Modelling and forecasting brand share: A dynamic demand system approach

This paper proposes a dynamic state-space AIDS model of brand share. The model structure separates the transitory and permanent components of the data and offers several key advantages over conventional share modelling approaches. Firstly, it provides a general framework to examine the time series properties of brand share. Secondly, it provides an accurate assessment of the short-run effects of marketing mechanics. Thirdly, the extracted permanent component can be used in a separate auxiliary analysis of the long-run effects of marketing mechanics. The model is applied to a small segment of the toiletries category comprising four differentiated brands and covering over five years of quad-weekly time series data. The results demonstrate a significant improvement over the conventional AIDS model and indicate short-run competition on the basis of price and promotional activity. Long-run analysis, however, suggests that such activity exerts no sustainable impact on the share of competing brands. Out of sample forecasts are satisfactory and provide an improvement over both a standard dynamic AIDS and time-varying-parameter MNL model. Brand share forecasts, in conjunction with price and segment value forecasts, are then used to produce brand volume forecasts.