Differentiation in Online Product Reviews: A Machine Learning Based Analysis
Online review of products and services has become a prevalent information source for consumers. This paper examines a key aspect of reviews and reviewer behavior: whether and how the content of a review systematically differs from reviews posted earlier. Content differentiation is particularly important when more reviews are posted as newer star ratings tend to converge providing limited room for the newer reviews to stand out. The authors apply machine learning and deep learning to classify restaurant reviews on Yelp.com. Employing first-difference models that account for dynamic panel bias, the analysis provides strong evidence for review differentiation: when previous reviews write more about the food (or non-food) dimension, a later review tends to write less about it. Review differentiation is greater as more reviews are published, when the star rating associated with the review deviates more from previous star ratings, and for regular (versus established) reviewers. The authors also show that review differentiation helps enhance the impact of a review. These findings suggest two important but distinct motivations for review differentiation: to enable a review (and its reviewer) to stand out from the crowd and to provide support for star ratings. Implications for reviews and review platforms are discussed.