To date, much effort has been made to overcome obstacles involved in inferring the provenance of sediments based on heavy mineral assemblages and geochemical fingerprints of minerals. These efforts have been concentrated on two main avenues. First, efforts have focused on disentangling physical and chemical properties of minerals and their behavior during transport. Second, efforts have centered on processing the obtained data in a way that takes into account the properties of heavy minerals and geochemical data sets. More specifically, to consider the limitations that arise from the nature of point count and compositional data. In this manuscript, a novel workflow for provenance is proposed based on the fact that point count data, although not compositional, carry compositional information. A case study is presented to infer the provenance of late Quaternary eolian sands from the southeastern Adriatic archipelago for which two different sources have been previously proposed. Heavy mineral data were represented in orthonormal log ratio coordinates. In this way, variables which compensate for hydraulic sorting and/or selective dissolution were obtained, with the major advantage that there are no restrictions for their use with statistical methods originally developed for real Euclidian space. To support conclusions based on heavy mineral data, we analyzed the geochemical composition of pyroxenes, which were also discriminated using a novel approach. As a result, the Albania-Montenegro shelf is determined to be a sediment source for eolian sands from Korčula, Lastovo and Mljet Islands. Hvar Island has a Neretva River signature, and Vis Island has a dominantly Neretva River signature with a small portion of sediments from the Albania-Montenegro shelf.