Grocery retailers need accurate sales forecasts at the Stock Keeping Unit (SKU) level to effectively manage their inventory. Previous studies have proposed forecasting methods which incorporate the effect of various marketing activities including prices and promotions. However, their methods have overlooked that the effects of the marketing activities on product sales may change over time. Therefore, these methods may be subject to the structural change problem and generate biased and less accurate forecasts. In this study, we propose more effective methods to forecast retailer product sales which take into account the problem of structural change. Based on data from a well-known US retailer, we show that our methods outperform conventional forecasting methods that ignore the possibility of such changes.