The impact of images on user clicks in product search

Abstract

Product search engine faces unique challenges that differ from web page search. The goal of a product search engine is to rank relevant items that the user may be interested in purchasing. Clicks provide a strong signal of a user's interest in an item. Traditional click prediction models include many features such as document text, price, and user information. In this paper, we propose adding information extracted from the thumbnail image of the item as additional features for click prediction. Specifically, we use two types of image features -photographic features and object features. Our experiments reveal that both types of features can be highly useful in click prediction. We measure our performance in both prediction accuracy and NDCG. Overall, our experiments show that augmenting with image features to a standard model in click prediction provides significant improvement in precision and recall and boosts NDCG.

Understanding Image Quality and Trust in Peer-to-Peer Marketplaces

“…Image Quality in Online Marketplaces. Previous work has shown that image-based features such as brightness, contrast, and background lightness contribute to the prediction of click-through rate (CTR) in the context of product search [5,11]. This line of work used hand-crafted image features, but did not actually assess the image quality as dependent variable.…”

“…In this work, we leverage our annotated dataset, and conducted the first computational analysis of the impact of common product photography tips on image quality. Unlike previous work [5,11] that analyzed the impact of image features on clicks, we evaluate directly on potential buyers' perception of image quality. In a later section, we then show how image quality can in turn predict sales outcomes.…”

“…These findings suggest that both image quality and aesthetic quality are associated with higher likelihood of sales for individual listings on online marketplaces, though both have limited power in improving the accuracy of sales prediction. Future work could explore the difference among different product categories, or the relationship between image quality and other metrics for online marketplaces, such as "sellability" [15], and click-through-rate [5,11].…”

Is a picture really worth a thousand words?

“…Recent study showed that certain image features can also help to improve click through rate in product search engine [11,7]. The authors conducted experiments to show that including image features in a machine learned click based ranking model improves the NDCG (normalized discounted cumulative gain) of the search results.…”

“…Because some of the these features are computed from RGB values (0-255), as a convention, the range of quality score is also kept in [0, 255], whereas a score of 255 means the best quality. Potential readers are referred to [7,11] for details.…”