Abstract Double-unit unrelated cord blood transplantation (DUCBT) is an option in patients for whom a single unit is not sufficient to provide an adequate number of cells. As current guidelines on UCB unit selection are mainly based on single-unit UCB data, we performed a retrospective analysis of 1375 adult recipients of DUCBT for hematologic malignancies to determine optimal criteria for graft selection. Cryopreserved total nucleated cells (TNCs; ≤3.5 vs >3.5 × 107/kg: hazard ratio [HR], 1.53; 30% vs 45%; P = .01), number of HLA mismatches (≥2 vs 0-1: HR, 1.28; 42% vs 48%; P = .01), and ABO compatibility (minor/major ABO incompatibility vs compatibility: HR, 1.28; P = .04) were independent risk factors for OS. Cryopreserved CD34+ cell dose ≥0.7 × 105/kg in the winning UCB was associated with improved OS (HR, 1.34; P = .03). Low TNC (≤3.5 × 107/kg) and CD34+ (≤1.4 × 105/kg) cell doses were related to decreased neutrophil recovery (HR, 0.65 [P = .01] and HR, 0.81 [P = .01], respectively). DUCBT recipients with ≥2 HLA mismatches had a higher incidence of grade II-IV and III-IV acute graft-versus-host disease (HR, 1.26 [P = .03] and 1.59 [P = .02], respectively). Low TNC dose (HR, 1.57; P = .02) and receiving UCB with ≥2 HLA mismatches (HR, 1.35; P = .03) were associated with increased transplant-related mortality. Our data support selecting adequately HLA-matched UCB units with a double-unit cryopreserved TNC dose >3.5 × 107/kg and CD34+ cell dose of ≥0.7 × 105/kg per unit in DUCBT candidates.