Unveiling Customer Engagement in Destination Marketing: A Multimodal Approach to Analyzing Official Promotional Videos

This study examines how features of official destination promotional videos (DPVs) shape customer engagement (CE) on social media. Using 5037 Instagram videos from 54 US national parks, we apply machine learning and computer vision to extract video content, visual features, multimodal cues (text, audio, thumbnails), and post metadata. LightGBM with SHAP analyses shows that thumbnail–video contrasts, color coherence, and emotional cues significantly influence CE. Simpler, familiar content drives passive engagement (likes), whereas emotionally resonant, high-aesthetic content fosters active engagement (comments). Beyond method, this study contributes a dual-pathway theoretical framework that integrates cognitive load, elaboration likelihood, and media richness theories to explain how multimodal features jointly shape CE. Practical implications highlight balancing content simplicity and depth, ensuring visual and emotional consistency, and optimizing posting strategies. Overall, this research advances destination marketing by providing a scalable and theory-informed framework for analyzing social media videos.