Cohort size, mentor pairing, and weekly cadence each independently predict completion. Together they explain 78% of variance in completion rate across our dataset.
The biggest single predictor is mentor pairing — programs without paired mentors hit a ceiling around 65% completion regardless of how good the content is.