To test the hypothesis that the outcome of hematopoietic stem cell (HSC) grafts is at least partially determined by the cellular composition of the graft, the National Marrow Donor Program (NMDP) analyzed the correlation of cellular phenotypes of unrelated grafts with graft outcome. Samples from 94 bone marrow (BM) and 181 peripheral blood progenitor cell (PBPC) grafts for transplantations at 40 U.S. transplant centers between 2003 and 2005 were analyzed at a single immunophenotyping reference laboratory. Samples were shipped from transplant centers upon receipt of graft. Graft cellular composition included analysis of leukocyte total cell numbers, and subsets of myeloid [CD34⁺, CD34⁺ CD38⁻], lymphoid [CD3⁺, CD3⁺ CD4⁺, CD3⁺ CD8⁺], and activated lymphoid cells [CD3⁺ CD25⁺, CD3⁺ CD69⁺, CD3⁺ HLA-DR⁺] coexpressing CD3⁺. There was substantial variability in the cellular composition of BM and PBPC grafts before and after graft processing by red blood cell (RBC) removal or plasma depletion in preparation for transplant. With BM grafts, cellular composition was not associated with hematopoietic recovery, graft-versus-host disease (GVHD), or survival. With PBPC grafts, survival rates were higher with CD34⁺>5×10⁶/kg, 59% compared to 34% with CD34⁺≤5×10⁶/kg at 1 year. Platelet recovery was higher with PBPC containing CD3⁺ CD8⁺ >8×10⁷/kg. Neutrophil recovery or GVHD could not be predicted by any cellular subsets of PBPC grafts. Although survival was superior with PBPC grafts containing >5×10⁶ CD34⁺/kg, an optimal graft mix of myeloid, lymphoid, and activated lymphoid subsets was not identified.