Background, Composition, and Focus: Machine-based Assessment to Understand Product Photo’s Impact on User Interest

Product photos are a very important part of product presentations in online shopping that directly aid consumers’ understanding of products. Previous studies have shown photo aesthetics matter to consumers. With the advances of image-processing and machine learning techniques, we are able to derive machine-based measures from photos and systematically examine their impact on e-commerce from multiple dimensions. In this study, we use a machine learning method to differentiate the foreground (i.e., product) and the background of a product photo and propose a background-composition-focus framework for understanding consumers’ perception of the photo. We conduct an empirical study using a real-world clothing dataset collected from one of the biggest fashion product websites in China. Using a hierarchical Bayesian model, we find that consumers prefer clothing products being shown on a simpler background, located vertically close to the center and horizontally following the rule-of-thirds. And, it is better to have a blurred background and no model face on the photo so that consumers can focus more on the product itself. These findings are of strong theoretical and practical value.