This paper develops a new framework for estimating the causal impacts on child skills and the mechanisms producing these impacts using data from a randomized control study of a widely evaluated early-childhood home visiting program. We show the feasibility of replicating the program at scale. We report estimates from standard procedures for reporting treatment effects as unweighted averages item scores and compare them with estimates adjusting for item difficulties. Such adjustments produce more interpretable estimates. We go beyond treatment effects and estimate individual-specific latent skills, comparing treatment and control skills and their impacts on test scores.