Using Natural Language Processing to Identify Effective Influencers

Identifying the right influencers for brands is often the starting point for a successful influencer campaign. However, influencer identification is understudied, and most previous studies have only discussed visible characteristics of influencers and their social networks, overlooking content-based metrics. Combining interdisciplinary theories and techniques from marketing, linguistics, and computer science, we propose a data-driven automated text analysis framework to identify characteristics of effective influencers using influencer posts. Specifically, we propose a model that incorporates influencer personality traits captured by natural language processing, accounting for traditional covariates, such as network structure and follower engagement. In addition, we use a dataset that attributes influencer social media activities to customer purchases to address fake engagement and showcase our automated textual analysis. The proposed framework can help marketers develop influencer profiles and predict optimal influencers for their campaigns.