We study the relative effectiveness, cost-effectiveness, and interaction of family- and school-based learning interventions using a randomized controlled trial in Colombia that assigns children to a parental engagement program, a teacher professional development program, both, or a control group. Both interventions are grounded in a child-centered learning approach that emphasizes active engagement and the progression from informal to formal mathematical understanding. Each intervention independently generates sizable and statistically similar gains in early numeracy (0.17σ and 0.20σ). Combining them produces no additional learning gains, suggesting that the two interventions act as substitutes over the time horizon and skill domain we study. When benefits accruing to future cohorts are taken into account, the teacher development program becomes at least as cost-effective as, and potentially more cost-effective than, the parental engagement intervention. Our results suggest that, in this setting, strategically concentrating resources on a single binding constraint – either at home or in school – maximizes the short-run learning gains per dollar spent.