Influencer videos on the TikTok online platform have emerged as a major, multibilliondollar force in marketing. We explore what differentiates influencer videos that drive many sales from those that drive only a few. We develop an algorithm that can be used to predict the sales lift of TikTok influencer videos. This algorithm uses a Convolutional Neural Network to quantify the extent to which the product is advertised in the most engaging parts of the video, creating what we call a motion score, or m-score for short. Videos with higher m-scores lift more sales, especially for products that are bought on impulse, are hedonic, or inexpensive. A video that only engages, or that simply features the product throughout, will not necessarily have a high m-score. It needs the product to be featured at the right place and the right time. Stakeholders in the TikTok ecosystem can use m-scores to aid video design, to select videos for various campaigns, and to align incentives for influencers and brands.
Influencer Video Advertising in TikTok
July 02, 2021