An association between salt pans or dry lake beds and distinctive crescentic lake‐floor sand mounds (1–10 m high, tens to hundreds of metres wide) is commonplace in desert systems. In the Makgadikgadi Basin of northern Botswana, a debate about the formative processes of these landforms has persisted despite numerous morphometric, sedimentary and geochronological analyses, with mound landforms variously inferred to be aeolian dunes, subaqueous dunes, spring mounds or shoreline remnants. We propose a new formative mechanism which draws on the interaction between uneven moisture distribution on the pan surface and mobile aeolian sediments. We use a numerical model (ViSTA), which couples vegetation and aeolian sand transport dynamics, together with optically stimulated luminescence (OSL) dating of a mound in the Makgadikgadi Basin to investigate the feasibility of this ‘sticky mound hypothesis’. We find that under a range of modelled environmental conditions, uneven moisture distribution on the pan surface can lead to the development and stabilization of crescentic aeolian dunes, with these dunes growing upwind from the point of initial deposition, corresponding with the chronological data gained from OSL dating of a mound feature. On removal of this moisture, the modelled dunes erode and dissipate. These findings suggest that the formative mechanism of the mounds could be dependent on the interaction between differential drying of the pan surface and the competence of the aeolian sediment transport system across the pan floor.