What makes information in online consumer reviews diagnostic over time? The role of review relevancy, factuality, currency, source credibility and ranking score

Highlights

• We study the antecedents of information diagnosticity in consumer reviews over time.

• Long reviews are not perceived as diagnostic information.

• Relevant reviews and overall ranking are perceived as diagnostic over time.

• The influence of review factuality and source credibility has evolved over time.

• Central and peripheral cues influence information diagnosticity in high-involvement.

Abstract

Online consumer reviews (OCRs) have become one of the most helpful and influential information in consumers purchase decisions. However, the proliferation of OCRs has made it difficult for consumers to orientate themselves with the wealth of reviews available. Therefore, it is paramount for online organizations to understand the determinants of perceived information diagnosticity in OCRs. In this study, we investigate consumer perceptions and we adopt the Elaboration Likelihood Model to analyze the influence of central (long, relevant, current, and factual OCRs) and peripheral cues (source credibility, overall ranking scores) on perceived information diagnosticity (PID). We consider the potential moderating effect of consumer involvement, and tested the robustness of the theoretical framework across time.

Based on two surveys carried out in 2011 and in 2016, this study demonstrates the dynamic nature of the antecedents of PID in e-WOM. We found that long reviews are not perceived as helpful, while relevant and current reviews as well as overall ranking scores are perceived as diagnostic information in both samples. The significance of the predicting power of review factuality and source credibility has evolved over time. Both central (review quality dimensions) and peripheral cues (ranking score) were found to influence PID in high-involvement decisions.

Introduction

Various online platforms (e.g. social commerce, ecommerce, online communities) are increasingly facilitating consumers in sharing their experiences, opinions, and feedback regarding people, products, services and brands in the form of online reviews, ratings, and ranking scores. Online consumer reviews, a form of electronic word of mouth (e-WOM), can be defined as any positive, neutral, or negative evaluation of a product, a service, a person, or a brand presumably posted by former customers on websites that host consumer reviews. According to a report from Mintel (2015) 81 percent of consumers aged 18–34 in the United States seek out opinions from others before purchasing a product or service. Research has established the power of online consumer reviews in predicting product sales and revenues in different product categories, such as books, beers, restaurants, movies, and hotels (e.g. Chevalier and Mayzlin, 2006, Clemons et al., 2006, Cui et al., 2012, Duan et al., 2005, Liu, 2006, Ye et al., 2011).

Different organizations are increasingly enabling consumers to leave a helpful vote to each review in an attempt to signal to consumers the most helpful reviews for assessing products and services’ quality and performance. Scholars have recently started to examine what makes online review helpful by importing the data from these e-retailers (e.g. Amazon) and using the voting mechanism to measure the characteristics of the reviews that receive more helpful votes (e.g. Ahmad and Laroche, 2015, Baek et al., 2012, Chua and Banerjee, 2016, Huang et al., 2015, Jabr and Zheng, 2014, Mudambi and Schuff, 2010, Pan and Zhang, 2011, Racherla and Friske, 2012, Yin et al., 2014).

However, there is a dearth of studies on the determinants of information diagnosticity from a consumer perspective in e-WOM research (Filieri, 2015). Although existing studies are useful, they have mainly investigated the ‘visible’ aspects of review helpfulness focusing on textual elements such as review extremity, review sentiment, review valence, review length or profile information of the reviewer (e.g. Chua and Banerjee, 2016, Mudambi and Schuff, 2010). Researching consumer's perception of information diagnosticity is important for several reasons: the voting mechanism can be easily manipulated (Filieri, 2016, Lim et al., 2010, Pan and Zhang, 2011); for example it is plausible to expect that given the importance that consumer reviews have on sales, managers may vote as more helpful those reviews that provide a positive rather than a negative evaluation of their business. Moreover, some important ‘qualitative’ information dimensions cannot be measured through quantitative textual analysis. For example, the perceived credibility of a source (i.e. the reviewer), the capacity of a review message to satisfy a consumer's specific information needs (i.e. review relevancy) or to provide plausible, fact-based information (i.e. review factuality) or up to date information (review currency). However, these factors can still be important to consumers to assess information diagnosticity.

This study applies the definition of information diagnosticity to consumer reviews and assesses consumers' perception regarding the ability of the information contained in OCRs to enable consumers to learn and to evaluate the quality and performance of services (information diagnosticity) before purchasing them. Understanding information diagnosticity is paramount for social commerce organizations because the higher the perceived diagnosticity of the information they host the better will be consumer's attitude towards shopping online (Jiang & Benbasat, 2007) and the higher will be the influence on purchase intentions (Filieri, 2015).

Additionally, most studies on e-WOM are cross-sectional and no study has measured the variations of the determinants of information diagnosticity in e-WOM over time. For instance, previous scholars have looked at the temporal evolution of e-WOM, investigating the evolution of different marketing variables and consumer posting behavior (Chen, Fay, & Wang, 2011), while other scholars emphasized the temporal dynamics in the evolution of ratings (Godes & Silva, 2012) or on how e-WOM volume evolves in movies' releases (pre-release and post-release) (Liu, 2006). In this study instead, we focus on the determinants of consumer perception of information helpfulness and how it evolves over time. We conjecture that due to the increasing importance of OCRs, the global echo produced by mass media on the phenomenon of fake reviews (e.g. Gartner, 2012, Smith, 2013, Tuttle, 2012), consumers may have become more cautious and attentive when they scrutinize the recommendations contained in websites publishing OCRs (Filieri, 2016). Thus, what makes a consumer review diagnostic may be subject to changes due to external factors (i.e. negative publicity from mass media) that might have changed consumers’ attitudes towards OCRs. An analysis at different points in time can provide us with some insights into how (and if) the influence of various antecedents of information diagnosticity has changed over time.

Elaboration Likelihood Model (ELM) has been adopted in e-WOM research to explain consumer cognitive processing of product reviews and evaluation of review messages (e.g. Park et al., 2007, Park and Lee, 2008, Zhang and Watts, 2008, Lee et al., 2008, Lee and Lee, 2009, Cheung et al., 2012, Filieri and McLeay, 2014). In this study, ELM (Petty & Cacioppo, 1986) has been used to investigate whether central and peripheral cues of information processing affect perceived information diagnosticity considering the potential moderating effect of consumer's involvement with a purchase. In line with the ELM, we have developed and tested a model that measures the influence of some central cues, namely length, relevancy, currency, and factuality; and of some peripheral cues of information processing, namely source credibility, and overall ranking scores; on information diagnosticity (dependent variable) considering the moderating role of consumer involvement with a purchase. The model was tested using regression analysis respectively in 2011 with 334 respondents and in 2016 with 297 respondents.