When having a set of measurement samples Θ at disposal, it is useful to create a definition of the region enclosing the samples — the domain of discourse. If the samples represent a system performing ℝⁿ → ℝ mapping, then this is about defining such region within ℝⁿ. In the case of clustering, the outcome region should be the sum of regions defined for single clusters. After having defined numerically this region, one can check if a certain testing sample (newly acquired), to be used as input to a ready model, is placed inside the validated region (VR) or outside it — in non-validated region (nVR). Such test allows to state whether the model output for this sample should be considered interpolation or extrapolation effect, and hence, how credible it is. This paper shows what approaches to VR definition could be chosen. Also, the paper presents the algorithm, the implementation of which would automatically define VR by producing a set of inequalities, forming the minimal convex region.