The goal of this paper is to provide guidance for applied education researchers in using multi-level data to study the effects of interventions implemented at the school level. Two primary approaches are currently employed in observational studies of the effect of school-level interventions. One approach employs intact school matching: matching schools that are implementing the treatment to schools not implementing the treatment that are similar in observable characteristics. An alternative approach disregards the clustered data structure and matches students in the treatment schools to students in any non-treated school. This approach essentially creates a synthetic comparison school for each treatment school composed of non-treated students that look like students in the treatment school regardless of what school these students attend. A primary purpose of this paper is to examine which approach performs better in practice. A related goal is to explore which of several variations of each strategy works best. With intact school matching, one can address any residual imbalance through regression adjustment or individual student matches that occur within a given school match. When matching at the student level to create synthetic groups, one can either match on student-level characteristics alone or also include school-level characteristics attributed to the individual. A secondary goal of this paper is to examine the relative performance of each of these approaches within the broader strategy of matching intact schools or individual students. Tables and figures are appended.