Residential preferences are often treated as exogenous and fixed. Challenging this assumption, this article elaborates how residential preferences are shaped by experienced neighbourhood conditions. In doing so, we acknowledge the mutual dependence of the neighbourhood context, residential preferences, and segregation patterns. Applying multilevel generalized linear latent and mixed logit models to unique, geocoded data from a choice experiment, it is demonstrated how heterogenous evaluations of the social and ethnic composition of available housing alternatives’ residential surroundings systematically vary with bespoke but not administrative neighbourhoods. These heterogeneous evaluations are mostly independent of respondents’ own social and ethnic background. Controlling for unobserved neighbourhood selection, however, removes the association with bespoke neighbourhoods’ composition. Nevertheless, even after accounting for unobserved selection processes, the evaluation of the social and ethnic composition of housing alternatives in the choice experiment systematically varies across bespoke neighbourhoods, pointing to unobserved neighbourhood influences that shape people’s residential preferences.