This paper presents a new two-step approach for analyzing big data with a binary outcome variable at the child level that is related to factors at both the child and parent levels. The first step develops a predictive model at the child level using logistic regression. The second step identifies the impact of parent-level variables on the child outcome using linear regression, with the difference between actual and expected outcomes as the dependent variable. The final model combines the child-level and parent-level factors to predict the binary outcome, weighting the contributions from each level. This two-level modeling approach has been successfully used to evaluate nursing home and customer retention performance.