Often, development organizations confront a tradeoff between program priorities and operational constraints. These constraints may be financial, capacity, or logistical; regardless, the tradeoff often requires sacrificing portions of a program. This work is concerned with figuring out how, when constrained, an organization or program manager can utilize social networks to take advantage of inherent tendencies that will allow a program to thrive. Specifically, this study looked at the playmate networks of children in 31 rural villages of central Afghanistan and how that relational information could improve programming of a rural schooling program. To accomplish this, a two-stage approach was used, where network structure and composition was estimated using exponential random graph models (ERGMs) and then related to individual child outcomes in math and language performance using multi-level models (MLMs). Unique in this work was translating ERGM parameters to MLM covariates by using the t-statistics from network estimations. Results of the MLMs indicated that individual ability drove most of a child's achievement, however, both network structure and composition were important in explaining children's academic achievement. Specifically, children maintained many reciprocated ties with other children, though more advanced network structures--such as triadic closure--were not fully developed in the networks. Compositionally, children tended to befriend others of the same gender and similar academic performance (homophily measures). This translated into MLM results of children doing better academically if they were friends with other children of a similar ability. Ultimately, the primary concern was how network information could improve program management, performance, and ultimately, impact. Key recommendations for utilizing networks included building in playtime during the school day to facilitate tie formation, identifying isolates and working to integrate them into the existing network, creating a "buddy" system for learning within schools that could provide the catalyst for more complex network structure, like triadic closure, and using visual depictions of networks to identify targeting opportunities for communication within networks. The following are appended: (1) Network Visualizations; (2) Survey Instruments; and (3) Exponential Random Graph Model MCMC Diagnostics.