Obstacle marks are instream bedforms, typically composed of an upstream frontal scour hole and a downstream sediment accumulation in the vicinity of an obstacle. Local scouring at infrastructure (e.g. bridge piers) is a well‐studied phenomenon in hydraulic engineering, while less attention is given to the time‐dependent evolution of frontal scour holes at instream boulders and their geometric relations (depth to width, and length ratio). Furthermore, a comparison between laboratory studies and field observations is rare. Therefore, the morphodynamic importance of such scour features to fluvial sediment transport and morphological change is largely unknown. In this study, obstacle marks at boulder‐like obstructions were physically modelled in 30 unscaled process‐focused flume experiments (runtime per experiment ≥ 5760 min) at a range of flows (subcritical, clear‐water conditions, emergent and submerged water levels) and boundary conditions designed to represent the field setting (i.e. obstacle tilting, and limited thickness of the alluvial layer). Additionally, geometries of scour holes at 90 in‐situ boulders (diameter ≥ 1 m) located in a 50‐km segment of the Colorado River in Marble Canyon (AZ) were measured from a 1 m‐resolution digital elevation model. Flume experiments reveal similar evolution of local scouring, irrespective of hydraulic conditions, controlled by the scour incision, whereas the thickness of the alluvial layer and obstacle tilting into the evolving frontal scour hole limit incision. Three temporal evolution phases—(1) rapid incision, (2) decreasing incision, and (3) scour widening—are identified based on statistical analysis of spatiotemporal bed elevation time series. A quantitative model is presented that mechanistically predicts enlargement in local scour length and width based on (1) scour depth, (2) the inclination of scour slopes, and (3) the planform area of the frontal scour hole bottom. The comparison of field observations and laboratory results demonstrates scale invariance of geometry, which implies similitude of processes and form rather than equifinality.