High‐resolution imaging with microseismic events requires the use of large and consistent data sets of seismic phase arrival times. In particular the S phase is important to derive physical parameters of the subsurface. Typically this phase is identified on one of the horizontal seismogram components by a change of signal amplitude and frequency as compared to the previous P phase. However, reliable S‐phase identification can be difficult for local events because of a signal overlap with the P coda, the presence of converted phases, and possible S‐wave splitting due to anisotropy. In this study we propose a new data processing technique aiming at uniquely identifying the S‐phase arrival using all available records from a seismic network. The technique combines polarization analysis of single three‐component recordings of an event with analysis of lateral waveform coherence across the network. This makes it possible to construct seismic sections in which the first arrival is the S phase. This graphical representation can support an operator in both the analysis of single events and in semiautomatic analyses of large datasets. In addition, an automated stacking velocity analysis provides S‐wave velocities from these sections. We demonstrate the applicability of this technique using synthetic seismograms, and we evaluate the efficacy on a dataset of three‐component velocimeter records from local earthquakes of the Campania–Lucania Apennines (southern Italy) recorded by the Irpinia Seismic Network (ISNet).