Compounding trials and following property testing of materials are time consuming and expensive. One method to save time and money is to reduce the number of trials and to search for mathematical laws binding the effect of one (or several) additive and one (or several) property.
Then it is supposed that a change in concentration of additive A (concA) leads to a new value of the property P obeying this law.
As to the modelling, it is a purely mathematical exercise, which doesn't predict the physical and chemical evolutions as soon as the experimental limits are left behind. So the user takes a risk, the more important as the real conditions get far away the experimental context needed for the modelling basis.
For example, as we can see on figure 1, if carbon black is used for conductive compounds:
* The resistivity decreases abruptly when a threshold of the amount of the well-dispersed conductive carbon black is added
* Then for a sufficient level, resistivity reaches a minimum and an overdose of carbon black is inefficient.