The in silico assignment of polyhydroxylated compounds represents a major challenge given the thus far unsolved problem of inappropriate description of the conformational amplitudes. Herein, we report a conceptually novel stochastic approach based on the creation and evaluation of random artificial ensembles, which could provide a new paradigm for computing NMR properties of flexible molecules. The strategy was successfully tested under the DP4/DP4+ platforms using a large set of compounds belonging to the hyacinthacine family.