Social Media Effects on Sales: Consumer Sentiment in a State-Space Model

Recently, we have grown considerably in using social media to allow consumers to exchange and share opinions and information about products and services. Companies increasingly use electronic word of mouth (e-WOM) to evaluate and improve their products. This paper proposes a state-space model approach for sales forecasting using social media sentiment as an additional signal to improve predictions. We test the model on two case studies: Gentilini Osvego Biscuits and FIAT L500. We collect two corpora from Twitter: Osvego Biscuits (1016 tweets) from January 1, 2013, to December 31, 2020, and FIAT 500L (21,495 tweets) from August 1, 2012, to December 31, 2018. We analyze the monthly sales volume for Gentilini biscuits and FIAT L500 for the same period. The results show that our models seem to outperform the ETS-AAA benchmarks.