The effectiveness of educational technology (EdTech) in improving the outcomes of poor, marginalized students has primarily been documented by studies conducted in developing countries; however, relevant research involving randomized studies in developed country contexts is relatively scarce. The objective of the current study is to examine whether an in-school computer assisted learning (CAL) intervention can improve the math performance (the primary outcome) and academic attitudes (secondary outcomes) of rural students in Taiwan, including a marginalized subgroup of rural students called Xinzhumin. We also seek to identify which factors are associated with the effectiveness of the intervention. In order to achieve this, we conducted a randomized control trial involving 1,840 sixth-grade students at 95 schools in four relatively poor counties and municipalities of Taiwan during the spring semester of 2019. According to the ITT analysis, the O-CAL intervention had no significant ITT impacts on the primary outcome of student math performance as well as on most secondary outcomes of the overall treatment group (who on average used the software for only about one quarter of the protocol’s minimum required time of 30 minutes per week, indicating that compliance was low). However, the LATE analysis revealed significant improvements in the math performance of the 30% most active students in the treatment group (who used the software for about two thirds of the minimum required time). Effect sizes of active users overall (0.16 SD-0.22 SD) increased in accordance with increases in usage and were larger for active Xinzhumin users specifically (0.21 SD-0.35 SD). A wide range of student-level and (in particular) teacher-level characteristics were associated with the low compliance to the intervention, which are findings that may help inform educational policymakers and administrators of the potential challenges of introducing school-based interventions that depend heavily on teacher adoption and integration.