A new salt velocity model building methodology is proposed which allows effective testing of different salt interpretation scenarios. In this methodology, we combine the strength of efficiency from interactive beam migration with the accuracy of localized RTM to derive a more accurate salt geometry. Using interactive salt geometry editing and efficient beam migration, a large number of salt interpretation scenarios are quickly tested and narrowed down to a small number of likely salt interpretation cases. This is followed by a reduced number of localized RTM runs to single out the final salt velocity model. Redatuming the wavefield from surface to a user defined subsurface datum plays a pivotal role in this methodology; it enables improvement in the quality of beam migration and in the efficiency of RTM.