Aggregate-level profile-based lectometry allows us to compare language varieties while aggregating over a wide range of linguistic features, thus allowing us to gauge robust, global differences between varieties. Moreover, refinements of the workflow were implemented that allow the researcher to have the best of both worlds, in the sense that next to global patterns, also information on how individual features relate or contribute to these patterns is visible. Even with these refinements, however, a remaining limitation of aggregate-level lectometry methods is that they assume, without testing, that the set of linguistic features fed to the method constitutes a “sensible set of features”. Sometimes, however, it is an empirical question whether aggregating over all features (as opposed to e.g. building separate solutions that aggregate over subsets of features) is the best choice. In this paper we propose a workflow, relying on Procrustes distances, that can help address this empirical question.