Herein, a meta-analysis of the device performance and transient spectroscopic results are undertaken for various donor:acceptor blends, employing three different donor polymers and seven different acceptors including nonfullerene acceptors (NFAs). From this analysis, it is found that the primary determinant of device external quantum efficiency (EQE) is the energy offset driving interfacial charge separation, ΔE. For devices employing the donor polymer PffBT4T blended with NFA and fullerene acceptors, an energy offset ΔE = 0.30 eV is required to achieve near unity charge separation, which increases for blends with PBDTTT-EFT and P3HT to 0.36 and ≈1.2 eV, respectively. For blends with PffBT4T and PBDTTT-EFT, a 100 meV decrease in the LUMO of the acceptor is observed to result in an approximately twofold increase in EQE. Steady state and transient optical data determine that this energy offset requirement is not associated with the need to overcome the polymer exciton binding energy and thereby drive exciton separation, with all blends studied showing efficient exciton separation. Rather, the increase in EQE with larger energy offset is shown to result from suppression of geminate recombination losses. These results are discussed in terms of their implications for the design of donor/NFA interfaces in organic solar cells, and strategies to achieve further advances in device performance.