LAUSR

Understanding what factors make a helpful online review is critical to increase sales and drive revenue for online retailers. This paper examined the impacts of both reviews’ numerical and textual characteristics on review helpfulness across three different review types including comparative, suggestive, and regular reviews. With an analysis of 30 338 product reviews collected from http://Amazon.com, the results indicated that the effects of numerical characteristics of reviews on review helpfulness are stronger for regular reviews than those for suggestive and comparative reviews. The impacts of text sentiment on review helpfulness are more significant for suggestive and comparative reviews when compared with regular reviews. Moreover, the text complexity of reviews has a significant invert U-shaped relationship with review helpfulness, and the relationships are stronger for regular reviews when compared with suggestive and comparative reviews. Furthermore, text sentiment has a negative effect on review helpfulness, and the effect is stronger for suggestive reviews than that for comparative and regular reviews. Finally, we employ a random forest method to predict review helpfulness based on its numerical and textual characteristics. This paper found that review length is the most helpful factor in predicting the helpfulness of online reviews. Our findings also indicated that the importance of numerical characteristics is greater than that of textual characteristics across three different review types. The theoretical and practical implications of the findings are presented.