One Picture Is Worth a Thousand Words? Empirical Evidence on the Informativeness of Product Images and Their Visual Appearance
Online presentations of products offer both visual and textual descriptions such that potential customers can identify products of interest. This paper contributes by comparing the informativeness of textual elaborations and product pictures. Our findings stem from a comprehensive dataset of more than 36,000 real estate listings from a website with classified advertisements in order to investigate the relationship of textual cues and visual appearance with prices. For this purpose, we train a deep neural network that can accurately infer the visual aesthetics of products from their images. The resulting image sentiment helps to explain a large portion of the variance in prices. A one standard deviation higher image sentiment links to a considerably higher price, and with a larger increase compared to narrative descriptions. Hence, images capture information beyond narrative explanations and encode valuable information for customers. As a direct implication, our work helps individuals and practitioners in the fields of electronic commerce, recommender systems, advertising and marketing to improve the presentation of their products and services.