Promotional forecasting in the grocery retail business

Abstract

Predicting customer demand in the highly competitive grocery retail business has become extremely difficult, especially for promotional items. The difficulty in promotional forecasting has resulted from numerous internal and external factors that affect the demand patterns. It has also resulted from multiple levels of hierarchy that involve different groups in the organization as well as different methods and systems. Moreover, judgments from the forecasters are critical to the accuracy of the forecasts, while the value of tweaking the forecast results is yet to be determined. In this business, the forecasters generally have a high incentive to over-forecast in order to meet the corporate goal of maximizing customer satisfaction. The main objective of this thesis is to analyze the effectiveness of promotional forecasting, identify the factors contributing to forecast accuracy, and propose suggestions for improving forecasts. In light of this objective, we used WMPE and WMAPE as the measures of forecast accuracy, and conducted analysis of promotional forecast accuracy from different point of views.

(cont.) We also verified our results with regression analysis, which helped identify the significance of each forecasting attribute so as to support the promotion planning without compromising forecast accuracy. We suggest several approaches to improve forecast accuracy. First, to improve store forecasts, we recommend three models: the bias correction model, the adaptive bias correction model, and the regression model. Second, to improve replenishment forecasts, we propose a new model that combines the top-down and bottom-up approaches. Lastly, we suggest a framework for measuring accuracy that emphasizes the importance of comparing the accuracy of forecasts generated from systems and from judgments.