Persistence of high youth unemployment and dismal labour market outcomes are imminent concerns for most European economies. The relationship between demographic ageing and employment outcomes is even more worrying once the relationship is scrutinized at the regional level. We focus on modelling regional heterogeneity. We argue that an average impact across regions is often not very useful, and that - conditional on the regions characteristics - impacts may differ significantly. We advocate the use of modelling varying level and slope effects, and specifically to cluster them by the use of latent class or finite mixture models (FMMs). Moreover, in order to fully exploit the output from the FMM, we adopt self-organizing maps to understand the composition of the resulting segmentation and as a way to depict the underlying regional similarities that would otherwise be missed if a standard approach was adopted. We apply our proposed method to a case-study of Germany where we show that the regional impact of young age cohorts on the labor market is indeed very heterogeneous across regions and our results are robust against potential endogeneity bias.