This paper studies the reliability of the calculated shear-wave velocity ($$V_{\mathrm{S}}$$ VS ) from different available $$V_{\mathrm{S}}$$ VS and SPT-N correlations in terms of seismic site response analysis. In the present study, various $$V_{\mathrm{S}}$$ VS –N correlations developed for different regions around the globe have been used to calculate the bound of $$V_{\mathrm{S}}$$ VS variations with depth at three different sites in Kolkata city. This bound has later been used to generate the random $$V_{\mathrm{S}}$$ VS profiles using the Monte Carlo simulation. Equivalent linear site response analysis has been performed to study the response of those generated profiles under different input motion excitations. Strong-to-weak ground motion records have been used for this purpose. The amplification spectra of the generated $$V_{\mathrm{S}}$$ VS profiles using all soil types and specific soil-type $$V_{\mathrm{S}}$$ VS –N correlations show significant variations. The study also shows that the $$V_{\mathrm{S}}$$ VS –N correlation may result in quite different $$V_{\mathrm{S},30}$$ VS,30 values and subsequently it may lead to the different site classes according to the NEHRP 2003 classification. So, the random choice of the $$V_{\mathrm{S}}$$ VS –N correlation, where the direct measurement of $$V_{\mathrm{S}}$$ VS is not available, may affect the outcome of seismic hazard analysis significantly. The study points out the need for accurate estimation of the $$V_{\mathrm{S}}$$ VS profile either from in-situ determination or using site-specific correlation.