To whom should I listen? Finding reputable reviewers in opinion-sharing communities

Abstract

Online opinion-sharing communities, which allow members to express personal opinions and preferences about specific products, provide important channels for consumers to learn about product quality and support their purchase decision process. Firms can use these reviews to understand customers' responses to their products and improve their products accordingly. Furthermore, opinion-sharing communities provide an alternative, effective marketing channel to firms by offering electronic word of mouth (eWOM). However, due to the openness and anonymity of opinion-sharing communities, their members face a challenging issue, that is, whether to believe or disbelieve information provided by other members. This study attempts to discriminate members (i.e., reviewers) with a high reputation from those with a low reputation on the basis of members' web trust network and review behaviors in an opinion-sharing community. We collected sample data pertaining to four product categories from Epinions.com to test our research model. The results indicate that four variables (trust intensity, average trust intensity of trustors, degree of review focus in the target category, and average product rating in the target category) successfully discriminate reviewers into the two groups, and product type is a significant control variable. These findings not only help firms identify reputable reviewers for marketing campaign purposes but also enable the members of an opinion-sharing community to determine who reputable reviewers are and whose reviews they should trust.

Introduction

With the rapid proliferation of e-commerce and Web 2.0 innovations, the Web has become an excellent platform for gathering and sharing consumers’ personal views on, preferences for, and experiences with products [8], [11], [13], [16], [44]. Many third-party product review websites (e.g., Epinions.com,¹ rateitall.com, Cnet.com, Edutainingkids.com, ZDNET.com) have established opinion-sharing communities to facilitate the exchange of consumer reviews about a variety of products. It also has become a common practice for retailers (e.g., amazon.com, target.com, bestbuy.com, tripadvisor.com, Shopping.com, Walmart.com) and product manufacturers (e.g., Hewlett-Packard, Nike, Levi's) to create their own opinion-sharing communities, where customers can express their opinions about products they have purchased or in which they are interested.

Consumer reviews are essential to firms (both retailers and product manufacturers) in their efforts to understand the general responses of customers to their products and improve their products accordingly [44]. In addition, opinion-sharing communities provide an alternative, more effective marketing channel to firms, in the form of electronic word of mouth (eWOM), that does not require huge advertising investments [18], [42]. Accordingly, prior studies indicate that consumer reviews can increase sales [5], [6], [13], [15], [27], and the presence of customer reviews on a company's website can improve customer perceptions of the website's usefulness and social presence [26], attract customer visits, increase the time spent on the site (i.e., stickiness), and create a sense of community among frequent shoppers [32].

From a customer perspective, consumer reviewers offer a key information source about product quality [6], [10], [19] and thus can facilitate purchase decision processes [32]. However, as the number of reviews in an opinion-sharing community increases, it becomes more difficult for a member to search or browse through the many reviews contributed by other members to obtain sufficient and useful information. The search process can be not only time consuming but also frustrating, because the quality of consumer reviews varies greatly. Some reviewers appear professional and devoted to the task, but others are subjective and biased. An even more serious concern arises from the openness and pseudo-anonymity of opinion-sharing communities, which confronts members with a challenging issue: should they believe or disbelieve the information provided by other members in the opinion-sharing communities [49]? Consumer reviews can be manipulated easily by interested parties [9]. For example, firms might anonymously post positive reviews of their own products to increase customers’ awareness and perceptions of those products.

One solution to this challenge is a reputation mechanism that reveals the reputation of reviewers who participate in an opinion-sharing community [8], [39]. With such a mechanism, members in the opinion-sharing community can attend to the reviews contributed by more reputable reviewers and avoid those written by less reputable reviewers. The firms also can promote new products to reputable reviewers, with the hope they will like the products and write favorable reviews. Many third-party product review websites (e.g., epinions.com, rateitall.com, ZDNET.com) and opinion-sharing communities hosted by retailers or manufacturers (e.g., Amazon.com, Walmart.com, Nike, Levi's) attempt a peer-evaluation-based approach (i.e., explicit reputation system) that enables members to evaluate and rate the reviews of other members. The site then can aggregate these evaluations (e.g., taking their average) as a measure of the reputation of a reviewer. While this practice appears feasible and involves straightforward computations for reputation estimation, whether to calculate reputation scores for members in the community or classify members into different reputation categories, this user-driven approach also suffers some limitations.

First, for comprehensive coverage, this approach requires at least one user evaluation on some reviews of each reviewer. If none of the reviews written by a reviewer ever gets evaluated by other members, his or her reputation cannot be derived. However, members may not bother to provide feedback (i.e., user evaluation) at all, thus creating a sparsity problem [39]. In addition, similar to other online behaviors, user evaluations tend to be skewed. That is, contributions by some reviewers receive hundreds of user evaluations, whereas others attract no user evaluations. Such sparsity and skewness can result in low coverage and significantly limit the applicability of the user-driven approach to reputation estimation. Second, the user-driven approach is vulnerable to malicious and inflated evaluation behaviors [12], [39]. For example, a member may intentionally give unfavorable evaluations to some members, regardless of the quality of their reviews. Or a member could attempt to inflate his or her own reputation dishonorably, for instance, by creating “fake” members and hiding behind these pseudonyms to create more positive user evaluations to his or her reviews. In both cases, the user-driven approach is ineffective for accurate reputation estimation of the members of an opinion-sharing community.

Motivated by the importance of reputation estimation and the limitations of the user-driven approach, this study develops a reputation estimation model to measure the reputation of a member in an opinion-sharing community. Specifically, our proposed reputation estimation model is based on factors derived from members’ web trust network and review behaviors in an opinion-sharing community. Because we do not rely on information pertaining to user evaluations, our reputation estimation model will not suffer from the limitations inherent to the user-driven approach (i.e., low coverage and vulnerability to malicious and inflated evaluation behaviors). To present this solution, we organize the remainder of our paper as follows: Section 2 reviews literature relevant to this study. We propose our research hypotheses in Section 3. In Section 4, we describe the research methodology and discuss important analysis results. Finally, we conclude with a summary and discussion of some future research directions in Section 5.