Studying sales promotions represents an extremely important research topic for both researchers and business professionals for the following reasons : – the amount of money invested in this field of study every year, – the emergence of new sources of data which add to or replace those which already exist. In France today, existing data comes from four sources : – consumer panels, – retailer panels, – test areas, where sales are recorded at outlets in a given area, thereby providing information about sales and the characteristics of a sample of consumers, – ad hoc experimentation. Further information from check-out counters at the time of purchase is also available to retailers. Collecting, managing and analysing the huge amounts of data available has become a big challenge for statisticians. Traditional statistical methods used to analyse the impact of promotions are : – time series analysis (using weekly sales data from outlets to study the effect of promotions on sales), – the linear regression model (which studies the effect on sales of different types of promotions and the activities of competing brands), – the logit model (brand choice models). In this paper, we set out to demonstrate the limitations of these traditional models and to put forward solutions to improve the models.
DUSSAIX, A.M. et INDJEHAGOPIAN, J.P. (1996). Etude de l’impact des promotions sur les ventes : méthodes de collecte des données et méthodes d’analyse statistique. ESSEC Business School.