Predicting livestream shopping success
利用淘宝直播数据，构建包含直播策略、产品组合、内容与互动指标的评分模型，将销售预测准确率提升25.6%，帮助识别有潜力的卖家。
Despite the popularity of the livestream shopping industry—projected to surpass $830 billion in China and $68 billion in the U.S. by 2026— research on identifying the factors that drive success in this industry remains scarce, primarily because of the complexity of this environment. In this study, we use data from Taobao Live, the largest livestream shopping platform, to predict the success of sellers hosting livestream sessions. We develop a scoring model using a comprehensive set of livestream strategies and metrics, including session strategies, product assortment strategies, livestream content metrics, and engagement metrics. We find that incorporating session and assortment strategies and livestream content and engagement metrics improves the accuracy of sales performance predictions by 25.6%. Additionally, we identify the key strategies and metrics that are most predictive of sales performance. The proposed predictive framework not only helps in identifying promising sellers but also enhances the understanding of the complex dynamics of the livestream shopping ecosystem, providing valuable insights for stakeholders such as investors, lenders, suppliers, and platform operators.