Determinants of consumer’s shopping behavior are of long-term interest to researchers. Since product photos directly aid consumers’ understanding of products, retailers often put a lot of effort into polishing them. However, there is limited research on the impact of product photos on shopping behavior. This research takes advantage of image-processing techniques to study product photos’ impact. These techniques allow us to investigate a large set of photo characteristics simultaneously in an empirical study. To rule out possible confounding factors, we use a real company dataset from a social shopping Website, which has a simple interface allowing consumers to judge products mainly based on their photos. We employ two-stage nested logit model embedded with differences-in-differences approach and examine product photo characteristics from the aspects of color, composition, complexity, and model face. We found that consumers prefer to click product photos with a warmer color, a larger key object, appropriate complexity.