The aim of this work is to verify if (1) the average of shear-wave velocity from the surface to 30 m depth (V-S30) is a good proxy of site amplification in a country with complex geology like Italy, (2) the grouping of V-S30 in different soil classes and relevant spectra in the Italian seismic codes is adequate, and (3) shallow shear-wave profiles (V-S10) could be a more economical tool for site classification. To answer the first two questions, we examined 40 estimates of V-S30 derived from microzonation projects we performed in Italy (27 obtained with downhole measurements, plus 13 velocity profiles obtained with surface techniques). In all the sites, we installed seismic instrumentation to record earthquakes and to estimate site response using horizontal to vertical spectral ratios (HVSR). The comparison between HVSR and V-S30 showed in about one-third of the sites that V-S30 is not a good proxy of observed amplification effects if the site does not have a monotonically increasing velocity profile. The reason V-S30 does not provide satisfactory estimates in Italy is linked to peculiar geological settings that are widespread in the country. We then compared the observed amplifications from earthquake HVSR data with the ones provided by the Italian seismic code, noting a substantial underestimation by the code, a somewhat unsettling situation because HVSR is usually considered to be a lower bound for amplification estimated with other techniques. Finally, we studied 45 V-S downhole profiles to 30 m depth performed also at sites where earthquake recordings are not available. On this data set, we noticed that V-S10 could predict site classification with the same performances of V-S30. We consider alternative soil classification schemes that include soil frequency besides the velocity profile. In this two-parameter approach, V-S10 could be substituted for V-S30.