Mining social media data via supervised topic model: Can social media posts inform customer satisfaction?

Abstract

Customer satisfaction is crucial for any firm. Traditional methods of measuring customer satisfaction, such as customer surveys, are resource-intensive despite their effectiveness. We develop an innovative approach that leverages social media posts to evaluate customer satisfaction. Specifically, we augment survey data with social media content and propose a supervised topic model to predict customer satisfaction. Method-wise, our model accommodates texts from various social media platforms, with or without explicit customer ratings. In addition, we address the challenges associated with integrating multiple data sources. To empirically validate our approach, we utilize data from various social media platforms combined with customer surveys from target firms in seven essential industries in Hong Kong. Our model exhibits higher prediction accuracy compared to baseline methods. This research provides a cost-effective and efficient tool for transforming vast amounts of social media posts into valuable insights on customer satisfaction.