Human-Centered Multimodal Insights for Short Video Social Media Marketing
(A2025-126419)

AUTHORS

Wanqing Zhang, Bayes Business School, City, University of London; Qinghua Li, northwestern university; Pradeep K. Chintagunta, Booth School of Business, University of Chicago; Peng Wang, Bayes School of Business; Liqiang Huang, Zhejiang University

ABSTRACT

Short video social media marketing has become a pivotal tool in today's digital landscape. In this study, we analyze a comprehensive dataset of short video advertisements to uncover the correlations between the success of short video marketing and diverse video features such as visual elements, human presence, gender, humor, and audience engagement, Utilizing data collected from DouYin (Chinese TikTok), we applied facial recognition, voice recognition, and pretrained transformer models to extract key features, including the human presence ratio, gender, emotion, and humor index. Our findings reveal that these features significantly correlated with sales of advertised products, with category-specific variations. This research highlights the importance of leveraging multimodal analytics to design effective, human-centered marketing strategies tailored to different product categories.