Considering economic indicators and dynamic channel interactions to conduct sales forecasting for retail sectors

Retail sectors consisting of hypermarkets, supermarkets, and convenience stores are closely related to the economy condition of a country because they satisfy basic requirements in Maslow's hierarchy of needs. These sectors sell substitutive product categories and mutually compete in customer groups. For instance, hypermarkets and supermarkets sell fresh vegetables, meat, fishes, and household supplies. Supermarkets and convenience stores sell drinks, snacks, bread, etc. To reveal managerial insights, this study presents a novel framework to achieve the following goals: (1) influential economic indicators and associated time lags are identified as effective predictors, (2) machine learning and deep learning are adopted to conduct sales forecasting, and (3) dynamic interactions between retail sectors and between competing firms are incorporated to estimate sales equilibriums. Experimental results show that consumer price index, retail employment population, and real wage concurrently affect sales revenues of the three retail sectors while seasonal factors are only critical to hypermarkets. Deep learning performs better and demonstrates good generalization without overfitting. Supermarkets and convenience stores suffer from the existence of hypermarkets and supermarkets are expected to decline 16% sales at equilibriums.