Abstract: When purchasing products in the live streaming room, consumers can not only gain a more intuitive understanding of product characteristics and performance through real-time video, but also consider the evaluation and attitude of the reference group as a crucial factor for decision-making. In the context of live streaming, the real-time reference group primarily influences users' purchase behavior by digitally displaying danmaku comments. Existing studies have predominantly examined reference groups within the context of reality, network communities, and online shopping platforms. However, there is still a need to explore how real-time reference groups during e-commerce live streaming influence consumers' purchasing behavior through danmaku. This study uses Python programming to extract 1,064,613 danmaku text data from 22 live broadcasts of the "Make a Friend" room on the TikTok platform, and collects relevant product data from the "Huitun" platform to explore the influence of real-time reference group on consumers' purchasing behaviors, while considering the moderating role of product types. The results show that the diversity, activeness, and sentiment of danmaku have an inverted U-shaped relationship with consumers' purchasing behavior. In terms of informational influence, search products strengthen this inverted U-shaped relationship compared to pseudo-experience products and experience products. In terms of normative influence, experience products strengthen this inverted U-shaped relationship compared to search products and pseudo-experience products. This study enhances the application scenarios of reference groups, extends the boundaries of research on e-commerce live streaming, and offers practical guidance for operating e-commerce live streaming.

Key words: e-commerce live streaming, real-time reference group, danmaku characteristics, product type