Earticle

Thumbnail image is one of the most common forms of video visualization used in online video platforms that let viewers to catch a glimpse of the content before watching the whole video. In recent years, leading online video platforms have adopted the feature of custom thumbnails with which video creators can actively generate or edit images to promote the content. Although prior literatures have revealed visual presentation to significantly influence human attention and purchase decisions, there has been little to no empirical study on whether and which visual features of a video thumbnail are effective in driving more views. In this paper, we conduct an exploratory study on the effect of thumbnail images on video views with a rich data set from the largest video-hosting platform, YouTube. Based on image features extracted from machine-learning models, thumbnail design factors associated with the video views are identified using fixed-effect models.