We share experiences from a mixed methods evaluation in rural India that combines a randomized controlled trial (RCT) of 400 villages with embedded case studies in four villages. Specifically, we present two lessons from the multi-stage sampling approach adopted to select the four case-study villages, which first prioritized key-informant observations regarding intervention status in order to shortlist locations and subsequently used data from the RCT’s baseline survey to select the final sample. In doing so, we highlight how large-scale mixed methods program evaluations in education can go beyond questions of “what works” to answering those of “how,” “why,” and “why not.”