A novel "Prediction and Confirmation" (PC) strategy was proposed for characterizing phosphodiesterase-5 inhibitor (PDE-5) derivatives in botanical dietary supplements (BDSs) for on-site detection. Discovery Studio (DS) and density functional theory (DFT) calculations were used for the "Prediction" step in order to estimate PDE-5 derivative structures and theoretical Raman shifts without synthesizing the derivatives. After 11 potentially bioactive sildenafil derivatives were acquired through DS, 32 common calculated Raman shifts were obtained through DFT. The mean absolute wavenumber deviation (δ, peak range) of the major bands and the minimum number (τ) of Raman spectral peaks matching the calculated common shifts were optimized, so that a positive result of an unknown sample could be reasonably produced. In this study, δ was set at ±10 cm-1 and the corresponding τ was set at 4-5 after optimization. Surface plasmon resonance (SPR) biosensor and surface-enhanced Raman scattering (SERS) detection were the "Confirmation" step to validate the reliability and accuracy of DS and DFT in the "Prediction" step, respectively. The optimized δ and τ criteria were used as indexes for on-site SERS detection after thin-layer chromatographic (TLC) separation of six real-world samples, one of which was preliminarily identified as "suspected positive samples." This strategy allows for a quick determination of the BDSs adulterated with sildenafil or its derivatives, independent of any standard materials.