Too obvious to ignore: Influence of popular reviews on consumer online purchasing decisions

Abstract

BACKGROUND:

Online reviews, as an important way for consumers to understand product information, have an important impact on consumers’ online shopping decisions. A lot of useful explorations have been made on the role of online reviews in existing empirical research, but the interaction between online reviews and its subdivided dimensions have not been explored.

OBJECTIVE:

Based on the two-step flow theory, this article aims to explore the impact of online review valence, review volume, and their interactions on online sales, focusing on the question of what are the factors that influence customer purchase decisions and what is the moderating effect of popular reviews on review valence.

METHODS:

Empirical analysis was done by tracking the product information and online sales data of mobile phone products and laptops in search goods category on the Amazon.cn website.

RESULTS/CONCLUSION:

The research results found that: (1) in terms of review valence, the average score significantly promotes online sales, and negative word-of-mouth significantly decreases online sales; (2) as for review volume, the number of total reviews and popular reviews have significantly promote online sales; (3) regarding the interactions between the review valence and review volume, popular reviews significantly enhance the impact of review valence on online sales, playing a complementary effect for review valence.

Keywords

Online review opinion leader popular comment online sales search goods

Xin Su is an assistant professor in School of Economics and Management, Beijing University of Posts and Telecommunications. She got her PhD in School of Business, Renmin University of China. Her research focuses on customer behaviour analysis and big data driven management innovation. Her work has appeared in Journal of Cleaner Production, Chaos, Solitons and Fractals, Journal of Intellectual Capital, Applied Economics et al. She can be contacted at: E-mail: xin.su@bupt.edu.cn.

Mingzi Niu is a PhD candidate in Department of Economics, Rice University. She got her bachelor degree in School of Economics, Peking University and master degree in Department of Economics, Duke University. Her research focuses on applied micro theory. She can be contacted at: E-mail: mingzi.niu@rice.edu.

1 Introduction

With the expansion of online shopping customer groups and the obvious differences among users, customers’ online shopping decisions will have a direct influence on the development of E-commerce platforms, at the same time, sellers or others will unavoidably influence customers as they shop online [1]. In traditional shopping environments, official announcements, third-party comments, and word-of-mouth from friends and relatives are the three major sources of information for customers [2]. The internet has added a fourth source of information; namely, online customer reviews provided by unidentified customers. As the principal form of online word-of-mouth publicity, online customer reviews are reviews of products or companies submitted by customers that evaluate products using text, images, or videos [3]. Information communication through the internet transforms traditional interpersonal communication into online communication. Since simultaneous one-to-many communication, unrestrained by time, place, or reputation, characterizes online word-of-mouth, it has drawn the attention of researchers and practical managers. User Generated Content (UGC) is a part of online reviews, the general term for content generated or created by users. Similar expressions include “User-Created Content” and “Customer Generated Information,” which refer to situations in which customers put original reviews on an internet platform, or offer reviews to other users [4].

Before customers make their purchase decisions, they can obtain relevant product information by reading online reviews, and some customers make purchases due to herd mentality [5]. Therefore, research on the factors that influence online purchase decisions has become a hot topic in the field of customer behavior [6]. Prior studies on customer decisions have analyzed the factors that influence customers’ online purchase decisions from the perspectives of technology acceptance model theory, customer involvement theory, the theory of planned behavior, transaction cost theory, and utility theory, constructing various theoretical frameworks for customer purchase decisions [7 –10]. Since the existing research in this area has made beneficial contributions that clarified the factors that influenced customer choice from the perspective of general customers or reviewers, newer studies should pay more attention to popular reviews, and explore details that are more specific.

Online reviews on the Amazon.cn are not displayed in the order of purchase date, they are displayed according to the usefulness of online reviews. Therefore, when consumers browsing reviews online, they will see the popular reviews first. Popular reviews enable potential consumers find useful review contents quickly, which greatly reduces the search costs of potential consumers. Due to the time and energy constraints of consumers, most of them will not go through all product reviews, and they are more likely to focus on the reviews that are placed at the top. Therefore, the number of words of popular reviews and the depth of the online reviews will largely affect consumers’ understanding of the products. The more words the review has, the more accurate it can reflect the product’s functional attributes and the shopping experiences.

Based on the analysis above, this study focuses on the following questions. First, what are the factors that influence customer purchase decisions in the online shopping environment? Second, based on two-step flow theory, what is the moderating effect of popular reviews on review valence? The rest of this paper is organized as follows: Section 2 contains the literature review, Section 3 presents the research design, Section 4 provides the empirical results, and Section 5 discusses the conclusions and discussions.

2 Literature review and hypotheses

2.1 User generated content

With the expansion of E-commerce, an increasing number of customers choose to shop online. While online shopping, UGC tends to influence customers. Broadly defined, UGC is the text, pictures, audio, and video posted on websites by customers. Because UGC comes from groups of customers, it has been the dominant factor influencing online shopping decisions [11]. The Organization for Economic Co-operation and Development (OECD) believes that UGC refers to content made by amateurs through non-professional channels, involving creative works that are available online [12]. Duan et al. believe that customers, rather than websites, submit UGC, and that real customers originally created the content or copied it from other sources [13]. The OECD listed three characteristics of UGC in its 2007 report: (1) someone published it on the internet, (2) it is, to some extent, innovative, and (3) neither professional users nor authoritative organizations created it. Chatterjee et al. believed that UGC should include at least two characteristics [14]. First, amateurs made it, and it involved some innovative content or the modification and editing of existing content. Second, other people must share the UGC on the internet [15].

Since UGC has a crucial impact on customers’ potential purchase decisions, understanding UGC’s influencing mechanism on customers in a social e-commercial environment is a key problem for platform enterprises [16]. Most existing studies have focused on analyzing the direct relationship between online customers’ UGC and their purchase decisions [17]. Berger et al. found that online customer reviews had an obvious influence on customers’ purchase desires and enterprises’ market performance [3]. In addition, Goes et al. analyzed the influencing mechanism under various conditions, such as UGC with different attributes (controversial reviews), product types, and publisher and reader characteristics [18]. However, a socialized E-commerce environment is a complex, influencing space in which UGC impacts customer group behaviors by constructing social network relationships, and then influences customers’ individual decisions [19]. The social transmission mechanism of UGC on individual customers is not yet evident, and there is a lack of research on the different effects and functions of UGC within various social network structures. Taking all of that into account, we propose the following hypothesis:

H1: The number of reviews has a significant and positive impact on online sales.

2.2 The two-step flow theory

Considering the integration of the digital, service, experience, and sharing economies via the internet and E-commerce, an increasing number of people have become spontaneous and keen on expressing their ideas on the open internet. The increasing popularity and commercialization of social media platforms or social networks has made socialized commercial activities, such as communication, interaction, recommendation, cooperation, and crowdsourcing among user groups, more profound [18]. In some situations, users even participated in the open cooperation and innovation of companies. Since major enterprises and organizations actively employ user reviews to carry out online marketing, the effect of popular reviews on users has increased as more users pay attention to them, which can spread and influence others’ decisions.

In the 1940s, Paul proposed two-step flow theory. According to the theory, opinion leaders pay close attention to mass media and pass on their interpretations of media messages to general audiences, rather than spread information directly to others, thus forming the “mass media-opinion leader-general audience” two-step flow model. Through a comparison of the literature, it was clear that opinion leaders had more subjective interests, and the frequency and quantity of their exposure to media was much higher than that of general audiences [19]. Opinion leaders then transmitted the knowledge they gained from mass communications to general audiences, which had a certain impact on the formation of public opinions [20, 21]. In the study of Dellarocas et al., mass media had a limited influence on public opinion, and it was more suited for providing information rather than influencing others [12]. Hsin et al. suggested that the media might influence opinion leaders first, and then influence others through opinion leaders [22]. Opinion leaders use the special norms and psychological advantages of groups to share their opinions and values and exert influence on other group members [23]. With the rise and development of the internet, people paid more attention to the influence of online opinion leaders. The internet’s unique environment provided material carriers for the appearance of opinion leaders, the diversified needs of netizens provided spiritual power to opinion leaders, and the dependence of general netizens on authority provided a psychological basis for the spread of information [24]. This leads us to propose the following hypothesis:

H2: The number of popular reviews has a significant and positive impact on online sales.

2.3 The average rating

With the rapid development of computer technology and the internet, online shopping has become one of the most important parts of the daily lives of netizens. More and more E-commerce websites have added an online review function where customers can review the quality of their products and services. The daily exponential growth of online review data creates a word-of-mouth effect, and customers have become accustomed to looking up online word-of-mouth information before making their purchase decisions [25]. To some extent, online reviews and ratings can reflect product quality and features. Customers are free to make reviews that are extensive, powerful, and help an enterprise expand its brand marketing, while also increasing the risk of exposing product defects. Customers also make decisions according to previous reviews and ratings based on their judgments [26].

In an online shopping environment, websites expose customers to a graph of the online word-of-mouth reviews of the product while browsing information on the product, a graph that visually presents the overall distribution of product evaluations made by existing customers [27]. If comments are entirely positive or negative, it is easy for customers to make their purchase decisions [28]. However, when the distribution of the product’s word-of-mouth reviews presents a mix of praise and criticism, there is uncertainty in the customer’s purchase decision [29]. Although marketers pursue beneficial comments through a variety of measures [30], such as cash back for positive comments, gifts, and letters of gratitude, there are still inconsistent online word-of-mouth conditions [31]. Therefore, we propose the following hypothesis:

H3: The average rating has a significant and positive impact on online sales.

2.4 The negative reviews

With the development of the internet, E-commerce platforms have improved, and online shopping has gradually become an important part of people’s lives [32]. While online shopping, online customer reviews are an important channel of communication among online customers, and they are an important publicity source for online sellers that hold commercial activities [33]. There is no denying that current consumption trends have changed from traditional offline consumption to online consumption. Online reviews have gradually become an important activity that influences the online purchase behaviors of potential customers.

The 1950s introduced the concept of word-of-mouth into the marketing arena. In certain situations, word-of-mouth had a more significant influence on customer behavior than other activities, such as personal selling or advertising. There is no doubt that online reviews, as an extension of offline word-of-mouth, reflect a broadening of traditional word-of-mouth [34, 35]. Liu et al. believed that negative word-of-mouth included when customers informed acquaintances of their unsatisfied experiences, or when customers expressed negative attitudes toward a product or an enterprise [32]. Berger et al. refined the concept of negative online reviews, and pointed out that their aim was to prevent others from buying certain goods [3]. Meanwhile, Dellarocas et al. defined negative online reviews as an information-interaction behavior where customers changed unhappy virtual experiences into text, which they regarded as an online user feedback system [12]. Based on prior studies, some scholars understood negative online reviews to be customers employing online platforms as communication channels to express negative information about enterprises or products, mainly through discontent or disappointing customer experiences. Although scholars differed in their definitions of negative online reviews, they also had some consistency, which primarily manifested as the idea that the main subject in the process was customers, online platforms worked as intermediaries, and the content of negative reviews was an evaluation and communication regarding the product that emphasized a “dissatisfied” feature. In short, negative online reviews take the shape of visuals, text, pictures, videos, and other forms that customers use on the medium of the network platform to relay abstract negative information about products and services, mainly consisting of unhappy customer experiences and dissatisfied emotions. As for customer groups, word-of-mouth has two contrary aspects. Positive word-of-mouth can promote customer purchases, while negative word-of-mouth can hinder sales. Negative word-of-mouth is more likely to cause a potential loss of customers. Therefore, we propose the following hypothesis:

H4: Compared to positive reviews, negative reviews have a greater influence on online sales.

2.5 The popular reviews

Prior studies have confirmed that both the valence and volume of online reviews affect online sales. Customers who have purchased products with higher average ratings and fewer negative reviews are often more satisfied with the quality and performance of the products, and usually have better user experiences. Customers are more likely to recognize such products, and they have higher sales [3]. Products with higher sales have more people paying attention to them due to their large sales, resulting in extensive discussions among customers about those products [8]. Others are more likely to transmit the comments of opinion leaders, which affects the evaluations of the products through two-step flow theory, which in turn affects the review valence. From this point of view, the valence and volume of online reviews are not independent; there is a certain interaction between them which affects the final online product sales. Thus, we propose the following hypothesis:

H5: The number of popular reviews strengthens the effect that review valence has on sales.

3 Research design

3.1 Research sample

A review of the research throughout the world revealed that many scholars obtained their results through data analysis of the Amazon website [36]. Amazon has a full range of Chinese products in detailed categories with many reviews, giving a high degree of data availability that satisfies the needs of this study well. In addition, as a giant global B2C E-commerce website, Amazon has built a relatively comprehensive online review system through years of accumulation. Based on the considerations above, this study selected the Amazon.cn website as its research object. In view of the impact of product types on research results, we selected search goods as our research object. There were a huge number of comments and descriptions of search goods on the Amazon.cn website [37], which ensured that the product information collected met the requirements of this study.

Based on the work of Bei et al., who chose mobile phones as their search goods, we selected mobile phones as our research objects for the following reasons. (1) The parameters of mobile phones, such as brand and type, are relatively stable; therefore, customers identified their targeted mobile phones through categories such as pixels, memory size, display size, and screen resolution when buying mobile phones on Amazon. (2) Because mobile phones can be recycled and the parts disassembled and separated for reuse, the purchase channel is an important guarantee of the quality of mobile phones. The guarantee of authenticity provided by the Amazon.cn website reduced customers’ worries about unqualified products and improved customers’ trust in shopping websites. (3) Mobile phones are high-tech products, so customers need to search and refer to a lot of information before making their purchase decisions. The Amazon China website, as a representative of B2C E-commerce platforms, provided potential customers with a detailed introduction to mobile phones as well as many post-purchase comments.

3.2 Data collection

During the data collection stage, this study collected data on product descriptions and customer-generated reviews twice. The first stage was preliminary research in June 2017, when we conducted static data collection on product information and the valence and volume of online reviews in the mobile phone category on the Amazon.cn website. Through data analysis on the preliminary survey, we initially verified the proposed research hypothesis. The second stage of data collection was in December 2017, when the mainstream Python program Scrapy scraped data from the Amazon website and automatically stored the data in a MySQL database. For analysis and mining, the data were output offline in the form of a CSV file. Constant, real-time monitoring of network data kept the data updated, and the updated consumption data was stored in the database. Due to restrictions on the Amazon website, we could only obtain information on the top 100 products in a certain category. During the data cleaning process, we removed outliers, and any samples that had missing information, from the sample data. Please see the sample of data records structure in the Appendix.

Through tests during the preliminary investigation, we removed uncorrelated variables and selected the name of the product, its sales ranking, suggested retail price, the actual price, the brand, the sales distributor, the number of comments, the number of popular comments, the average rating, and the numbers of 5, 4, 3, 2, and 1-star ratings as the variables for data collection. Since the suggested retail price of a commodity is highly correlated with its actual price, this study used the discount (the actual price divided by the suggested retail price) as a variable for analysis. Considering the collinear problems between prime distribution, sales distributors, and commitments to return and exchange, the sales distributors were also research variables in this study. Due to the high correlation between 1-star and 5-star reviews, and we primarily prepared to investigate the impact of negative reviews on online sales, this study only recorded the number of 1-star reviews.

3.3 Research variables

To explore the influencing model of online reviews on online sales, this study selected the sales ranking of commodity i in category j on day t on the Amazon.cn website as the dependent variable, customer perceived value and perceived risk as control variables, and the valence and volume of online reviews as the independent variables. Table 1 provides specific definitions for the variables utilized in the model.

Table 1
The Definitions of variables in the model

Variables Definitions

Online sales rank Sales ranking (1 represents the highest sales)

Perceived value price The suggested retail price

dct Actual price/retail price

Perceived risk brand The brand category

merch The sales distributor

Online reviews nofcmt The number of reviews

nofpop The number of popular reviews

averate The average rating (1 represents the lowest rating)

nof1 The number of 1-star rating

Variables		Definitions
Online sales	rank	Sales ranking (1 represents the highest sales)
Perceived value	price	The suggested retail price
	dct	Actual price/retail price
Perceived risk	brand	The brand category
	merch	The sales distributor
Online reviews	nofcmt	The number of reviews
	nofpop	The number of popular reviews
	averate	The average rating (1 represents the lowest rating)
	nof1	The number of 1-star rating

3.4 Research model

The dependent variable was the sales ranking of the goods in the mobile phone category, and it was an ordinal variable. The independent variables, such as the number of reviews, the number of popular reviews, the average rating, and the number of 1-star ratings, were continuous variables. The control variables of brand and sales distributor were categorical variables. Therefore, this research initially adopted an ordinal logistic regression model to analyze the relationships between the variables. According to prior studies, when ordinal variables have 10 or more levels, they can be treated as continuous variables in the model. In addition, since the sales rank has numerical values in a reversed order, the main model utilized the reciprocal of the rank, written as Rank1. Finally, this study adopted multiple linear regression to analyze the degree of fit for the data and hypotheses.

Based on the above hypotheses regarding the variables, the theoretical model of this study is as follows. According to H1, H2, and H3, the number of reviews, the number of popular reviews, and the average rating had a positive impact on online sales; namely, α1 > 0, α2 > 0, and α3 > 0. According to H4, negative reviews had a greater impact on online sales, meaning that α4 < 0. $Rank 1_{itj} = β_{1} P V_{itj} + β_{2} P R_{itj} + β_{3} O R_{itj} + λ_{itj}$ (1)

According to the theoretical model proposed in this study, Equation (1) reflects the factors that influence the online sales of search goods on Amazon’s website. Customer perceived value (PV), perceived risk (PR), and online reviews (OR) all influenced the online sales of commodity i in category j on day t on the Amazon.cn website. PV is measured using price and discount, PR is measured using brand and sales distributors, and OR is measured using the number of reviews, the number of popular reviews, the average rating, and the number of 1-star ratings. We expressed the OR measurement equation as follows:

$\begin{matrix} O R_{itj} = α_{1} * NofCm t_{itj} + α_{2} NofPo p_{itj} + α_{3} \\ * AveRat e_{itj} + α_{4} * NofP 1_{itj} + δ_{itj} \end{matrix}$ (2)

For commodity i in category j on day t, λ_itj refers to any factors, other than the influencing factors proposed in this study, that explain changes in online sales. Similarly, δ_itj refers to any variables that had an impact on online reviews other than the measurement factors affecting online reviews proposed in this study.

4 Results

4.1 The descriptive statistics

Of the top 100 products by sales in the mobile phone category, we removed 2 samples with missing values and retained 98 samples, for an effective sample rate of 98%. From the descriptive statistics on mobile phones, the market shares of various mobile phone brands on Amazon.cn website vary greatly. Among the brands, Huawei accounts for 46% of sales, Apple 26%, and Xiaomi only 7%, with the three brands accounting for a combined 79% of total sales. Moreover, prices differed greatly between the products, and the discount rate was low. The average rating for mobile phones was 4.13, indicating that most customers were very satisfied with their mobile phone purchases on Amazon.cn. 5-star reviews accounted for 66% of all reviews, indicating that most customers give a 5-star review to their mobile phone purchase experience, with the second and third most frequent scores being 4 star and 1-star reviews, indicating that nearly all post-purchase mobile phone reviews by customers fell on the extreme ends of the ratings spectrum. Figure 1 presents the frequency distribution statistics for the different brands in the mobile phone sample.

Fig. 1

The brand frequency distribution of mobile phones.

4.2 The correlation analysis

We employed linear correlation analysis to study the degree of linear correlation between two variables, where r represented the correlation coefficient. In general, if x and y moved in the same direction, namely if r > 0, it meant that the variables were positively correlated, r < 0 indicated that the variables were negatively correlated; |r| > 0.95 that the variables were extremely correlated; |r| ≥ 0.8 that the variables were highly correlated; 0.5 ≤ |r| < 0.8 that the variables were slightly correlated; 0.3 ≤ |r| < 0.5 that there was a low correlation between the variables; and |r| < 0.3 that the correlation was very weak, and the variables were uncorrelated.

As shown in Table 2, for the research variables in the mobile phone samples, the correlation coefficient between most of the variables was less than 0.3, except for the significant correlation between the number of 1-star ratings and the number of comments. The correlation between some variables was lower than 0.05, indicating very low correlation between those variables. Most of the variance inflation factors were below 2, which was much lower than 10 and indicated that there was no multicollinearity problem between the variables. It is worth noting that there was a significant and positive correlation between the number of popular reviews and the average rating, indicating that, for mobile phone products, the more popular reviews they had, the more likely they were to have a higher rating.

Table 2
The correlation analysis

Mean SD VIF 1 2 3 4 5 6 7 8

1rank1 0.05 0.17 2.30 1

2price 1753.28 1.00 1.75 0.05 1

3dct 0.95 0.11 1.02 0.01 0.03 1

4merch 0.76 0.09 1.12 0.24 –0.19 0.09 1

5averate 4.13 0.83 1.84 0.15 0.29 –0.08 0.08 1

6pstar1 0.07 0.09 2.02 –0.32 –0.17 0.03 –0.04 –0.53 1

7nofcmt 175.61 134.27 1.19 0.18 –0.13 0.05 –0.09 0.01 –0.05 1

8nofpop 14.32 4.63 1.67 0.19 0.07 0.06 0.07 0.38 –0.42 0.38 1

	Mean	SD	VIF	1	2	3	4	5	6	7	8
1rank1	0.05	0.17	2.30	1
2price	1753.28	1.00	1.75	0.05	1
3dct	0.95	0.11	1.02	0.01	0.03	1
4merch	0.76	0.09	1.12	0.24	–0.19	0.09	1
5averate	4.13	0.83	1.84	0.15	0.29	–0.08	0.08	1
6pstar1	0.07	0.09	2.02	–0.32	–0.17	0.03	–0.04	–0.53	1
7nofcmt	175.61	134.27	1.19	0.18	–0.13	0.05	–0.09	0.01	–0.05	1
8nofpop	14.32	4.63	1.67	0.19	0.07	0.06	0.07	0.38	–0.42	0.38	1

The sales ranking selected for this study was the sample data of the top 100 sellers in the mobile phone category. Table 3 presents the regression model and results of this study after taking the reciprocal of sales ranking as a proxy for sales volume and treating them as continuous variables. Models M1–M3 verified hypotheses H1–H3 of this study. Of those, M0 was the null model, M1 was the main effect model, and models M2 and M3 revealed the moderating effect.

Table 3

The regression analysis

	M0	M1	M2	M3
price	0.007	0.005	0.004	0.002
	(0.67)	(0.52)	(0.50)	(0.22)
dct	–0.460	–0.320	–0.308*	–0.278
	(–1.32)	(–1.59)	(–1.35)	(–1.57)
merch	0.009	0.008	0.006	0.007
	(1.21)	(1.04)	(1.13)	(1.03)
averate		0.012^**	0.031^*	–0.010
		(1.00)	(1.87)	(–0.59)
lgnof1		–0.003^*	–0.020^*	–0.005
		(–0.83)	(–1.63)	(–1.33)
nofcmt		0.000^***	0.000^***	0.000^***
		(8.31)	(8.26)	(7.47)
nofpop		0.005^**	0.004^**	0.008^*
		(1.30)	(0.92)	(01.62)
Brand dummy		Include	Include	Include
averate_nofpop			0.002^**
			(1.37)
lgnof1_nofpop				–0.000^**
				(–4.64)
_cons	0.301^*	0.279^*	0.137^*	0.198^*
	(2.17)	(1.92)	(2.03)	(1.99)
N	98	98	98	98
R squared	0.3121	0.2637	0.3402	0.3101

t statistics in parentheses. ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

According to M1, review volume (the number of reviews) promoted online sales (t = 8.31), but the coefficient was relatively small; therefore, this supported H1. The number of popular reviews had a statistically significant impact on online sales; therefore, this supported H2. In terms of review valence, the average rating had a statistically significant impact on online sales; therefore, this supported H3. Negative word-of-mouth had a statistically significant and negative impact on online sales (t = –0.83); therefore, this supported H4. M2 revealed that the number of popular reviews enhanced the positive effect of the average rating on online sales, while M3 confirmed that the number of popular reviews intensified the adverse effect of negative reviews on online sales. To summarize, the number of popular reviews enhanced the impact of online review valence on online sales, which supported H5.

4.4 The robustness test

In order to verify the robustness of the regression results, we performed a robustness test on two groups of models in this study, where we replaced the dependent variable in M4 and the product samples in M5. For the central effect of M1, since the original dependent variable was sales ranking, we used the reciprocal of sales ranking in M1 in order to recognize the influence of online reviews on online sales more conveniently. Thus, M4 once again verified the main effect after choosing the original online sales ranking as the dependent variable.

Search goods were the research object in this study. We utilized mobile phones to represent search goods in M1. Based on prior studies on search goods, we selected the laptop category as the research object for M5 and followed the same techniques that we utilized for the mobile phone models, collecting the top 100 sellers in the laptop category. Of those 100 laptops, we removed 6 examples with missing values and retained 94 samples, making the effective sample rate 94%. In the laptop sample, HP accounted for 27%, Lenovo occupied 29%, and Dell only compromised 21% of sales, with the three major brands accounting for 77% of all sales in the laptop market.

The robustness test results (Table 4) indicated that, after changing the dependent variable and product category, the regression results for the main variables were mostly consistent with the original results.

Table 4
The robustness test

M1 M4 M5

price 0.005 –0.29 0.17^*

(0.52) (–0.08) (3.92)

Dct –0.320 1.02^* –0.05^***

(–1.59) (4.95) (–8.37)

merch 0.008 –0.008^* 0.00^*

(1.04) (–4.52) (2.57)

averate 0.012^ –3.37^* 0.00^

(1.00) (–3.11) (0.89)

lgnof1 –0.003^* 0.33^* –0.00^

(–0.83) (1.09) (–1.85)

nofcmt 0.000^* –0.96^* 0.00^*

(8.31) (–3.42) (0.73)

nofpop 0.005^** –0.87^* 0.05^*

(1.30) (–2.25) (1.36)

Brand dummy Include Include Include

_cons 0.279^* –8.87^* 0.59

(1.92) (–0.45) (2.99)

N 98 98 94

R squared 0.2014 0.2992 0.3037

	M1	M4	M5
price	0.005	–0.29	0.17^***
	(0.52)	(–0.08)	(3.92)
Dct	–0.320	1.02^***	–0.05^***
	(–1.59)	(4.95)	(–8.37)
merch	0.008	–0.008^*	0.00^***
	(1.04)	(–4.52)	(2.57)
averate	0.012^**	–3.37^***	0.00^**
	(1.00)	(–3.11)	(0.89)
lgnof1	–0.003^*	0.33^*	–0.00^**
	(–0.83)	(1.09)	(–1.85)
nofcmt	0.000^***	–0.96^***	0.00^***
	(8.31)	(–3.42)	(0.73)
nofpop	0.005^**	–0.87^*	0.05^*
	(1.30)	(–2.25)	(1.36)
Brand dummy	Include	Include	Include
_cons	0.279^*	–8.87^*	0.59
	(1.92)	(–0.45)	(2.99)
N	98	98	94
R squared	0.2014	0.2992	0.3037

t statistics in parentheses. ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

5 Conclusions and discussions

5.1 Conclusions

Based on two-step flow theory, this study explored the factors influencing the online sales of search goods from the perspective of opinion leaders, and drew the following conclusions.

First, in terms of review volume, the number of reviews promoted online sales considerably. Customers collected a lot of descriptive information about a good’s functions and parameters before purchasing the search good [38]. Online reviews from people who had purchased the product provided a large quantity of detailed and authentic information about the products that supplemented the product information provided by the seller [39]. The more product reviews there were, the more information potential customers had about the shopping experience and product features. The richer the content of the online reviews, the more customers learned about product features, and the more likely they were to make purchase decisions [40]. The number of popular reviews also had a statistically significant impact on online sales, because the content in popular reviews was widely discussed by customers, who reflected the attitudes of opinion leaders in virtual online communities about the products and the shopping experience. Driven by the herd effect, potential customers tended to follow the crowd, and made purchase decisions based on the reviews of opinion leaders [41].

Second, in terms of review valence, the average rating had a statistically significant and positive effect on online sales. Before purchasing their mobile phones, customers often sought out a great deal of information about them, in order to become familiar with the product’s performance and evaluate the product [42]. The online ratings given by customers who already purchased the product were not only an indication of the quality and performance of the product, but they also evaluated the overall shopping experience (such as express delivery or the after-sales service) [43]. Potential customers made purchase decisions based on online ratings [44]. In addition, customers’ evaluations of mobile phone products displayed a high level of consistency; mobile phone products that already had large numbers of favorable comments were often preferred by customers, which increased their online sales [45]. In terms of negative reviews, the conclusions of this study confirmed that negative word-of-mouth had a statistically significant and adverse impact on online sales. Negative reviews often revealed deficiencies in the product’s performance, and for search goods, the online shopping conversion cost was very low for customers. Therefore, when potential customers discovered, via negative reviews, that a product’s performance did not meet their needs, they tended not to make purchase decisions, thus negatively affecting the product’s online sales [46].

Finally, we conclude that there was an interactive relationship between review valence and volume [47]. Customers more easily accepted products with higher average ratings and fewer negative reviews [48], which promoted the products’ sales. Furthermore, customers with limited time and energy tended to mimic the choices of others [49]; that is, customers exhibited herd behavior, and were likely to be influenced by opinion leaders who were very popular in their online communities [50].

5.2 Contributions

Theoretically, we discovered that the number of reviews, the number of popular reviews, and the average rating of the online reviews had a positive effect on sales, while negative reviews had a significant and adverse effect on sales, which verified some results from prior studies on online sales [51]. Existing studies have generally explored the effect of online review valence and volume on online sales using various product categories, while paying less attention to the interaction between the different dimensions of online reviews. Utilizing the two-step flow theory perspective, this study investigated the moderating effect of the number of popular reviews on review valence. It revealed that popular reviews markedly boosted online review valence’s influence on sales, which also verified the role of opinion leaders in prior studies on online reviews [52].

In terms of practical implications, as the information E-commerce enterprises provide becomes more abundant, customers have gradually become research-oriented customers who no longer obtain product information passively, but actively choose products according to their own needs and considerations. While purchasing products, research-oriented customers gain a comprehensive understanding of the product by examining the purchase behaviors and evaluations of customers who have already purchased the products, comments from opinion leaders, product logistics, after-sales returns, and replacement services. If customers collect useful information for their purchase decisions, they not only determine the real quality of those goods, reducing the risk associated with online shopping, but they also save time [53]. For online retailers, the results of this study indicate that the type of goods online retailers sell has a considerable impact on final sales. For search-oriented products, customers can collect a great deal of information about those products on the internet, and have a full understanding of the products and the appropriate psychological expectations, before buying them. Since the conversion cost for products purchased by customers on different shopping websites is very low, it is very important for shopping websites that offer search goods to reduce product prices, improve their service, and guarantee their quality [54].

5.3 Limitations and future directions

Due to data availability, although this study made several contributions, it still had some research limitations. Follow-up studies can be promoted and perfected in the following ways. First, we conducted this study in June 2017 and December 2017, and we utilized web crawling technology to collect data from two product categories on the Amazon website. However, due to the limitation of the data collection period, the sales ranking changes in the two categories were not obvious. In prior studies, Chevalier and Mayzlin (2006) collected four years of data, from 1998 to 2002 [10], while Duan et al. (2008) collected one and a half years of data [13]. Therefore, this study does not reflect the long-term impacts some variables had on sales, although future studies can gather measurements over longer periods in order to improve the statistical significance of certain factors. Based on prior studies, we selected the representative category of mobile phones for the main effect in the study, and tested the regression result’s robustness via the laptop category. However, due to the limited representativeness of those two categories, we cannot apply the research results of this study to all search goods; therefore, future research must expand on these research categories. In addition, there were no direct sales data on the Amazon.cn website. Although prior studies have revealed a positive correlation between sales rankings and actual sales, this study replaced sales volume with sales ranking [25]. However, constrained by Amazon’s ranking lists for sales, each category only included the ranking of the top 100 products. Future studies should try to conduct research with more numerous samples that include actual sales data, such as film box office returns and lottery sales. Finally, future studies may validate our findings by examining Amazon’s competitors, and arrive at interesting results.

Footnotes

Appendix

Sample of data records structure

Acknowledgments

This study was supported by the Fundamental Research Funds for the Central Universities of Beijing University of Posts and Telecommunications (Grant No: 2020RC31 and 500419218) and the Social Science Program for Chinese Ministry of Industry and Information Technology (Grant No. 2019R22).

References

Babin

, Lee

, Kim

. Modeling Consumer Satisfaction and Word-of-mouth: Restaurant Patronage in Korea, Journal of Services Marketing (2005;19(3):133–9.

Bei

, Chen

, Widdows

. Consumers’ Online Information Search Behavior and the Phenomenon of Search vs. Experience Products. Journal of Family and Economic Issues. 2004;25(4):449–67.

Berger

, Sorensen

, Rasmussen

. Positive Effects of Negative Publicity: When Negative Reviews Increase Sales. Marketing Science. 2010;29(5):815–27.

Boadi

, et al.Customer Dissatisfaction and Unfavorable Word of Mouth. Human Systems Management. 2018;37(4):445–51.

Boadi

, Li

, Philip

. Customer Satisfaction, Co-creation Intentions and Diligent Business Response, Human Systems Management (2017;36(3):251–7.

Chatterjee

. Online review: Do Consumers Use Them? Advances in Consumer Research. Advances in Consumer Research. 2001;28:129–33.

Chatterjee

, Ravid

. How Critical Are Critical Reviews? The Box Office Effects of Film Critics, Star Power and Budget. Journal of Marketing. 2003;67(4):103–17.

Chen

, Wang

, Xie

. Online Social Interactions: A Natural Experiment on Word of Mouth Versus Observational Learning. Journal of Marketing Research. 2010;48(2):238–54.

Chen

, Xie

. Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix. Management Science. 2008;54(3):iv–626.

10.

Chevalier

, Mayzlin

. The Effect of Word of Mouth on Sales: Online Book Reviews[J]. Journal of Marketing Research. 2006;43(3):345–54.

11.

Clemons

, Gao

, Hitt

. When Online Reviews Meet Hyper differentiation: A Study of the Craft Beer Industry. Journal of Management Information Systems. 2006;23:149–71.

12.

Dellarocas

, Zhang

, Awad

. Exploring the Value of Online Product Reviews in Forecasting Sales: The Case of Motion Pictures. Journal of Interactive Marketing. 2007;21:23–45.

13.

Duan

, Gu

, Whinston

. Do Online Reviews matter?——An Empirical Investigation of Panel Data. Decision Support Systems. 2008;45:1007–16.

14.

Indārs

, Savin

, Lublóy

. Herding behaviour in an emerging market: Evidence from the Moscow Exchange. Emerging Markets Review. 2019;38:468–87.

15.

Jiménez

, Mendoza

. Too Popular to Ignore: The Influence of Online Reviews on Purchase Intentions of Search and Experience Products. Journal of Interactive Marketing. 2013;27(3):226–35.

16.

Forman

, Ghose

, Wiesenfeld

. Examining the Role of Relationship Between Reviews and Sales: The Role of Reviewer Identity Disclosure in Electronic Markets. Information Systems Research. 2008;19:291–313.

17.

Gasimli

, et al., The Informational Role of Average Rating and Variance of Customer Ratings in the Differential Patterns of Consumer Behavior. Human Systems Management. 2020;39(1):1–10.

18.

Godes

, Mayzlin

. Using Online Conversations to Study Word of Mouth Communication. Marketing Science. 2004;23(4):545–60.

19.

Godes

, Mayzlin

, Chen

, et al.The Firm’s Management of Social Interactions. Marketing Letter. 2005;(5):23–26.

20.

Godes

, Mayzlin

. Firm-Created Word-of-Mouth Communication: Evidence from a Field Test, Marketing Science. 2009;28(4):617–808.

21.

Vickery

, Wunsch-Vincent

. ParticipativeWeb and User-Created Content: Web 2. 0, Wikis and Social Networking. Paris: Organization for Economic Cooperation and Development (OECD), Paris, 2007.

22.

Lin

H-C

, Bruning

, Swarna

. Using online opinion leaders to promote the hedonic and utilitarian value of products and services. Business Horizons. 2018;61(3):431–42.

23.

Hsu

L-L

, Chen

S-J

, Chiu

M-C

, Chen

JCH

. “Determinants of successful online transactions-effects of transaction assurance seal and reputation rating affecting trust and purchase intention of consumers”. Human Systems Management. 2015;34(2):105–18.

24.

, Liu

, Zhang . Do Online Reviews Affect Product Sales? The Role of Reviewer Characteristics and Temporal Effects. Information Technology and Management. 2008;9(3):201–14.

25.

, Pavlou

, Zhang

. Can Online Word-of-Mouth Communication Reveal True Product Quality? Experimental Insights, Econometric Results, and Analytical Modeling. 7th ACM conference on Electronic commerce. 2006;324-30.

26.

, Li

, Xu

, Fujita

. Construction of a technology adoption decision-making model and its extension to understanding herd behavior. Knowledge-Based Systems. 2015;89:471–86.

27.

Kumar

, Benbasat

. The Influence of Recommendations and Consumer Reviews on Evaluations of Websites. Information Systems Research. 2006;17:425–39.

28.

Kusumawati

, Sri Rahayu

. The effect of experience quality on customer perceived value and customer satisfaction and its impact on customer loyalty. Human Systems Management. 2020;39(2):219–32.

29.

, Zhang

, Wang

, Su

. Predicting the sequential behavior of mobile Internet users based on MSM model. International Journal of Market Research. 2019;1-15.

30.

Liu

, Su

, Akram

, Rasool

. Tourism Apps Women Prefer and Their Tipping Point[J]. Human Systems Management. 2019;38(3):209–20.

31.

Liu

, Su

, Akram

, Abrar

. The User Acceptance Behavior to Mobile Digital Library. International Journal of Enterprise Information Systems. 2020;16(2):38–53.

32.

Liu

. Word of Mouth for Movies: Its Dynamics and Impact on Box Office Revenue. Journal of Marketing. 2006;70:74–89.

33.

Luqiu

, Schmierbach

, Ng

Y-L

. Willingness to follow opinion leaders: A case study of Chinese Weibo. Computers in Human Behavior. 2019;101:42–50.

34.

Lynch

, Ariely

. Wine Online: Search Costs Affect Competition on Price, Quality, and Distribution. Marketing Science. 2000;19:83–103.

35.

Mayzlin

. Promotional Chat on the Internet. Marketing Science. 2006;25:155–63.

36.

Qiu

, Su

, Wang

. Factors influencing commercial buildings to obtain green certificates. Applied Economics. 2016;49(20):1937–49.

37.

Schlosser

. Posting Versus Lurking: Communicating in a Multiple Audience Context. Journal of Consumer Research. 2005;32:260–5.

38.

Seyed

, Hosseini

, Ildar

, Qu

. Opinion leader detection: A methodological review. Expert Systems with Applications. 2019;115:200–22.

39.

Shim

, Lee

. Internet Portals’ Strategic Utilization of UCC and Web 2.0 Ecology. Decision Support System. 2009;(47):415–23.

40.

, Gao

, He

, Zhu

. Effect of Leader’s Implicit Followership Prototypes on Employee’s Internal and External Marketability. Social Behavior and Personality. 2019;47(12):e8470.

41.

, Yu

. Research on early warning algorithm for economic management based on Lagrangian fractional calculus. Chaos, Solitons and Fractals. 2019;128:44–50.

42.

Sun

. A longitudinal study of herd behavior in the adoption and continued use of technology. Mis Quarterly. 2013;37(4):1013–41.

43.

Susan

, Schuff

. Research Note: What Makes a Helpful Online Review? A Study of Customer Reviews on Amazon.com. MIS Quarterly. 2010;185-200.

44.

Villanueva

, Yoo

, Hanssens

. The impact of marketing-induced versus word-of-mouth consumer acquisition on consumer equity growth. Journal of Marketing Research. 2013;45(1):48–59.

45.

Wang

, Li

, Zhang

, Su

. How impacting factors affect Chinese green purchasing behavior based on Fuzzy Cognitive Maps. Journal of Cleaner Production. 2019;240.

46.

Wang

, Su

, Wang

, Zou

. Intellectual Capital and Technological Dynamic Capability: Evidence from Chinese Enterprises. Journal of Intellectual Capital. 2019;20(4):453–71.

47.

Wang

, Zeng

. Toward a service-innovation ecosystem of enterprises in China. Human Systems Management. 2019;38(3):279–95.

48.

Warner

, Zhao

. Knowledge transfer, indigenization and human-development resources management in China. Human Systems Management. 2018;37(2):151–60.

49.

Welbourne

, Blanchard

, Wadsworth

. Motivations in virtual health communities and their relationship to community, connectedness and stress. Computers in Human Behavior. 2013;29(1):129–39.

50.

Wen

, et al.Customer-knowledge Enabled Innovation: An Empirical Assessment. Human Systems Management. 2019;38(2):179–96.

51.

Yasmeen

, et al.Service-innovation Capability Founded on Knowledge from Customers. Human Systems Management. 2019;38(1):29–41.

52.

Zameer

, et al.Corporate Image and Customer Satisfaction by Virtue of Employee Engagement. Human Systems Management. 2018;37(2):233–48.

53.

Zhou

, Liu

, Su

, Xu

. Gossip fiercer than a tiger: Effect of workplace negative gossip on targeted employees’ innovative behavior. Social Behavior and Personality. 2019;47(5).

54.

Zhu

, Zhang

. Impact of Online Consumer Reviews on Sales: The Moderating Role of Product and Consumer Characteristics. Journal of Marketing. 2010;74:133–48.