Information Networks to Derive Value from Social Media

The rise in electronic interactions has made information networks ubiquitous. Correspondingly, research across multiple domains has begun to acknowledge the social and economic value of these networks for business decision-making. One such type of information networks is implicit brand networks that provide a direct digital window into the “interest space of millions of brand fans” on social media. These new types of digital artifacts effectively condense the preferences of millions of Twitter users into a reduced manageable format; thus, providing a novel methodological tool for managers to obtain close to real-time estimates of within-industry competition and across-industry complementarities. Unlike conventional social networks that involve direct interaction between the participating entities, links within a brand network are tacit and arise due to the co-followership activity of digital users. To statistically examine the tacit information contained in brand-brand links, the authors employ Exponential Random Graph Models from network theory and reveal a mix of network and brand level characteristics responsible for the observed network structure. Some of the significant effects include homophily based on category, cross-category interactions between certain pairs (Automotive-Sports, Travel-Restaurants, etc.) and frequency of the brand’s engagement with fans. While most existing studies focus on within-category competition, the notion of examining cross-category effects has largely been ignored. This work aims to bridge this gap by highlighting statistically relevant co-interest patterns between brands of the same as well as different categories. The paper concludes by discussing the practical applications of the information network, particularly for the automatic extraction of category-specific brand ratings. As information pertaining to category-specific ratings (e.g., sports, tech, luxury, etc.) is rarely found in online users’ comments, the brand network’s ability to automatically reveal such insights, with minimal a-priori assumptions, is a significant contribution of this study