We propose an indirect approach for assessing criteria-importance weights from overall outranking statements that a decision maker may acknowledge in a multiple-criteria decision aid process. Our approach takes advantage of the bipolar characteristic valuation of the overall outranking and the stability of its associated median-cut, crisp outranking relation. Numerical criteria-importance weights are inferred with the help of a mixed-integer linear programming model for which we present tests results under different configurations and tunings.