The streamwise effective slope ( \({\rm ES}_x\) ), which is the mean absolute streamwise gradient of the roughness, is considered to be a key parameter in predicting the drag penalty of rough-wall turbulent boundary-layers. However, many real-world rough surfaces are multi-scaled. For such surfaces, \({\rm ES}_x\) can be unbounded and its value can be dominated by scales of the topography that are invisible to the flow. To illustrate this, a campaign of drag balance measurements was conducted with a set of machined surfaces. A baseline surface is prepared with fine machining parameters. By coarsening the machining precision, artefacts called ‘scallops’ are introduced which increases the ‘measured’ \({\rm ES}_x\) without changing other geometrical statistics. The drag of the ‘scalloped’ surfaces is higher than the baseline surface, with the relative drag increase scaling with the viscous scaled scallop height, but only when their height exceeds \(2\sim 3\) times the viscous length scale. Further, the drag of one of the ‘scalloped’ cases, even when the scallop height is \(\mathcal {O}(10)\) viscous units, is seen to be much lower ( \(\sim 28\) %) than a case with matched \({\rm ES}_x\) , but where the \({\rm ES}_x\) results from larger scale features. These findings confirm that for multi-scaled surfaces, \({\rm ES}_x\) may be a misleading topographical metric for drag (or \(k_s\) ) prediction and one must consider which scales contribute to the average slope of a surface.
