We examine the vertical structure of the tidally driven bottom boundary layer during nearly homogeneous (N2 < 10−5 s−2) and strongly stratified (N2 ∼ 10−4 s−2) conditions in a shallow coastal region dominated by semidiurnal tides. From moored array and shipboard measurements taken at a 76‐m‐deep study site on the southern flank of Georges Bank, we infer tidal velocity profiles, bottom stress estimates, Richardson numbers, and turbulent dissipation rates. On the basis of our measurements, we discuss changes in tidal boundary‐layer dynamics in the presence of weak and strong stratification. We compare observations to results from two different one‐dimensional numerical circulation models: a two‐layer eddy viscosity model with linear eddy viscosity distribution in the lower layer and constant eddy viscosity in the upper layer (2LK), and a continuously varying eddy viscosity model (both in time and in the vertical) with Mellor‐Yamada level 2.5 closure (MY2.5). Both models compare favorably with observations during nearly homogeneous conditions, but show disagreement with data when the water column is strongly stratified. In the case of 2LK, the model overestimates the bottom stress and does not reproduce the observed velocity maximum at mid‐depth. This behavior is clearly related to the absence of buoyancy effects in the simplistic turbulence closure scheme. The advanced MY2.5 scheme, on the other hand, reproduces the observed velocity distribution and bottom stress well. However, the model also predicts an abrupt adjustment from the turbulent bottom boundary layer to a nearly nonturbulent region above which is not supported by our Richardson number estimates and observed turbulent dissipation rates. Potential reasons explaining the discrepancies between observations and MY2.5 include high‐frequency internal‐wave mixing and underestimation of the critical Richardson number used by the model to describe the transition from active to decaying turbulence.