Photoinduced charge-transfer (CT) processes play a key role in many systems, particularly those relevant to organic photovoltaics and photosynthesis. Advancing the understanding of CT processes calls for comparing their rates measured via state-of-the-art time-resolved interface-specific spectroscopic techniques with theoretical predictions based on first-principles molecular models. We measure charge-transfer rates across a boron subphthalocyanine chloride (SubPc)/C60 heterojunction, commonly used in organic photovoltaics, via heterodyne-detected time-resolved second-harmonic generation. We compare these results to theoretical predictions based on a Fermi’s golden rule approach, with input parameters obtained using first-principles calculations for two different equilibrium geometries of a molecular donor–acceptor in a dielectric continuum model. The calculated rates (∼2 ps–1) overestimate the measured rates (∼0.1 ps–1), which is consistent with the expectation that the calculated rates represent an upper bound over the experimental ones. The comparison provides valuable understanding of how the structure of the electron donor–acceptor interface affects the CT kinetics in organic photovoltaic systems.