Many phenomena of interest to science and engineering occur at interfaces; however, access to, or discrimination of, interfacial interactions has been challenging, especially at buried interfaces. Vibrationally resonant sum-frequency generation (VR-SFG) spectroscopy is a powerful tool for investigating the molecular structure of free or buried interfaces, but spectral analysis has relied on many assumptions. To claim accurate new insights, practitioners must be able to make unique determinations from the data without experimental artifacts affecting the final results. For example, two independent and overlapping studies for the polystyrene/air interface were carried out, but reported different surface structures. Initially, this difference was attributed to the use of different substrates, but we have shown that the surface structure is independent of substrate by experimental suppression of the interfering nonresonant signal. These results show difficulties in SFG analysis that have led to faulty determinations of structural changes. Similar problems have been observed in systems assumed to have negligible nonresonant interference, demonstrating the need for proper experimental design instead of relying solely on post-experimental analysis of the data. We have investigated the inherent limitations imposed on the technique from the nature of the signal generation and nonresonant interference, and have developed methods to overcome such difficulties, depending on what is desired from the data. By nature of nonlinear spectroscopy, the desired frequency response is affected by overlapping interactions in the time domain, and these time domain interactions can be exploited to overcome challenges in analysis. By delaying the upconverting pulse, the nonresonant signal can be removed to enable accurate qualitative comparison or even quantify change; however this removal results in incomplete upconversion, or apodization, of the resonant signal, causing distortion in the observed resonant response. If absolute parameters are desired, additional work is necessary to correct the distortion of the resonant response. Correction can be accomplished by further exploiting time domain effects by collecting spectra at various delay times of the upconverting pulse, and this additional data also aids in interpretation of congested spectra. Many practical applications, however, only require a means to quantify change, and measurements of change are unaffected by the effects of apodization. These techniques have been used to more accurately analyze polystyrene and octadecylsilane surfaces.