Moving boat acoustic Doppler current profiling (ADCP) is increasingly used to measure discharge in large rivers. We investigate whether useful information about bed shear stress can be recovered from such data. Alternative ways to estimate local bed shear stress using the logarithmic law of the wall and spatial averaging are tested using ADCP transects across lower Fraser River, Canada. Repeatability is assessed by comparing estimates from outward and return boat tracks. The most precise method uses the vertically averaged mean velocity and a zero-velocity height based on bed grain size information. The accuracy of the assumed zero-velocity height can be judged by consistency between estimates using mean velocity and near-bed velocity. Shear stress estimates from unconstrained log-law fits are less repeatable and tend to overpredict, and mean shear stress estimates using the depth-slope product are unreliable in this river because of nonuniform flow.