This research investigates the effectiveness of a non-convex clustering criterion with the ability to discriminate clusters by means of quadratic boundaries that take into account cluster variance. No algorithms have been shown to work efficiently and effectively for this kind of criterion, so we introduce and evaluating a generalized version of the incremental one-by-one clustering algorithm of MacQueen (1967) that is suitable for general variance-based criteria, whether convex or otherwise. Results show that the criterion performs remarkably well, both in synthetic and real-world data sets. We also have successfully tested the approach on the financial domain problem of finding opportunities to short sell securities that translate into good investment decisions.