Using Consumer-Generated Social Media Posts to Improve Forecasts of Television Premiere Viewership: Extending Diffusion of Innovation Theory

Abstract

This research investigated how social media can be quantified and used as an input to improve audience forecasts for television show premieres. There were two key findings. First, Twitter activity (volume of tweets and retweets) drives viewership of shows that are unscripted (i.e., reality or competition shows). From a practical perspective, Twitter activity improved prediction accuracy beyond that of forecasting inputs typically employed by the industry. The second key finding was the interaction of media attention given to a show and audience size of the show leading into (i.e., preceding) the premiere. One practical implication for network scheduling and promotion efforts is to use strong lead-in shows along with large mass media efforts for a television season’s “flagship” series that are most important for a network’s success. This research also extends Diffusion of Innovation theory (Rogers, 2003) and diffusion modeling to television entertainment consumption. Diffusion of Innovation theory predicts the importance of information dispersion across heterophilous groups (i.e., groups that are diverse). The Bass (1969) diffusion model, a statistical model introduced to describe the diffusion of innovation, also predicts that media presence (an external factor) in conjunction with consumer eWOM (an internal factor) drive premiere ratings.