315 / 2024-04-24 12:35:21

Demand Forecasting during Grand Promotion for Online Retailing

Demand forecasting;,Grand promotion,Bayesian,LASSO

Grand Promotion, a notable promotion strategy adopted by online retailers, can boost product sales significantly while also presenting unique challenges to operations management. Online retailers must rely on accurate demand forecasts to effectively prepare for procurement, inventory, and logistics. We propose a wavelet-based forecasting framework to predict demand during grand promotions. Our framework effectively utilizes the discrete wavelet transformation to depict the intense fluctuations of daily sales curves. Exploiting the sparsity of the wavelet coefficients and feature sets, we design a Bayesian LASSO method to handle the high dimensionality of parameters. We examine the error bound of this framework to provide a theoretical guarantee and evaluate its prediction performance using JD.com's real data. Compared to JD.com's existing method, our framework can decrease the prediction error by about 4$\sim$11\% among different metrics. Furthermore, our method reveals valuable insights about grand promotions, which can effectively illustrate consumer behavior patterns and provide crucial guidance for the operational management of online retailers.