We show how using one expert system allows us to go further and quicker in the analysis of the enormous amount of data collected by scanner in a sample of stores. In particular the first step of our research is to be able to condense, to structure this data in a automatically generated report, underlining the significant elements.
The use of retail panels for measuring, the impact of a new brand on the market, is an invaluable and indispensable method. A new brand must, if successful, make a measurable difference to the existing market, though how far test market results can be used to forecast national sales is a matter for some conjecture. However, test markets are used not only for assessing the performance of new brands but are frequently used in an endeavour to measure the effect, on established brands, of advertising promotions, couponing and sample operations, advertising weight tests, media tests and so forth. I hope to demonstrate, in this paper, that natural and apparently fortuitous variations occur in over-counter sales, which, in some cases, far outweigh those to be expected from promotional activities.