Evidently there are competing concepts of relevance. One concept builds the attribute relevance on the frequency of (co-)occurence of observations. Or one can build on the concept of ideal values whose absence is tolerated to a certain extend. This tolerance-oriented approach is taken here.
Critically important factors must show certain values to make a variant successful. These factors must have ideal values. Subcritically important factors can show non-ideal values to a certain extend without endangering success of the variants. The more important a factor is the less non-ideal values are tolerated. The example base contains variants that fulfill the target criterion acceptably. Therefore we can infer how tolerable a factor is. The absence of a specific value is called tolerable if it does not endanger the success of the alternative. Values of a factor are tolerable if there is at least one acceptable variant that differs in this factor and many more factors. It has to be stressed, that the importance/tolerance of a factor inherently depends on the importance/tolerance of all other factors.