Author(s): Matthew KutchWith increasing financial pressure on higher education, concern for student retention is growing. Most previous work on retention focused on factors unobservable to a faculty member, including measures of social, academic, and financial support. This study uses weekly data on attendance, low-impact high-frequency assignments, and high-impact low-frequency exam performance to model course performance and retention. The models confirm the value of an early warning system for course performance (e.g., midterm grades) after the first high impact exam. The models suggest potential value in adding an end-of-semester warning system for retention based on performance on low-impact high frequency assignments.