This study investigates the relative impact of online delivery of supplemental instruction (SI) to undergraduate students enrolled in STEM courses at a large, four-year, Hispanic Serving Institution (HSI) in the southern United States. A multi-condition, randomized trial comparing the relative performance of in-person to online SI was implemented across 20 classroom-based courses over six academic semesters. Application of multilevel ordinal regression models to data obtained from implementation of two experimental conditions revealed that students assigned to in-person SI had higher final course grades than their online SI counterparts, but only when the online SI condition did not include access to the session recordings. These results suggest that synchronous online delivery of SI can be as efficacious as in-person SI when supplemented with asynchronous course resource supports. Implications for the design and future testing of SI models are discussed.