Previous diagnoses of surface seismic velocity models with vertical seismic profile (VSP) data in the Gulf of Mexico have indicated that shallow velocities were poorly constrained by VSP due to ringing caused by multiple casing strings. This ringing also hampered direct measurement of the seawater average velocity (SWAV) at a rig site with direct arrivals of a zero-offset VSP (ZVSP). We have directly measured the SWAV at a rig site with a known water depth by using differential times between primary water bottom multiples (WBMs) and direct first arrivals acquired in a marine VSP survey. We developed a procedure to process ZVSP-WBM signals for SWAV measurement. This WBM method is successfully applied to VSP data recorded at 27 rig sites in the deep-water environments of North and South America. Our results suggest that VSP processors should implement this method and add the SWAV measurement in their future velocity survey reports. We have estimated water bottom depths using differential times. We found that the estimated water depths are comparable with those acquired from sonar measurements by autonomous underwater vehicles, but with large uncertainties. The WBM method is extended by using data from a vertical incidence VSP to measure a profile of the SWAV along the path of a deviated well and evaluate possible lateral variations of SWAV. This method can potentially be applied to a time-lapse VSP to monitor temporal variations of SWAV. We also evaluated the application scope and limitations of the WBM method.