Theoretical gravitational-wave models of compact-binary mergers need to be accurate, but also fast in order to compare millions of signals in near real time with experimental data. Various regression and interpolation techniques have been employed to build efficient waveform models, but no study has systematically compared the performance of these methods yet. Here we provide such a comparison. For analytical binary-black-hole waveforms, assuming either aligned or precessing spins, we compare the accuracy as well as the computational speed of a variety of regression methods, ranging from traditional interpolation to machine-learning techniques. We find that most methods are reasonably accurate, but efficiency considerations favor in many cases the simpler approaches. We conclude that sophisticated regression methods are not necessarily needed in standard gravitational-wave modeling applications, although machine-learning techniques might be more suitable for problems with higher complexity than what is tested here.