The Predictive Power of Social Media: Using Twitter to predict Spotify streams for newly released music albums

Kalampokis, Tambouris and Tarabanis (2013) categorized social media predictor variables in three categories: volume-related variables, sentiment-related variables and profile characteristics of online users. Previous research has shown the predictive power of Twitter when using a single type of predictor variable. However, few studies have focused on the combination of multiple types of social media predictor variables. Moreover, research on the predictive power of social media in music often solely focused on volume-related variables and focused on blogs and Myspace. Nonetheless, Twitter plays an increasingly important role in the music industry and in academic research. This study tests the predictive power of all three categories of social media predictor variables on Spotify streams of newly released music albums. Over 2.4 million tweets were collected over a period of five weeks using keywords related to the artist and the album title. Multiple regression analyses were performed in order to test the relationship between Twitter predictor variables and Spotify streams. Results show the importance of volume-related variables in predicting Spotify streams. The sentiment-related variables and profile characteristics of online users were found to have no significant predictive power on Spotify streams. The volume of tweets related to the album performed better at predicting Spotify streams than the volume of tweets related to the artist. A daily time series of the volume of tweets containing the album title was able to predict first week streams with high accuracy.