The increase of the quality of advertising offers an important possiblity to improve the success of a brand without costs being automatically increased. Pretest measures for commercials as e.g. the AD*VANTAGE test system are thus intensively used by the manufacturers. With its quantitative and qualitative analysis potential, these tests can supply precious information for the assessment of TV commercials. Nevertheless, in some cases it was not possible until now to provide definite recommendations. For example, if commercials of different length are compared, the longer one very often achieves better quality scores, but is more expensive when being aired. The question whether the better quality is worth the additional costs cannot be answered from the test alone. To answer this kind of question a model is needed which connects pretesting data and market share development. This is exactly what the AD*VANTAGE model does which was developed for frequently purchased consumer goods. The model is based on a non-stationary markoff process of individual purchase data. The purchase probabilities for the own product and for the competitive products are influenced by distribution, price (own product and competition), advertising (own product and competition, commercial quality and advertising impact) as well as by promotions. The model requires only data which are normally available in big companies. The structure of the model is explained, its limitations discussed. The validations available up to now show good results. Two case studies demonstrate the use of the model for prognosis and simulation.