The FACTORFINDER-Approach requires to divide all known examples into those with acceptable success and those that fail. For each variant we can tell whether it is acceptable in respect to the success criterion.
That reduces the effort to measure success of the variants. Even partial data enable the analyst to attribute the success of an alternative to certain factors.
The factors that determine the success of alternatives could be mapped to a causal model of all relevant factors. The fact that such a model shows different states of its components under different assumptions is lost when putting causal chains into a linear order. This renders causal models inadequate for certain types of analysis and promotes the modelling by means of an example base. The structured combination of attributes forming an example represents a multitude of actors and actions that condense to a snapshot of the current state of all factors. This type of representation makes it superfluous to search for causal chains, to differenciate between causes and consequences and to search for indirecty acting factors. Actors and causes can be left open.
The only requirement for each alternative is to assign values to each factor and to classify each example as being acceptable or inacceptable. This bi-valued differentiation into success and failure makes an explicit model of the generation of each example - e.g. using a polit-economical or psychological model - unneccessary. This endeaver would be without success anyway if we consider early stages of analysis mostly due to the multitude of acting factors and constraints.
Each example has to be classified as being acceptable or inacceptable. In some cases, an analyst may hesitate when deciding. This can indicate a missing link in the applied model of the object analysed or can indicate a missing factor in the model. By making use of these signals, the nowon deepened analysis may lead to the discovery of formerly tacid knowledge.