Personality-Based Content Engineering for Rich Digital Media

Abstract

Firms have increasingly turned to rich digital media, such as videos and photos, to attract attention and boost awareness. Although extant research may help firms promote these media more effectively, the marketing process truly begins with creation of the media. Content creators may thus benefit from understanding what content features (specifically, we focus on personalities) are likely to help their media achieve greater popularity. We develop an empirical approach to understanding the effect of content on the consumption of online videos, and apply this approach to a unique dataset of 16,414 videos from 363 YouTube channels. Our method labels videos as high- or low-performing relative to comparable videos, and employs natural language processing to characterize videos by the extent to which their captions reflect each of the “Big Five” personality traits. We then leverage non-linear, data-driven machine learning inductive techniques to identify whether, and which personalities improve video performance. Our analysis uncovers novel predictive, economic, and prescriptive insights. We find that using just their personality, we can predict with 72% accuracy whether videos will perform better than comparable media. Furthermore, videos associated with high-performing personalities can expect a 15% increase in cumulative consumption relative to those with low-performing personalities. Finally, we examine which personalities are associated with high consumption, confirmed by counterfactual analysis, offering prescriptive insights for content engineering.

Keywords: content engineering, personality, rich digital media, random forests, natural language processing