Utility function is an interesting tool generally used to measure the preference over a set of attributes. Calculation of the utility function is possible by substituting preferences related assumptions in predefined mathematical models. The method may be challenging while the number of preferences available are reduced. The paper, presented herein, proposes a new methodology for a more accurate estimation of the utility function. Two experimental designs tools are suggested to find the best approximation to the utility function meeting decision makers preferences. First, Latin Hypercube Sampling is employed for filling the experimental space with meaningful representative data. Second, the model response surfaces are analyzed and optimized.