Experimental approaches can help disentangle the impacts of policies from the effects of individual characteristics, but the heterogeneity of implementation inherent in studies with complex program designs may mask average treatment impacts (Morris & Hendra, 2009). In the case of the Opportunity NYC-Family Rewards (ONYC-Family Rewards), families earned a range of differing incentives, as a result of variation in engagement in the activities rewarded by the program. Given that the effect of ONYC-Family Rewards on key domains of interest (education, health, and employment) are thought to occur "through" such engagement (as well as the increases in income that accompany it), this heterogeneity is critical to explore. In this study, the authors examine the extent to which effects of ONYC-Family Rewards are moderated by key family and child-level characteristics that are associated with a families' propensity to earn the rewards offered by the program. They use multiple baseline characteristics to predict the amount of dollars earned from the rewards as a means to identify theoretically important, multivariate-defined groups of children for whom program effects might be more concentrated. Specifically, the authors ask: (1) Does the likelihood of earning rewards moderate the impact of ONYC-Family Rewards on children's educational outcomes? (2) Are the impacts of ONYC-Family Rewards on children's educational outcomes greater among the group of children whose families are most likely to earn rewards? Findings suggest that the program may be improving academic outcomes for those youngest children whose families are likely to experience the greatest increase in reward income. Further analyses will explore the mediating processes underlying these effects, to gain further confidence in these results. Appended are: (1) References; and (2) Tables and Figures. (Contains 1 table.)