A causal forecasting model for the sales of a consumer product

This paper presents a forecasting model for bimester trade sales of Johnson & Johnson, Lda. (J&J) product in Portugal. The trade sales of this market leader show great variability, as a result of seasonality, promotional activity and retailer's stock policies. It was necessary to understand the logistic chain, from when the product leaves J&J until it reaches final consumer, and to model final consumer sales before trade sales could be modelled. A causal model was adopted, including three groups of explanatory variables – marketing, seasonality and trend, and a variable that represents the product's trade stock level. This variable has a different effect in the two models; if trade stock is high, sales to the final consumer increase reflecting retail promotional activities, but trade sales decrease as retailers have full store rooms. Models to forecast sales by size, as a function of total product sales, are also presented. A discussion of the effect of promotional sales on the total product sales is also included.