The effect of online reviews on product sales: A joint sentiment-topic analysis

https://doi.org/10.1016/j.im.2018.04.007Get rights and content

Abstract

This research examines the business impact of online reviews. It empirically investigates the influence of numerical and textual reviews on product sales performance. We use a Joint Sentiment-Topic model to extract the topics and associated sentiments in review texts. We further propose that numerical rating mediates the effects of textual sentiments. Findings not only contribute to the knowledge of how eWOM impacts product sales, but also illustrate how numerical rating and textual reviews interplay while shaping product sales. In practice, the findings help online vendors strategize business analytics operations by focusing on more relevant aspects that ultimately drive sales.

Introduction

Online reviews play an essential role in shaping customers’ awareness and perceptions about products [1], [2], [3]. As a major source of electronic word-of-mouth (eWOM), online reviews serve as a reliable source of information about the quality of goods, particularly goods that cannot be easily characterized before its use [1]. On e-commerce platforms, online product reviews enable customers to evaluate and compare alternatives before making purchase decisions [4]. Therefore, it is considered as a main driver for future product sales [5].

A considerable amount of research has studied the relationship between online reviews and product sales [5], [6], [7], [8], [9]. Although most evidence suggests that collectively eWOM has an impact on future sales, the findings are not always consistent. For example, Duan et al. [6] find that the volume of eWOM has a positive effect on future movie revenues, while Chintagunta et al. [7] show that only the valence (star ratings) of reviews matters. The key to resolving these conflicting findings is to understand how consumers process the information embedded in eWOM. As Hu et al. [5] point out, consumers pay attention to contextual information beyond the simple statistics such as ratings and volume. As a result, the influence of reviews on sales hinges on other factors such as the strength of the brand [10], reviewer reputation [5], reviewer location [11], and review text [12], [13]. A better understanding of how the information embedded in the reviews drives sales can help businesses exploit the value of eWOM through more accurate forecasting, promoting new products, and attracting and retaining shoppers.

We contribute to the literature by proposing a new mediation model, whereby numerical “star rating” partially mediates the relationship between review texts and product sales. A typical product review contains two types of information – the numerical rating and the review text. The numerical rating is a quantitative summary of the reviewer’s experiences, attitudes, opinions, or sentiments toward a product or service, usually expressed as number of stars. The review text is an open-ended textual description of the reviewer’s opinions toward the product or service [14], [15]. Extant research on the economic impact of eWOM focuses on numerical ratings but rarely addresses textual reviews [16], in part due to the complexity of text analysis. Few studies that incorporate textual reviews use techniques such as sentiment polarity [12] or frequent noun phrases [13]. Yet, much of the value of product reviews lies in conveying “attributes and attribute dimensions using the ‘voice of the consumer’” [17]. Capturing the full economic impact of online reviews may require us to uncover the dimensions that consumers care about. New text analytics methods that go beyond sentiment analysis and counting phrase are needed.

We introduce a methodological tool to the online reviews literature – the joint sentiment-topic model (JST) [18]. This unified machine learning model achieves two goals at the same time: it not only summarizes the sentiment in the review text, but also identifies the aspects of the product that the reviewer is happy with or critical of. JST provides a much richer representation of the qualitative review data. The outputs allow us to investigate, among other things, how positive or negative valence of specific product features lead to changes in future sales. We proceed to study the impact of textual reviews and numerical ratings on the actual sales of 312 tablet PC products using a panel dataset. Furthermore, we conduct a mediation analysis to study the interplay of textual review and numerical review ratings using Baron and Kenny's [19] approach.

The findings from JST and mediation model enhance the understanding of how online reviews provide information cues and shape product sales. We show that reviews with positive and negative valence focus on different sets of product aspects. More importantly, the positive and negative aspects have different impacts on sales performance. The numerical ratings mediate the effects of textual reviews that discuss negative aspects of a product. But the effects from textual reviews that carry positive valence persist in the mediation model. In a nutshell, reviews that highlight the positive aspects of the product provide an extra boost to sales that cannot be captured simply by a “5-star” rating. Our findings underscore the importance of analyzing social media data to e-commerce. Our research framework can also help online vendors strategize their business analytical initiatives by focusing on more relevant aspects of eWOM. Additionally, we demonstrate an innovative approach to analyzing textual data along with numerical data, which may be valuable for similar research in the future.

The rest of the paper is organized as follows. Section 2 discusses the theoretical and literature background of the research and formulates hypotheses. Section 3 presents the data and research method used for the study. Section 4 reports the analysis and results. Section 5 concludes the paper with a discussion of implications, limitations, and avenues for future research.

Section snippets

Literature review and hypothesis development

Many mechanisms can account for how eWOM affects future product sales. First, online reviews can serve as a signaling device in the context of imperfect information [9], [20]. In online shopping, the prospective buyers usually lack the experience that product reviewers have. Through prior purchases and usage, reviewers possess valuable information about the product such as quality, value, and potential issues – the information that prospective buyer needs but lacks for comparing alternatives.

Data

The online reviews dataset is titled “Market Dynamics and User-Generated Content about Tablet Computers” and is provided by Wang, Mai and Chiang [51]. The dataset contains 88,901 consumer reviews on 794 tablet computer products or SKUs. The weekly market dynamics and reviews data were collected using a Java web crawler during a 24-week period from February 1 to July 11, 2012.1

Empirical model and estimation

To establish the relationship of numerical ratings and textual reviews on sales, we follow [13], and use a dynamic panel data (DPD) model with the following estimation equationyit=αyi,t1+βxit+γzit+εit,εit=ζi+uitwhere yit is log(saleit), or the log-sales for product i at week t. xit is the review variables, which could be numerical ratings, overall sentiment, or positive and negative aspects of the products. zit is a vector of control variables such as price, and measure of product newness such

Discussion

This study examines the influence of online word-of-mouth on the sales performance of products. In particular, we investigate the impacts of numerical star ratings and sentiments expressed in textual reviews on sales. We use JST to mine the topics as well as the sentiments associated with each topic. The results reveal two prominent positive topics – “hedonic experience” and “hardware” and two dominant negative topics – “user interface” and “logistics and customer service”. The results suggest

Acknowledgement

The first author would like to acknowledge the financial support of The National Natural Science Foundation of China (Project No. 71471125) for the completion of this paper.

Xiaolin Li is an Associate Professor of e-Business and Technology Management with the College of Business and Economics of Towson University. He received his Ph.D. in Management Systems from Kent State University. Dr. Li’s current research emphases are electronic commerce and big data. His research has been published in Information Systems Research, Journal of the Association for Information Systems, Decision Sciences, International Journal of Production Economics, Electronic Commerce Research,

References (60)

  • N. Hu et al.

    Do online reviews affect product sales? The role of reviewer characteristics and temporal effects

    Inf. Technol. Manag.

    (2008)
  • P.K. Chintagunta et al.

    The effects of online user reviews on movie box office performance: accounting for sequential rollout and aggregation across local markets

    Mark. Sci.

    (2010)
  • J.A. Chevalier et al.

    The effect of word of mouth on sales: online book reviews

    J. Mark. Res.

    (2006)
  • F. Zhu et al.

    Impact of online consumer reviews on sales: the moderating role of product and consumer characteristics

    J. Mark.

    (2010)
  • N.N. Ho-Dac et al.

    The effects of positive and negative online customer reviews: do brand strength and category maturity matter?

    J. Mark.

    (2013)
  • C. Forman et al.

    Examining the relationship between reviews and sales: the role of reviewer identity disclosure in electronic markets

    Inf. Syst. Res.

    (2008)
  • N. Hu et al.

    Ratings lead you to the product, reviews help you clinch it: the dynamics and impact of online review sentiments on products sales the mediating role of online review sentiments on product sales

    Decis. Support Syst.

    (2013)
  • N. Archak et al.

    Deriving the pricing power of product features by mining consumer reviews

    Manage. Sci.

    (2011)
  • S. Mudambi et al.

    What makes a helpful online review? A study of customer reviews on amazon.com

    MIS Q.

    (2010)
  • S. Moon et al.

    The impact of text product reviews on sales

    Eur. J. Mark.

    (2014)
  • T. Lee et al.

    Automated marketing research using online customer reviews

    J. Mark. Res.

    (2011)
  • C. Lin et al.

    Joint sentiment/topic model for sentiment analysis

    Proceeding 18th ACM Conf. Inf. Knowl. Manag. – CIKM ’09, ACM

    (2009)
  • R.M. Baron et al.

    The moderator-mediator variable distinction in social the moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations

    J. Pers. Soc. Psychol.

    (1986)
  • N. Hu et al.

    Can online word-of-mouth communication reveal true product quality?

    Proc. 19th Int. Conf. Inf. Qual.

    (2014)
  • L.L. Hellofs et al.

    Market share and customers’ perceptions of quality: when can firms grow their way to higher versus lower quality?

    J. Mark.

    (1999)
  • O. Müller et al.

    Using text analytics to derive customer service management benefits from unstructured data

    MIS Q. Exec.

    (2016)
  • D.M. Blei et al.

    Latent dirichlet allocation

    J. Mach. Learn. Res.

    (2003)
  • E. Brynjolfsson et al.

    Consumer surplus in the digital economy: estimating the value of increased product variety at online booksellers

    Manage. Sci.

    (2003)
  • A. Ghose et al.

    Evaluating pricing strategy using e-commerce data: evidence and estimation challenges

    Stat. Sci.

    (2006)
  • P. Miller

    Theories of Developmental Psychology

    (1994)
  • Cited by (209)

    View all citing articles on Scopus

    Xiaolin Li is an Associate Professor of e-Business and Technology Management with the College of Business and Economics of Towson University. He received his Ph.D. in Management Systems from Kent State University. Dr. Li’s current research emphases are electronic commerce and big data. His research has been published in Information Systems Research, Journal of the Association for Information Systems, Decision Sciences, International Journal of Production Economics, Electronic Commerce Research, Journal of Organizational Computing and Electronic Commerce, Journal of Computer Information Systems, among others. He received the “Distinguished Research Paper Award” (in innovative education track) from the DSI Annual Conference in 2009 and Towson University College of Business and Economics Outstanding Scholarship Award in 2012.

    Chaojiang Wu is Assistant Professor in LeBow College of Business at Drexel University. He obtained his PhD from the University of Cincinnati. His research focuses on data analytics and its business applications such as Finance and Information Systems. His research develops quantitative methods in analyzing complex data for better decisions and applies such methods to applications in finance, online word of mouth, and scholarly communications. He has published in journals such as Journal of Financial and Quantitative Analysis, IIE Transactions, Statistics and Computing, and Computational Statistics and Data Analysis, among others.

    Feng Mai is an Assistant Professor of Information Systems at the School of Business, Stevens Institute of Technology. He holds a Ph.D. in Business Administration from the University of Cincinnati. Professor Mai’s research interests are social media, marketing analytics, and FinTech. His work has been published in journals such as Marketing Science, Journal of Consumer Research, and European Journal of Operational Research.

    View full text