). The study, posted on arXiv (2606.25308), positions delayed start as a content-independent detector—unlike traditional usage metrics tied to specific assignments. This work aims to bridge education research with classroom practice by providing teachers with interpretable, session-level behavioral cues that align with observations they already make. The implications are clear: automatically flagging delay patterns could enable earlier, more targeted interventions to boost student outcomes across subjects.