Thumbnail power: Visual cues and video popularity on UGV platforms

User-generated video (UGV) has emerged as a powerful marketing channel capable of enhancing consumer engagement, fostering trust, and driving sales. This study integrates machine learning and deep learning techniques to extract visual attributes—specifically, complexity and coherence—from a dataset of 22,958 UGV thumbnails. Employing panel data econometric models and two randomized experiments, this study examines how these visual attributes influence video popularity, with a focus on the moderating roles of video type, typicality, and consumer involvement. Results reveal an inverted U-shaped relationship between visual complexity and video popularity, and a U-shaped relationship for visual coherence. These effects are more pronounced for utilitarian videos and those perceived as highly typical. Experimental findings further show that individual-level characteristics—namely, involvement, mindfulness, and thinking style—moderate these relationships. Collectively, the findings advance understanding of visual processing in digital environments and offer actionable guidance for optimizing UGV thumbnail design to enhance consumer engagement.