Conclusion

Previous  Top  Next

The FACTORFINDER-Relevance-Model is based on the theoretical construct of ideals. Ideals are not real. They are composed of specific values in each factor. Failing to accomplish these specific values in real examples may lead to a result with inacceptable success. But goal-adverse values do not neccessarily lead to failure. The absence of ideal values is tolerated in real examples to a certain extend. The value that represents the importance of that factor is measured by its tolerance in respect of non-ideal values.
The analysis is difficult as we can not directly judge from observed variants whether a value acts goal-promoting or goal-adverse. The only thing we can be sure of is that all goal-adverse values of a successful alternative are not important enough to let that variant fail. Whenever different values of one factor can be observed in acceptable variants, there are two options left open:

·

Either a factor is critically important. Then each value must be ideal as this factor grants zero tolerance against non-ideal values.

·

Or factors are less than critically important. Then a factor of an acceptable alternative can show non-ideal values.

To the analyst, it is not known and only can be infered, whether factors of a real example show non-ideal values. Alternatives more likely fail the more important the a non-ideal factor is. In factors with little relevance, any values are tolerated. The tolerance of non-ideal values depends on the relevance of the factor.
The class of problems the FACTORFINDER-Model is designed for is characterized by a threshold value of the target criterion. Because of this threshold value even non-ideal variants are considered as being successful just as if they were ideal. That holds true as long the threshold of the target criterion is surpassed. The reward structure of the object analysed is characterized by a threshold value that divides all variants into a group of acceptable alternatives and a group of inacceptable alternatives.
In situations where only one allocation of resources reaches the threshold acceptably and all other variants fail, all dimensions analysed seem to be important. Other values in any factor are not tolerated without endangering the success of the variant. That leads to the conclusion that each factor is critically important.

A vehicle without spare wheel depends critically on each wheel. As soon different alternatives exist (e.g. a spare wheel could replace one wheel), the existance of this wheel would be tolerable without endangering the function of the vehicle. Now we can assign different grades of importance to each wheel.

The same can be explained with an economic example: Assume a restaurant with its location and its menue as critically important factors for its success. A restaurant with an inacceptable location or with an inacceptable menue would remain unsuccessful. Would the menue be less than critically important, then a restaurant with any menue could be successful - as long as all the other factors remain ideal. Whether this restaurant fails if other factors turn non-ideal depends on the importance of the other non-ideal factors.

This example shows that the underlying notion of relevance. Different grades of relevance depend on the interaction of a multitude of factors. The more acceptable alternatives, that differ from each other in different combinations of values and still are successes, are known, the less important the differing factors appear. Non-ideal values lead to failure. The more important a factor is, the more likely non-ideal values contribute to failure. At the same time, the analyst does not know with certainty what the ideal values are. It can only be infered from comparisons of all known examples.

As the relevance model is designed for categorial (and not continuous) factors, we can differenciate values in goal-promoting and goal-adverse. Different grades - as it is senseful with continuous factors - seems inadequate with categorial factors. If several factors show non-ideal values, the relevance of all non-ideal factors adds up to a sum that determines the success of the alternative. This value is called deviation. The effect of non-ideal values remains hidden as long all factors with non-ideal values are together not important enough to let that variant fail. That is why it seems as if the influence of non-ideal values is compensated by other, ideal values. This supposed compensation fails as soon further factors turn non-ideal. Just on its own, this non-ideal factor would not lead to failure of that alternative. But in conjunction with other - taken on their own non-critical - factors, this minor influence leads to failure of the alternative. That justifies to call it a non-compensatory relation between all factors.

The value that represents the relevance of the factors faciliates comparisons between variants. The sum of the relevance of all differing factors is called variance. The variance is either subcritically or surpasses the threshold value. It is plausible, if an example base contains for each acceptable example at least one other successful example with a subcritical variance.
On the other hand, if an example with a subcritical variance is missing from the example base, this indicates a strongly differing example or an inappropriate relevance hypothesis. The benefit of this discovery is the indication of changes within the object to be analysed. This may suggest further inquiries.
The FACTORFINDER-Procedure describes the influence of ideal and non-ideal values on the success of variants. The FACTORFINDER-Procedure requires as data input to record all known alternatives and to mark them as being acceptable or inacceptable. A more sophisticated measurement of success is superfluous. As well a model of the interaction - any further than the additive relation between all factors - is superfluous. It may be left open, what values act goal-promoting. This as well as the relevance hypothesis is the result of the analysis.