On steep beaches, the cross-shore movement of sand in response to ‘erosive’ storm waves and ‘accretive’ swell waves can lead to temporal changes between a barred winter profile and a non-barred summer profile with a pronounced berm in the upper swash zone. Despite recent improvements in predicting berm formation and evolution within process-based morphodynamic models, substantial demand for improvement in understanding swash processes and associated surf–swash sand exchange remains. Here, we analyze bed level data collected on a near-prototype, 4.5-m high and 75-m wide sandy beach (median grain diameter D50=430μm) with a lagoon situated at its landward side. In particular, we distinguish between surf–swash sand exchange (time scale of tens of minutes to hours), the net effect of single and multiple swash events on the entire beach face (time scale of a few seconds to hours), and instantaneous bed variability at 3 cross-shore locations within individual swashes. During ‘erosive’ waves (Hs=0.8m, Tp=8s) sand on the initially 1:15 planar profile was predominantly eroded from the inner surf zone to be deposited in the outer surf zone as a sandbar, indicating minimal surf–swash sand exchange. Subsequent ‘accretive’ waves (Hs=0.6m, Tp=12s) caused substantially larger surf–swash sand exchange: the pre-existing sandbar migrated onshore and decayed, with the sand ending up on the beach face in a prominent (up to 0.7m high), steep (1:6) berm. We found the dynamics of the berm to be governed primarily by wave conditions and the antecedent morphology, with ground water gradients of additional importance when morphodynamic feedback between swash flow and the berm is small. The observed bed level change within a swash and averaged over a swash event could be substantial (several centimeters) during all wave conditions, but the net (i.e., averaged over multiple swash events) bed level change was strongly suppressed because erosive and accretionary swashes nearly balanced. In addition, the local beach face slope could be instantaneously ≈25% steeper or shallower than the median slope, or the initial or final slope. We anticipate that our data will stimulate new model development, as to increase the range of conditions and settings in which morphodynamic models can be applied realistically and reliably.