A New Method to Understand the Online Word of Mouth Dynamics: Multi-View Sequential Canonical Covariance Analysis

Abstract

In today’s economy, nearly all products and services generate WOM communication on social media. However, at least three challenges hinder the analysis of online WOM. First, online WOM is usually unstructured data in various communication forms. However, the process of transforming unstructured data can generate a large number of variables, increasing the need for dimension reduction. Second, online WOM can be continuous or bursty. The volume and valence of online WOM may dramatically change in a short period before and after an incident. Third, important events might trigger symmetric or asymmetric reactions in online WOM across rival products or services. We introduce a new method—multi-view sequential canonical covariance analysis to solve these methodological challenges. This new method can solve the myriad WOM conversational dimensions, detect WOM dynamic trends, and examine its concurrent effects across multiple firms. It also provides greater computational efficiency and thus can be referred to as a more advanced manifold optimization approach. We illustrate the advantages of this new method through an empirical example—the 2017 United Express Flight 3411 incident. We find the shared WOM across all airlines significantly increased in April and May 2017. United Airlines and its rivals all experienced a sudden increase of negative emotions and a sudden decrease of positive emotions most likely because of the Incident, yet the magnitudes of the changes were more dramatic for United Airlines. This new method provides a novel insight into the online WOM dynamics and can contribute to a wide range of fields.