The input of the FACTORFINDER-Software is a list of all acceptable variants. The analyst has to decide during analysis what factors shall andergo the evaluation by the program. The user interface guides the analyst through the recording of data. At the same time, the completeness and consistency of the input is verified by the program.
It has not to be predetermined by the user what values are ideal as this will be the result of the analysis. Additionally, there is no need to differenciate between different degrees of acceptablity.
The FACTORFINDER-Software compares all examples pairwise. Each pair justifies different inferences in respect to the relevance of the analysed factors. As described in the previous chapter, this is done by exploring the tolerability of non-ideal values. Initially, a random tolerance hypothesis is generated by the FACTORFINDER-Procedure. The tolerance test vector is put into relation with the known variants and their respective success and thereby examined in respect to its plausibility. Based on this test a new tolerance hypothesis is generated. This tolerance hypothesis differs slightly with random variations. Again, this new tolerance hypothesis is analysed in relation with the example base. In repeated adaption cycles the FACTORFINDER-Software improves its tolerance-/relevance-hypothesis. Through computer assistance thousands of adaption cycles with new hypothesis and plausibility tests are feasable with little effort.
Using the FACTORFINDER-Procedure the analyst implicitly assumes that values of a factor are either goal-promoting or goal-inhibiting. Only irrelevant factors act neutral. If each factor is associated with only one single ideal value and each factor is critically important, then only one single variant can be observed as being of acceptable success. All other alternatives will fail. Whenever more than one value can be observed in successful variants, two options are left: Either the factor is critically important. Then each value has to be ideal and goal-promoting as this factor grants zero tolerance against non-ideal values. Or the second option holds true: The factor is less than critically important. Subcritically important factors tolerate non-ideal values in successful variants. One or more factors may show non-ideal values. The more ideal values exist and the less important the factors are, the more acceptable variants can be observed. Non-ideal values may let an alternative fail - but they must not. Non-ideal values lead to failure - the more important a factor the more likely failure is. The effects of several non-ideal values add up - as assumed by the FACTORFINDER-Procedure - to a value that determines success or failure of an alternative.
Provided that the example base contains enough examples, as a result the FACTORFINDER-Software identifies the ideal values, the relevance of the factors and the variants that successful despite differing extremely from those variants that are known to be ideal.
The FACTORFINDER-Software may be applied to a broad range of analytical problems. The analytical tasks are characterized by the interaction of non-metric factors and by variants whose success can be measured binary as acceptable or inacceptable.