Education leaders need valid metrics to predict students' long-term success. We use a unique data set with cognitive skills, self-regulation, behavior, course performance, and test scores for eighth-grade students from a Northeast school district. We link these data to students' high school outcomes, college enrollment, persistence, and on-time degree completion. Survey-based cognitive and self-regulation measures predict high school and college outcomes. However, these relationships become small and lose statistical significance when test scores, grade point average, and an absences-suspensions index are included in the predictive models. For leaders hoping to identify the best on-track indicators for college completion, the information collected in student longitudinal data systems better predicts both short- and long-term educational outcomes than the survey-based self-regulation and cognitive measures.