Parametric CAD Systems promise great potential with regard to product optimization. A given parametric set may be varied to produce variant instances of the geometry that further are going to be functionally assessed by means of computational methods, so-called CAE methods. If such a methodology is being embedded within a numerical optimization procedure, e.g. evolutionary strategies, the parametric set that rules the geometric and thereafter the functional behavior may converge to an optimum or at least may lead to an improvement of the corresponding product. In the present investigation some limitations of parametric CAD systems are figured, if intended to be applied in the context of numerical optimization. Specifically, the non-uniqueness of geometric constraint solving leads to robustness issues when applied in automatic geometry update cycles upon parameter variation. The suggested enhancement with regard to the geometric interpretation of distance and tangency strives towards a more reliable parametric CAD system.