The Capacitated Facility Location Problem (CFLP) is a core problem in location science. We show that the combinatorial element of the MIP formulation manifests itself to different degrees in the level of interdependence between facilities serving similar subsets of customers. This induces an implied separation of the sets of candidates and customers into regions within which location decisions interdepend more strongly. Although these regions are easily identifiable in visual representations of allocation decisions or the spatial distribution of candidates and customers, detecting them solely from decision vectors is challenging. We show that spectral biclustering, a pattern recognition technique, can be used to retrieve implied regions from linearly relaxed solutions. This opens novel directions for algorithm development.