Forming Limit Curves are surrounded by uncertainty - in particular the uncertainty surrounding the calculation of limit strains from experimental data. The position dependent method outlined in ISO12004-2 is known to make assumptions regarding the behavior of sheet metal at the onset of necking. Recent time dependent methods better utilize measured strain data from digital image correlation (DIC) techniques to determine the onset of localized necking. Both approaches generate a series of forming limit strains which require subsequent interpretation to produce a deterministic forming criterion. By acknowledging the inherent heterogeneous plastic deformation behavior of sheet metal, a new statistical approach to quantifying formability is proposed. Gaussian Mixture Models are used to characterize DIC-measured strain data and to determine both onset of localized necking and the size and location of the neck itself. By avoiding any pre-conceptions regarding the size or characteristics of the localized neck, a more realistic and robust probabilistic forming criterion is attainable.