The FACTORFINDER-Procedure uses each acceptable variant as a fix point for comparisons. One by one example is selected as fix point for comparisons with all other successful alternatives. By this, it is evaluated how tolerable it is not to fulfill specific values in each factor. Several evaluation-cycles repeatedly take place. For each acceptable variant there finally is a hypothesis on how tolerable non-ideal values in each factor are. This hypothesis is called the intermediate result vector. All intermediate result vectors are condensed to one single relevance hypothesis. Counter values serve for processing and memorizing competing tolerance hypothesis. The counter values express how strong a relevance/tolerance hypothesis is supported by the available examples. All counter values together form a matrix, the counter matrix. The FACTORFINDER-Procedure develops a characteristic distribution of counter values assigned to each candidate variant. The intermediate result vector represents this specific counter value distribution within the counter matrix. That is why the counter matrix has to be initialized each time the procedure starts with a new candidate.
Very important factors make non-ideal values look little tolerable. Only small values will find confirmation as a tolerance hypothesis. Though, minor important factors will enable high values to be confirmed as a tolerance hypothesis and at the same time low values. Low values can be confirmed at any time because they are less critical. By repeated evaluation cycles, high and low values are tested to find out the highest value that represents tolerability of non-ideal values and that is confirmed by at least one pair of acceptable variants.
The intermediate result vectors indicate what variants can be considered being ideal. Later, from the intermediate result vectors the final relevance vector is calculated.
The FACTORFINDER-Procedure requires preparing steps as compiling the example base and setting assumptions on the behaviour of the procedure. The core of the procedure consists of two encapsulated loops: The outer loop takes one acceptable variant after the other and generates an intermediate result assigned to each acceptable variant. That intermediate result reflects properties of this example. To reach that intermediate result, the inner loop generates many tolerance test vectors.
As soon as the outer loop is finished, the intermediate results are processed. From the intermediate results, the ideal variants can be directly identified. By calculating the average of all intermediate results in each factor the final result is derived.
Consequently, the ideal values and a relevance hypothesis are known. With this information, the plausiblility of the success of all variants can be verified and at the same time the result of the analysis can be checked.