Using administrative data from an urban school district, two series of predictive models were tested for their ability to project a student's high school graduation status. The models included student grades, attendance, behavior, demographic predictors, and school-level variables. Eighth and ninth-grade variables were tested for two graduation cohorts (n = 2773) and multilevel logistic regression analyses were conducted to analyze the data. Most early warning indicator (EWI) models include end-of-course grades as predictors. This study operationalizes course failure as having failed a single marking period in the eighth grade or having failed a single semester in the ninth grade. Findings offer practitioners an actionable course grades based EWI, allowing for intervention after a single marking period of failure.