546 / 2024-04-25 20:48:17

Video Ranking Based on Reference Effect

consumer search,reference effect,product ranking,short-form video platform

By analyzing a public dataset from NetEase Cloud Music, we empirically verify that within short-form video platforms, where viewers can swipe down to see the next video, there exists an unconscious tendency for viewers to compare new videos with the best ones they have previously watched. Based on such empirical evidence, we study the sequential product ranking problem considering this peak reference effect. Traditional video recommendation literature typically ranks videos based on their popularity. However, our study takes into account the peak reference effect and finds that platforms should provide mediocre videos first and then gradually provide better videos to manage viewers' expectation. We consider both pre-set ranking scenario and responsive ranking scenario. Although the dynamics in the model make the problem challenging, we present structural results and design tractable algorithms with guaranteed performance for both scenarios.