|
Abstract:
|
Promotion analysis is an important part of effect retail management. Because of the many uncertain factors such as discount range and price, weather variations, holidays, and demographics, all of which prove to change considerably, promotion effects on sales can be very complex. The focus of this paper is to develop accurate sales forecasting models to aid in informed promotion analysis. For sales forecasting, we establish a semiparametric varying coefficient model to account for the data varying over time and over space. In the forecasting model, we estimate confidence bands for the predictions that can be further integrated in the promotion analysis. The accuracy of the proposed forecasting model is investigated through a simulation study. Also, real-life data from a national-wide retail company demonstrate the importance of the model and illustrate the potentials of the model.
|
