Experiments (in the narrow sense) and surveys, both in the sequel called experiments (in the wider sense), are the most important tools in empirical research. Because research resources must be used as efficient as possible, the design and the analysis of experiments should be done in an optimal way. Experimentation should start with the precise formulation of the experimental question, a procedure which can be repeated after the following steps. One of them is the choice of an appropriate (statistical) model, another the definition of the precision requirements and of the optimality criterion. Then the experiment can be planned in an optimal way, performed, and optimally analysed. In the paper this procedure will be described for some practical situations using linear and non-linear statistical models.