The condition for a successful discovery of new structural materials is a systematical analysis of collected data, its structuring and the embedding of all results in a search algorithm. Especially a formalization and evaluation of descriptors and the resulting determination of the predictor function are of primary importance. The predictor function is determined by multi-criteria-decision-making and a system of hypothesis based on formal methods. The techniques of multi-criteria-decision-making assist finding relational operators for descriptors, which are essential for the valuation function of the search algorithm. The system of hypothesis enables a validation of if- and- then relations in the collected data. Here a domain-specific language is developed that allows experts to give complex requests to the data. A proven hypothesis strengthens the predictor function as generalization. If the hypothesis is proven false, a counterexample is provided which is additional informative. Also an analytic approach describes and inverts the data and functions.