We present a novel, hybrid algorithm for quantifying the ore grade variability that has central importance in ore reserve estimation. The proposed algorithm has three stages: (1) fuzzy clustering, (2) similarity measure, and (3) grade estimation. The method first considers data clustering, and then uses the clustering information for quantifying the ore grades by means of a cumulative point semimadogram function. The method provides a measure of similarity and gives an indication of the regional heterogeneity. In addition, grade estimations can be obtained at different levels of similarity using a weighting function, which is the standard regional dependence function (SRDF).