Social Advertising Effectiveness Using a Large-Scale Randomized Field Experiment

Study 1: Social Advertising Effectiveness

Study 2: Social Advertising Effectiveness Across Products,  forthcoming in Marketing Science

Study 3: Identifying Subgroups with Enhanced Peer Influence Using High-dimensional Data

We examine social advertising effectiveness and its heterogeneous effects across products, individuals, and social ties, by identifying the causal relationships among social influence, products, and network-embedded human behaviors. Social advertising places social cues (e.g., likes) in ads, utilizing the power of social influence (the effects of social cues in ads) to encourage ad engagement. We collaborate with a world-known social networking app for a large-scale randomized field experiment on its social ads. In the experiment, the presence and the number of social cues were randomly assigned among 57 million ad-user pairs (more than 37 million subjects and across 71 products in 25 product categories). Integrating the experimental evidence and the data of individuals, products, ads and network structures, our studies also address the incentives, magnitude, contagion patterns and viral factors (i.e., characteristics of products, behaviors, and individuals) of social influence in social advertising and product adoptions.

How Emotions Impact Online Content Spread in a Massive Social Network

To investigate the impact of emotions on the spread of online content, we analyzed a random sample of 387,486 online articles and their diffusion cascades, in which more than 6 million unique individuals shared the articles on one of the world's largest social networking site. We detected the degree of eight discrete emotions (i.e. surprise, joy, expect, love, anxiety, sadness, anger, and disgust) embedded in the content of each article with a newly generated domain-specific and up-to-date emotion lexicon. Our results suggest that articles with a higher degree of emotion (sum of discrete emotions embedded in the content) reached a significantly larger number of people and diffused significantly deeper, more broadly and virally but more slowly. Anxiety and love significantly increased cascading size, depth, breadth and structural virality, while sadness significantly decreased them. We found that the articles with a greater degree of anxiety and love were spread more among older and more central users (with more friends indicating higher social status), who associated with larger and faster cascades. On the opposite, the articles with greater sadness were transmitted more among younger users. The articles with greater love were diffused more through weak ties, while those with more sadness were transmitted more among strong ties. We also observe that the articles with a larger percentage of the weak ties in its diffusion cascade spread more slowly. The clusterness of the seed users did not impact any cascading dimensions, which indicates that social reinforcement is not at work in the diffusion of online articles. Our work has important practical implications for content producers, marketers, and campaign managers to promote articles, products, and elections.