Cross-borehole radar (XBHR) is widely used for the quantification of pore-scale liquid water in geologic materials, inferred from bulk velocity variations caused by differences in electromagnetic properties between the water and the surrounding material. The XBHR can accurately and repeatedly measure variation at depth, with sampled material remaining under natural stresses, while maintaining good lateral sampling. However, even small errors in measured radar velocities result in large errors in water content estimates, emphasizing the need to quantify and minimize errors. We have rigorously assessed the sources of uncertainty in XBHR surveys undertaken in a glaciological setting. We have summarized and quantified the three main areas of uncertainty in data collection: (1) instrument time drift, (2) first-break picking, and (3) borehole geometry. Our analysis of field data indicated that contemporary acquisition procedures can produce velocity errors of [Formula: see text] ([Formula: see text]), equivalent to [Formula: see text] water content. We have developed several revisions to produce improved data acquisition. Through enhancement of existing techniques, the velocity uncertainties were improved to [Formula: see text]. We also found the measurement of borehole diameter during hot-water drilling, which could hypothetically further reduce the velocity uncertainty to [Formula: see text], equivalent to [Formula: see text] water content. The need for such precise measurement is clear because an increase in englacial water content, from 0% to 0.8%, has been proven to triple the strain rate and soften the ice. Liquid water between ice crystals has also been linked to faster velocities in ice streams and surging events.