Early intervention has the potential to improve the outcomes of children with autism. As more evidence-based early intervention approaches become available, data generated from studies testing their efficacy provide the unique opportunity to examine whether specific interventions differentially benefit children based on their demographic and behavioral characteristics. In this study, we retrospectively applied a probability-of-intervention-benefit framework to data from a randomized controlled trial comparing the outcomes of children with autism receiving community intervention as usual to those receiving the Early Start Denver Model (ESDM) delivered by university-based clinicians. Data from 85 children, randomized to one of the two intervention groups were analyzed. The primary outcome was the Brief Observation of Social Communication Change, a video coded measure of autism related social communication and repetitive behaviors. Children were classified as making reliable change or not on this outcome based on their reliable change index. 73% of children in the ESDM group and 57% in the community group made reliable improvements over the 24 months of intervention. The probability of intervention benefit in the community group depended on children's baseline characteristics. Those with more significant cognitive delays and greater autism severity were less likely to make reliable change in the community group, but not the ESDM group. Overall, the data highlight the variability of change that individual children make in early interventions, the role of various factors contributing to change, and finally, the importance of analyzing change at both the group and the individual levels.