Results of 2 parallel phase 2 trials of transplantation of unrelated umbilical cord blood (UCB) or bone marrow (BM) from HLA-haploidentical relatives provided equipoise for direct comparison of these donor sources. Between June 2012 and June 2018, 368 patients aged 18 to 70 years with chemotherapy-sensitive lymphoma or acute leukemia in remission were randomly assigned to undergo UCB (n = 186) or haploidentical (n = 182) transplant. Reduced-intensity conditioning comprised total-body irradiation with cyclophosphamide and fludarabine for both donor types. Graft-versus-host disease prophylaxis for UCB transplantation was cyclosporine and mycophenolate mofetil (MMF) and for haploidentical transplantation, posttransplant cyclophosphamide, tacrolimus, and MMF. The primary end point was 2-year progression-free survival (PFS). Treatment groups had similar age, sex, self-reported ethnic origin, performance status, disease, and disease status at randomization. Two-year PFS was 35% (95% confidence interval [CI], 28% to 42%) compared with 41% (95% CI, 34% to 48%) after UCB and haploidentical transplants, respectively (P = .41). Prespecified analysis of secondary end points recorded higher 2-year nonrelapse mortality after UCB, 18% (95% CI, 13% to 24%), compared with haploidentical transplantation, 11% (95% CI, 6% to 16%), P = .04. This led to lower 2-year overall survival (OS) after UCB compared with haploidentical transplantation, 46% (95% CI, 38-53) and 57% (95% CI 49% to 64%), respectively (P = .04). The trial did not demonstrate a statistically significant difference in the primary end point, 2-year PFS, between the donor sources. Although both donor sources extend access to reduced-intensity transplantation, analyses of secondary end points, including OS, favor haploidentical BM donors. This trial was registered at www.clinicaltrials.gov as #NCT01597778.