Cascading Behavior in Yelp Reviews

TL;DR

This paper analyzes information cascades in Yelp reviews across various cities, revealing social influence patterns, cascade characteristics, and predictive features that can inform business strategies and consumer behavior analysis.

Contribution

It provides the first comprehensive analysis of Yelp review cascades, including topology enumeration, cross-city comparison, and the importance of non-root review features.

Findings

Significant presence of cascades in Yelp reviews

Heavy-tailed distribution of cascade sizes

Initial reviews can predict cascade size accurately

Abstract

Social media has changed the landscape of marketing and consumer research as the adoption and promotion of businesses is becoming more and more dependent on how the customers are interacting and feeling about the business on platforms like Facebook, Twitter, Yelp etc. Social review websites like Yelp have become an important source of information about different businesses. Social influence on these online platforms can result in individuals adopting or promoting ideas and actions resulting in information cascades. Research on information cascades have been gaining popularity over the last few years but most of the research has been focused on platforms like Twitter and Facebook. Research on the adoption or promotion of product using cascades can help determine important latent patterns of social influence. In this work, we have analyzed the spread of information i.e. cascades in Yelp across different cities in Europe and North America. We have enumerated and analyzed different cascade topologies that occur in the Yelp social networks. Some of our significant findings include the presence of a significant number of cascades in Yelp reviews indicating the importance of social influence, heavy-tailed distribution of cascades and possibility to accurately predict the size of cascades on the basis of initial reviews. In addition, we have also found that the characteristics of the non-root nodes and the non-root reviews are much more important type of feature as compared to the properties of the root nodes and root reviews of the cascade. These findings can help social scientists to analyze customer behavior across different cities in a much more systematic way. Furthermore, it can also help the businesses in a city to figure out different consumer trends and hence improve their processes and offerings.