We show for a series of six small donoracceptor dyads that the energy difference between non-charge transfer (non-CT) and charge transfer (CT) excited states, as well as the squares of the electronic couplings between these states, can be predicted from first-principles using variational orbital adapted configuration interaction singles (VOA-CIS) theory. VOA-CIS correctly predicts the observed experimental trends in these values and provides quantitative accuracy roughly on par with a modern long-range corrected density functional, omega B97X. Using VOA-CIS and ?B97X, the experimental energy difference between the non-CT and CT excited states is predicted with root mean squared errors of 0.22 eV and 0.21 eV, respectively. The square of the electronic coupling between these states is predicted with root mean squared errors of 0.08 eV(2) and 0.07 eV(2), respectively. Orbital optimized CIS (OO-CIS) and CIS(D), two perturbative corrections to CIS, provide a significant correction to the errant relative energies predicted by CIS, but the correction is insufficient to recover the experimentally observed trend.