Cellular therapy products are often thought of as generally homogeneous, with standard quality control criteria during manufacturing focusing on broad genotypic or phenotypic characterizations (e.g., proportion of cells transduced). In many cases, however, cells utilized for manufacturing are often extremely heterogeneous, each with their own potentially distinct physiological experiences, developmental trajectories, and corresponding epigenetic differences that can influence therapeutic efficacy. For instance, in a pediatric clinical study utilizing CD19 chimeric antigen receptor-modified (CAR) T cells, we found that a subset of cells within products ultimately contributes to the majority of cellular expansion and effector response in patients. Although high-resolution analysis of the pre-infusion products alone was insufficient for identifying cellular subsets with particularly high effector potential, joint analysis of pre- and post-infusion CAR T cells linked by T cell receptor sequencing allowed us to trace successful lineages to a subset of cells within the pre-infusion product. By analyzing this pre-infusion subset in the context of patient samples, we then characterized unexpected gene expression signatures and phenotypic patterns corresponding to cells that gave rise to highly effective CAR T cells in patients. Functional validation confirmed that these cells were both more cytotoxic and less prone to exhaustion than other cells in the product upon antigen encounter. Using this study as an example, we provide an experimental and analytical framework for identifying biomarkers of success in cellular therapy products.