A realist evaluation framework asks what works best for whom under what circumstances. This study utilized a realist evaluation framework to determine which students benefited under what circumstances from supplemental instruction, an academic support program using peer-assisted problem-solving techniques. The study setting was a large urban community college that was also a Hispanic-Serving Institution (HSI), with 80% Hispanic students and over 50% of all students qualifying for financial aid. Two sources of existing administrative data were analyzed using a quasi-experimental design. All science, technology, engineering, and math (STEM) courses in one semester with supplemental instruction (SI) (n = 18 courses in spring 2015) were matched to STEM courses that did not have SI (n = 35). Student-level comparisons based on the number of SI sessions attended were also made (N = 1,851). Three research questions were analyzed: (a) what was the pattern of SI participation?; (b) Did course pass rates increase with SI participation?; and (c) Were grade outcomes different based on SI participation? Results indicated that as expected, SI participation was higher among students in SI STEM courses (33.4%) compared to those in non-SI supported STEM courses (5.3%). In addition, increasing numbers of students passed their courses with increasing SI session attendance. Mean grade differences of just under half a grade were found, favoring students who attended one or more SI sessions. Across all students and in math courses, more students who attended SI earned higher grades (A's, B's, C's) and fewer withdrew or received an F compared to students who did not attend SI.