Aim: Understanding how ecosystem functioning and evolution shape taxonomic as- semblages is a lively debate basically involving two major opposite views: the niche- and dispersal-assembly hypotheses. Here, we introduce a new method allowing for the identification of the first-order process of assembly underlying a set of taxonomic assemblages. Methods: Building on Clarke’s SIMPER (for “similarity percentage”) analysis of a taxon/ locality occurrence data set, we develop a permutation-based algorithm named PER- SIMPER, allowing for the identification of the first-order process—either niche- or dispersal-assembly—that drives species distribution within two or more groups of assemblages. We demonstrate the reliability and robustness of the method through cellular automaton-like simulations generating niche-assembled and/or dispersal-as- sembled species occurrence data sets. Sensitivity analysis further allows evaluation of its accuracy and robustness to sampling effort, including reduced numbers of sampled localities and/or species. Main conclusions: Niche- and/or dispersal-assembled communities generate very dif- ferent SIMPER profiles, which, in turn, allow for the accurate and consistent identifica- tion of the first-order process of assembly operating within two or more groups of species assemblages through a threefold randomization procedure named PER-SIMPER. The PER-SIMPER method appears robust to varying sampling efforts that may affect the number of sampled localities and/or species, especially when one of the two processes of assembly dominates the other. The PER-SIMPER analysis can be achieved on any empirical occurrence data set using a dedicated R function available as Supporting Information.