Visual Uniqueness in Peer-to-Peer Marketplaces: Machine Learning Model Development, Validation, and Application

Peer-to-peer (P2P) marketplaces have seen exponential growth in recent years featured by unique offerings from individual providers. Despite the perceived value of uniqueness, scalable quantification of visual uniqueness in P2P platforms like Airbnb has been largely unexplored. Addressing this gap, our research develops, validates, and implements an unsupervised machine learning model, utilizing a dataset of over 14,121 Airbnb properties in New York City spanning 13 months. First, we develop a machine learning model based on contrastive loss to assess visual uniqueness in over 481,747 property images, demonstrating both high accuracy and interpretability. Next, we validate our model through three distinct experiments involving various populations and methods, confirming that our model's visual uniqueness scores and heatmaps align closely with human judgment. Finally, we apply this model to longitudinal Airbnb demand data, revealing an inverse U-shaped relationship between visual uniqueness and demand and two significant moderation effects: properties with superior host reputations or higher quality gain more from visual uniqueness. This research not only addresses methodological and conceptual voids in understanding visual uniqueness but also provides valuable managerial insights for sharing economy platforms like Airbnb and their hosts, emphasizing the strategic use of visual uniqueness to enhance visual appeal and market performance.