Cornell Tech grad student Xiao Ma led a research team to study image quality's impact on online sales.

Courtesy of Cornell University

Cornell Tech grad student Xiao Ma led a research team to study image quality's impact on online sales.

January 22, 2019

Cornell Research Shows Poor Image Quality Can Affect Online Sales

Print More

Think twice before using either a stock photo or a low-quality image when trying to sell your used items online because buyers are more likely to trust high-quality images, a new Cornell Tech study warns.
The study finds that buyers in online marketplaces are more likely to trust sellers that used high-quality photos in their listings. As a consequence, products with higher-quality images are also more likely to be sold.

One reason for this may be that these types of photos “give an accurate depiction of what the product looks like,” whereas stock photos may feel “impersonal” or “too good to be true,” according to the study.

Secondhand peer-to-peer marketplaces like eBay and Facebook Marketplace, where individuals can buy and sell used items from each other in a streamlined process, are increasing in popularity. Citing data from PricewaterhouseCoopers, lead researcher Xiao Ma said that these online marketplaces are estimated to be worth $355 billion by 2025.

“Understanding how images mediate interactions in online marketplaces could improve the efficiency of these marketplaces dramatically,” Ma told The Sun in an email.

Although there is existing research on the impact of images in online marketplaces, new technology has opened up new possibilities for studying these images, according to Ma.

“In recent years, the advancement of deep learning — especially in computer vision — presents an opportunity to revisit these topics, achieving better performance,” Ma said. Deep learning, a form of artificial intelligence, attempts to mimic brain function to process data, according to the MIT Technology Review.

Ma’s study focused on images of shoes and handbags on the online marketplaces LetGo.com and Ebay.com. Her team, including her Cornell Tech advisor Prof. Mor Naaman, information science, built an algorithm that classifies images based on quality, which achieved 87 percent accuracy.

This algorithm could help online marketplaces “pick out good quality images, and thus help ranking or thumbnail image choosing,” Ma said.

Ma believes this research could also help individual sellers take better images of their products. For example, the study concluded that images that were brighter and well-proportioned were more likely to be perceived as high-quality.

Ma said this type of research could have vast implications on the future of online marketplaces, such as future sellers using augmented reality applications that can provide “real-time feedback and guidance” of their images.

“Computer vision offers an opportunity to scale such process of improving image quality,” she said.