1. The recognition of management units (MUs) that respect inter-population distinctions in management needs is central to many biological applications addressing species conservation, biological invasions and ecosystem processes. 2. We present a methodological approach for the evaluation of population-level differences in the host compatibility of natural populations of affiliate (dependent) species. Two experiments were performed to diagnose the sources of variability in the relationships between an endangered freshwater mussel Unio crassus and its host fish species in a fragmented river system in Central Europe. 3. Despite the common approach applied to the conservation of U. crassus throughout the study area, we identified differences in the ability of U. crassus to infest particular host fish species between nearby and recently isolated mussel populations. These differences could strongly influence their reproductive success under a constant conservation regime. 4. The observed differences were accompanied by genetic and morphometric distinctions, as revealed by microsatellite, mtDNA and elliptic Fourier descriptor analyses. The genetic data indicated substantial reproductive isolation, which could have resulted in adaptive or random changes responsible for the differences in host compatibility. 5. We showed that experimental testing of physiological host compatibility can be effectively used for the detection of different MUs of U. crassus, with direct implications for the management of host populations. In combination with traditional genetic and morphometric techniques, this approach may provide a general framework for the recognition and care of management units of affiliate species. 6. Synthesis and applications. Population-level evaluations of host compatibility should be utilised in management unit recognition and the formulation of management targets for affiliate species. We demonstrated that small-scale cross-compatibility testing can be effectively used to diagnose the sources of variability in host relationships with direct management implications. This approach can enhance the targeting of management actions in many biological applications addressing species conservation, biological invasions and ecosystem processes.