Do online reviews have different effects on consumers’ sampling behaviour across product types? Evidence from the software industry

Abstract

Previous research shows that online reviews may have different effects for search goods and experience goods. However, as a typical type of experience goods, software can be further divided into different categories based on product characteristics. Little research has been conducted regarding the different effects of online reviews for different types of software. Furthermore, to offer free samples is another common practice of software firms to alleviate consumer uncertainty prior to purchase. To fill the corresponding research gap, this research focuses on the interaction effects between online reviews and free samples for different types of software. Through our empirical analysis, we find that user ratings significantly increase consumers’ sample downloads. Furthermore, consumers download more samples for some categories than for others. Finally, user and editor ratings might have differential effects for different types of software.

Keywords

Free samples online reviews product characteristics product types

Introduction

Software is a type of experience goods whose product attributes are difficult for consumers to evaluate prior to purchase [1]. Since consumers are uncertain about experience goods, they usually tend to rely heavily on others’ (which may include other users or experts) reviews when making purchase decisions [2 –4]. Internet technology facilitates this process through online review websites, which allow consumers to post and publish their comments and may also recruit experts to publish their reviews of a particular product or service. Various review websites provide online reviews for different types of businesses, among which CNET.com is a major website that focuses on providing online reviews for Information Technology (IT) products such as software. Investigations by both industry and academics show that consumers’ adoptions of software are largely impacted by online reviews. For example, 92% of the top 100 iOS app downloads have a rating of three stars or better [5]. Furthermore, several scholars also verified the importance of online reviews on software adoption through both empirical and theoretical research [6 –9].

Offering free samples is an alternative strategy of software firms to reduce consumer uncertainty before purchase. Software firms may offer different forms of free samples, including time-limited free samples [10,11], functionality-limited free samples [12,13] or mixed free samples [14], which are combinations of the previous two. Using free samples, consumers can learn software functionalities and characteristics through their own experience before making a purchase decision.

To summarise, consumers may resort to two different information channels before they make a purchase decision on software: online reviews or free samples. Most commonly, both information channels are simultaneously available to software consumers. For example, on CNET.com, consumers can download free samples and refer to other users’ and editors’ aggregated ratings of software. It therefore would be interesting to investigate how one information channel affects the other. That is, how do user and editor ratings affect consumer downloads of free samples? Such an investigation would help to reveal the mechanism of the interaction between the two information channels when both are available to consumers before purchase.

Furthermore, software can be divided into various categories, and each category may serve different aspects of consumer needs. For example, CNET divided all software into 22 categories, including securities, business software and games. Software from different categories may have different characteristics. Therefore, it will be interesting to explore how different software categories affect the relationship between online reviews and free samples.

The previous literature has examined the effects of online reviews for different product types. Most of these studies follow the framework by Nelson [1], which categorised products as search and experience goods. In comparison to experience goods, search goods are defined as those whose product attributes can be determined prior to purchase. Relevant study found that the effect of online reviews is stronger for experience goods than for search goods because the product information on experience goods is more difficult to obtain than that of search goods [15 –17].

However, little research has been conducted to explore the different effects of online reviews of experience goods that involve differential product characteristics. As mentioned above, although software is a type of experience goods, it includes many different forms of applications and can be further categorised. Our research aims to fill this research gap by exploring the interaction between online reviews and free samples for different types of software.

In summary, this article aims to address the following research questions:

What is the overall effect of online reviews (including user and editor ratings) on consumer downloads of free software samples?

Do consumers show different sample downloading tendencies across different software categories?

Do online reviews have different effects on consumers’ sample downloads across various software categories?

Based on consumers’ purpose of use, we classified software into four categories: Platform-related, Entertainment-related, Work-related and Decoration-related. Software in the Platform-related category includes software necessary for setting up or operating a computer environment, including operating systems, network applications and security software. The primary purpose of using Entertainment-related software is for leisure and relaxation in daily life, including games, music and video players. Work-related software is mainly adopted for professional use such as business software and developer tools. There is also software whose usage is mainly for the purpose of ‘decoration’, including screensavers, customised desktops and photo editors. Such software is defined as Decoration-related software in this article. More importantly, the different categories of software may have different characteristics to different extents, including network externalities, customization and innovation, which will be elaborated in detail in section ‘Theoretical framework and hypothesis development’.

This study makes several contributions to the literature on online ratings and free sampling. On the theoretical front, our research suggests that the software type moderates the impacts of user and editor ratings on consumers’ sample downloads. To the best of our knowledge, we are the first to categorise software products into more detailed types and examine the moderating role of software types in the relationship between online reviews and consumers’ sample downloads. On the practical front, our research might offer important guidelines for software firms. For example, they may need to adopt different strategies to utilize online reviews to induce more downloads for different categories of software. Analogously, this categorization effect may extend to the contexts of the relationship between mobile app ratings and their sample downloads.

The rest of the article proceeds as follows. In section ‘Theoretical framework and hypothesis development’, we discuss the theoretical foundations and develop our hypothesis. Section ‘Methodology’ presents the empirical models. Section ‘Results’ presents the research results. In section ‘Discussion’, we discuss the empirical findings and practical implications. Section ‘Conclusion’ concludes the article.

Theoretical framework and hypothesis development

Research overview

Software users may obtain product information prior to purchase mainly through two channels: online reviews and free samples. First, consumers can collect others’ opinions and evaluations of software, which will help them to form their own decisions. Second, consumers can learn about software attributes based on their own experiences using free samples before purchase. It thus will be interesting to investigate how one information channel influences the other information channel or how online reviews of software affect consumer adoption of free samples. Furthermore, online reviews may include user ratings and expert ratings. It is also unclear whether reviews from peers (other users) or experts have different effects on consumer adoption of free samples when both are available. Our research will examine the relationship between software ratings (user versus editor ratings) and consumer sample downloads (H1a and H1b).

Software is a wide-ranging product that can be classified into different categories. Software from different categories serves different aspects of users’ needs. It is thus possible that consumers’ intentions to use free samples are different across categories due to consumers’ different demands for software. This categorization effect will be examined by H2.

More importantly, it is meaningful to explore the interaction effect between online reviews and the categorization effect; that is, how user and editor ratings influence consumers’ sample downloads might be different for different software categories. For example, relying on lab experiments and professional knowledge, editor ratings can more effectively transfer product information for products that are frequently updated using cutting-edge innovations, such as Entertainment-related software, than for other products. However, for software that features customised tastes, such as Decoration-related software, other users’ opinions may not be considered an important reference by consumers. We study this moderating effect of the software category on the relationship between software ratings and sample downloads by testing H3a and H3b. The conceptual model is shown in Figure 1.

Figure 1.

Research framework.

Effect of user and expert ratings

Numerous previous studies have verified the influence of online reviews on consumers’ adoption decisions. For example, Dellarocas et al. [18] showed that user ratings have a significant influence on movie revenues. Cui et al. [19] revealed the positive effect of user ratings on new product sales. Li et al. [20] established the influence of both positive and negative reviews on product sales. Ham et al. [21] moved one step further and examined how online reviews influence consumer decision-making in different ways.

As mentioned above, software is a type of experience goods, and thus, software users are urged to obtain production information prior to purchase [1,22]. Using online reviews is one easily accessible information channel for consumers. Several scholars have empirically shown that software users’ adoption decisions are influenced by online reviews [9,23,24].

In addition to online ratings, free samples serve to provide consumers with product information. Previous research has shown that online reviews and free samples might interact with each other and influence product sales [25]. In this article, we are interested in investigating how one information channel influences the other. One noteworthy difference between these two information channels lies in their associated costs. On one hand, software ratings, as a direct and accessible channel that delivers information, require only very limited efforts from consumers to read. On the other hand, free samples require consumers to make additional efforts to download, instal and learn to acquire information. Due to these costs that are associated with using a free sample, it is reasonable to conjecture that if a consumer perceives a low expected evaluation from software ratings, then she will not be willing to download a free sample.

In addition, software ratings typically involve user ratings and editor ratings. They may have different effects on consumers [26,27]. Although editor ratings can provide comprehensive and professional information through lab tests and technical expertise [28,29], user ratings may be more trustworthy and suitable for current consumers, since they represent previous users’ own experiences. Previous studies found that the role of editor ratings might be weakened for various reasons. For example, consumers may suspect whether an editor is indeed knowledgeable and objective [30]. Lee and Tan [31] found that experts usually select software to review based on their sponsorship or partnership with software developers or review platforms, resulting in consumer concerns regarding the integrity and credibility of editor ratings.

In summary, higher software ratings could raise consumer willingness to purchase the software, thereby incentivizing consumers to download a free sample to further learn about the product; however, this impact may vary based on the rating sources and depending on the credibility of the released information. Based on the previous discussion, we propose the following hypotheses:

H1a. User ratings have a positive impact on consumers’ sample downloading behaviour.

H1b. Editor ratings do not have a significant impact on consumers’ sample downloading behaviour.

Software categories

Previous studies have empirically shown that consumers’ purchase intentions might be different between search goods and experience goods [15 –17]. In line with these studies, several scholars consider different product categories. For example, Lian and Lin [32] classified products by three dimensions: cost and purchase frequency, value proposition and the degree of differentiation. Pascual-Miguel et al. [33] divided products into digital and non-digital goods.

Although software is considered a digital and experience good, it includes so many different applications that it can be further classified into several more categories. Our research defined four categories of software, including Platform-related, Entertainment-related, Work-related and Decoration-related, which reflect consumers’ different needs for software.

It is straightforward to conjecture that software users’ intentions to adopt are different across these four categories. First, almost all consumers need to instal Platform-related software to set up and maintain their computer environments, regardless of their occupations, lifestyles and tastes. Second, Entertainment-related software, such as games and music players, is very frequently used in people’s daily lives. Software in these two categories thus would generate large demands. Third, Work-related software (e.g. statistics software and CRM software) is mainly used by consumers with specific professions. Finally, Decoration-related software (e.g. photo editors and desktop themes) is also mainly designed for particular uses. Software that belongs to these last two categories thus has relatively small demands. In addition, Platform-related software guarantees the fundamental operation environment for the usage of desktop or mobile devices. Thus, consumers make their decision more seriously and are more probably to download samples for testing purposes.

Consumers do not download a software sample that they have no need to use. Therefore, the number of consumers’ sample downloads should be proportional to their demand for a product. Based on this discussion, we thus test the following hypothesis:

H2. Compared with the Platform-related category, consumers download fewer software samples in the Decoration-related and Work-related categories and a similar number of samples in the Entertainment-related category.

The interaction between software ratings and product categories

The product category can reflect the different product characteristics. Most previous research categorises products as search or experience goods and examines the impact of their different product characteristics on the effects of online reviews [15 –17]. Several scholars provided a different perspective on product categorization. For example, Cheema and Papatla [34] examined how online information sources differently influence utilitarian and hedonic goods. In their research, utilitarian goods are characterised as those whose consumption is mainly for functional or practical purposes, and hedonic goods are those whose consumption is mainly for relaxation and sensual pleasure. Few studies have explored the differential effects of online reviews within one particular product such as software. One of the few exceptions is Zhu and Zhang [35]. Their findings indicate that online reviews might have different effects for products with different characteristics.

These previous studies largely inspired our research. Although software products are usually considered experience goods, different software applications have various characteristics. Based on the purpose of consumer use, we further divided software into four categories: Worked-related software, Decoration-related software, Platform-related and Entertainment-related software. Among these categories, Worked-related software and Platform-related software are used more for utilitarian purposes, Entertainment-related software is used more for hedonic purposes, and Decoration-related software can be used for both purposes. These different purposes are related to different product characteristics that can affect the underlying mechanism of consumer perception of the online rating in terms of affecting their purchase decisions. The purchase intentions in turn influence their sample downloading behaviour.

First, software products usually exhibit network effects or network externality, which leads to the scenario that users have stronger purchase intentions for a software product when the total number of users of this software increases [12,36 –38]. Different types of software will exhibit different levels of network externalities [39,40]. Work-related software would exhibit strong network effects, since its clients are mainly companies, governments and universities, who tend to purchase software from established makers and have large user bases in the same industry. Moreover, compatibility is an important issue here. For example, one consulting firm wants its business reports to be compatible with its clients. As a result, users’ purchase decisions would be largely influenced by network effects, reducing the effects of online reviews.

Second, consumer expectations for some software are highly customised based on their different preferences and tastes [41]. Among the four categories, Decoration-related software largely exhibits these characteristics. Consumers usually purchase desktop themes or photo editors based on their heterogeneous tastes. Accordingly, compared with Platform-related software, which is a category with relatively homogeneous evaluation standards, the users’ ratings for Decoration-related software have less impact on users’ own purchase intentions.

Moreover, in terms of network externalities and customised tastes, there is no significant difference between Platform-related software and Entertainment-related software. As a result, the positive impacts of user ratings on consumers’ sample downloads would be similar for these two categories.

In addition, some software features proliferated innovations and frequent updates [42]. For example, each year, numerous new games will be released, and existing games will publish updates. For these Entertainment-related software products, experts’ opinions, compared with previous consumer experiences, can keep pace with the market trend and the development of latest version of the software, leading to timely, relevant and comprehensive judgements of the software. As a result, editor ratings can help consumers evaluate the brand-new features of this software. On one hand, due to the lack of credibility, higher editor ratings may not lead to higher purchase intentions for Platform-related, Work-related or Decoration-related software. On the other hand, by relying on lab tests and professional knowledge, editor ratings may generate a more positive impact on consumers’ purchase intentions for Entertainment-related software than for other categories of software.

To have a better understanding of how product characteristics (according to the different software categories) are integrated with the online ratings from different sources in terms of affecting consumers’ sample downloads, we propose the following hypotheses:

H3a. Compared with Platform-related software, the positive impact of user ratings on sample downloads would be weaker for Work-related and Decoration-related software and similar for Entertainment-related software.

H3b. Editor ratings have no significant impact on consumers’ sample downloads of Platform-related, Work-related or Decoration-related software but will have a significantly positive effect on consumers’ sample downloads of Entertainment-related software.

Methodology

The data set of our empirical analysis is from CNET.com, which is an American website that provides reviews, news, articles, blogs, podcasts and videos on technology and consumer electronics. The download section offers free and legal software sample downloads by category. Each sample on CNET.com has an associated webpage that displays information on its corresponding software such as the title, price, file size, user ratings, editor ratings, sample licence and operating system. Thus, before downloading a software sample, a consumer can learn about the product information by reviewing the software ratings.

We collect data using a Python-based crawler from the download section of the website in February 2018. Overall, our data set includes 5453 observations of software on CNET.com. On its website, CNET has divided all software into 21 categories, including drivers, browsers and business software. We adopt a design of four categories rather than the 21 categories provided by CNET for the following reasons. First, if we divide software into 21 categories, the numbers of observations in some categories will be too small to offer meaningful results. Second, software from different categories (from the 21 categories by CNET) may share a high degree of similarities. For example, software products in Games and Entertainment Software categories both serve consumers’ need for entertainment.

Therefore, we divided software into four categories, based on consumers’ purpose of use. The Platform-related category corresponds to software that is necessary for setting up or operating a computer environment. Entertainment-related software is mainly used for leisure and relaxation in daily life. Work-related software includes software that is mainly adopted for professional use. Decoration-related software corresponds to software whose usage is mainly for the purpose of ‘decoration’. Table 1 presents the details of each category of software.

Table 1.

Design of software categories.¹

Software categorization	Corresponding types of software by CNET
Platform-related	Utilities and operating systems, drivers, Internet software, browsers and security software
Work-related	Educational software, business software, productivity software and developer tools
Entertainment-related	Games, entertainment software, travel, home software, mp3 and audio software, communications and video software
Decoration-related	Digital photo software, desktop enhancements, graphic design software, screensavers and wallpaper

To test hypotheses, such as H1 (the impact of online ratings on consumers’ sample downloads) and H2 (the impact of the software category on consumers’ sample downloads), we examine the following

\begin{array}{l} D o w n l o a d_{i} = a_{0} + a_{1} E d i t o r r a t i n g_{i} + a_{2} U s e r r a t i n g_{i} \\ + a_{3} D e c o r a t i o n_{i} + a_{4} E n t e r t a i n m e n t_{i} + a_{5} W o r k_{i} \\ + a_{6} F r e e_{i} + a_{7} F r e e - t o - t r y_{i} + a_{8} S i z e_{i} \\ + a_{9} P r i c e_{i} + a_{10} D e s k t o p_{i} + α_{11} Y e a r_{i} + ε_{i} \end{array}

(1)

where the dependent variable download_i is the number of sample downloads for software i in the last week. As the total number of downloads increases over time on CNET.com, a longer release period will probably lead to more downloads. Thus, we use the number of sample downloads for a 1-week period as the dependent variable to avoid this effect. The independent variables include the software ratings from two different sources (editor rating_i and User rating_i representing the degree of satisfaction with software i on a scale from 0 to 5, and the software category dummies ( $Decoractio n_{i}$ , $Entertainmen t_{i}$ and $Wor k_{i}$ ), which take a value of 1 if software i belongs to the corresponding category and 0 otherwise. Since the Platform-related category is omitted from the regression, the coefficients of the other three categories are relative values compared with it. In addition, we include the following control variables: $Fre e_{i}$ and $Free - to - tr y_{i}$ are dummies of the sample licence model for software i,² $Siz e_{i}$ represents the file size of software i, $Pric e_{i}$ is the associated price of software $i,$ and $Deskto p_{i}$ is the operating system dummy. $Deskto p_{i}$ equals to 1 for Windows or Mac (desktop) software or equals to 0 for Android or iOS (mobile) software. The time dummies $Yea r_{i}$ are also included to indicate the year in which the software sample was released.

To test H3 (the moderating effect of product characteristics), we examine the following

\begin{matrix} Downloa d_{i} = α_{0} + α_{1} Editor ratin g_{i} + α_{2} User ratin g_{i} + α_{3} Decoratio n_{i} + \\ α_{4} Entertainmen t_{i} + α_{5} Wor k_{i} + β_{1} Editor ratin g_{i} \times Decoratio n_{i} + \\ β_{2} Editor ratin g_{i} \times Entertainmen t_{i} + β_{3} Editor ratin g_{i} \times Wor k_{i} + \\ β_{4} User ratin g_{i} \times Decoratio n_{i} + β_{5} User ratin g_{i} \times Entertainmen t_{i} + \\ β_{6} User ratin g_{i} \times Wor k_{i} + a_{6} Fre e_{i} + a_{7} Free - to - tr y_{i} + a_{8} Siz e_{i} + \\ a_{9} Pric e_{i} + a_{10} Deskto p_{i} + α_{11} Yea r_{i} ε_{i} \end{matrix}

(2)

The estimations of $β_{1}, β_{2}, β_{3}, β_{4}, β_{5} and β_{6}$ are used to evaluate H3, which investigates whether the impact of software ratings on consumers’ sample downloads depends on product categories. Since the Platform-related category and its interaction terms with user and editor ratings are omitted from the regression, the coefficients of $Editor ratin g_{i}$ and $Userratin g_{i}$ represent the rating effects on sample downloads for Platform-related software; the coefficients of $Editor ratin g_{i} \times Decoratio n_{i}$ and $User ratin g_{i} \times Decoratio n_{i}$ represent the relative rating effects for Decoration-related software compared with Platform-related software; the coefficients of $Editor ratin g_{i} \times Entertainmen t_{i}$ and $User ratin g_{i} \times Entertainmen t_{i}$ represent the relative rating effects for Entertainment-related software compared with Platform-related software; and the coefficients of $Editor ratin g_{i} \times Wor k_{i}$ and $User ratin g_{i} \times Wor k_{i}$ represent the relative rating effects for Work-related software compared with Platform-related software. IT product characteristics (externalities, customization and innovation) differ across these four categories. Therefore, their moderating roles on the effect of online reviews can be captured by the coefficients of the category interaction terms.

Results

The descriptive statistics are summarised in Table 2, including the means and standard deviations for the main variables and their Pearson correlation coefficients. As shown in Table 2, the average editor rating is 3.73 and the average user rating is 3.49. The average decoration dummy is 0.20, showing that approximately 20% of software products belong to the Decoration-related category. The average of the entertainment dummy is 0.23; thus, approximately 23% of software belongs to the Entertainment-related category. The average of the institution-used dummy is 0.17, indicating that approximately 17% of software belongs to the Work-related category. The remaining (approximately 41%) products are thus Platform-related software. The average free licence dummy is 0.55, showing that approximately 55% of the software products release a sample with a free licence. The average free-to-try licence dummy is 0.44, which indicates that approximately 44% of the software products release a sample with a free-to-try licence. The remaining software products (approximately 1%) thus release a sample with a purchase licence. The average desktop dummy is 0.98, indicating that 98% of the software is used in Windows or Mac systems.

Table 2.

Descriptive statistics.

	Mean(SD)	Downloads	Editor rating	User rating	Decoration	Entertainment	Work	Free licence	Free-to-trylicence	File size	Price	Desktop
Downloads	369.51	1
	(4936.61)
Editor rating	3.73	0.01	1
	(0.76)
User rating	3.49	0.04***	0.13***	1
	(1.10)
Decoration	0.20	−0.02	−0.01	0.02	1
	(0.40)
Entertainment	0.23	−0.00	−0.01	0.01	−0.27***	1
	(0.42)
Work	0.41	−0.02	0.07***	0.01	−0.22***	−0.24***	1
	(0.37)
Free licence	0.55	0.04	0.03*	−0.01	−0.01	−0.12***	−0.04***	1
	(0.50)
Free-to-try licence	0.44	−0.04	−0.03**	0.01	0.01	0.11***	0.04***	−0.98***	1
	(0.50)
File size	16.74	0.01	0.05***	0.03**	−0.02*	0.09***	0.01	−0.07***	0.05***	1
	(89.59)
Price	17.25	−0.01	0.04***	0.03**	−0.01	−0.01	0.10***	−0.32***	0.31***	0.12***	1
	(59.30)
Desktop	0.98	0.00	−0.03**	−0.07***	0.06***	−0.10***	−0.02	−0.12***	0.13***	0.00	0.04***	1
	(0.15)

SD: standard deviation.

*p< 0.10, **p < 0.05 and ***p < 0.01.

The overall impacts of online ratings and the software categories

The regression results of equation (1) that are summarised in Table 3 offer several interesting findings. First, the coefficient of $Userratin g_{i}$ is significantly positive, while the coefficient of $Editorratin g_{i}$ is insignificant. These results suggest that on average, user ratings positively affect consumers’ sample downloading behaviour but editor ratings do not. H1a and H1b are supported. Therefore, the influence of one product information channel, however, is clearly revealed. On CNET.com, editor ratings are based on the CNET Labs’ testing method, which employs the expertise of CNET Labs’ staff and editors. In addition, CNET.com reviews only selected software products that enter their top 100 most popular rankings or new products. Thus, due to experts’ partnerships with the website and the underlying product selection process, editor ratings might be biased and not credible. Accordingly, editor ratings fail to affect consumers’ purchase decisions, especially when user ratings, which are a more credible source of product information, are available. In contrast, higher user ratings can improve consumers’ evaluations of software and their willingness to purchase and thus motivate consumers to download a free sample to further learn about the product information.

Table 3.

Basic regression.

Variables	Downloads
Editor rating	5.2592
	(49.8500)
User rating	102.9509***
	(28.9420)
Decoration	−337.8711**
	(141.0413)
Entertainment	−182.4322
	(172.0586)
Work	−499.5669***
	(158.9097)
Free licence	712.2494***
	(223.7052)
Free-to-try licence	369.4224*
	(209.8522)
File size	−0.1055
	(0.4279)
Price	−0.1434
	(0.1595)
Desktop	660.6446***
	(223.7528)
Released year dummies	Yes
Observations	5454
R-squared	0.0283

^*p < 0.1, **p < 0.05 and ***p < 0.01.

Second, the coefficients of the Decoration-related and Work-related category dummies are significantly negative, while the coefficient of the Entertainment-related category is insignificant. These results suggest that compared with Platform-related software, Decoration-related and Work-related software has fewer sample downloads, while Entertainment-related software has a similar number of sample downloads. A majority of CNET.com’s sample downloads are Work-related software, accounting for 41%. The other three categories (Decoration-related, Entertainment-related and Platform-related) account for 20%, 23% and 16% of the total number of software sample downloads at CNET.com, respectively. Thus, the impact of the software category on sample downloads is not driven by whether a certain category widely releases samples on the website. Instead, the impact may be explained by consumers’ differential demand for various categories of software. Decoration-related and Work-related softwares are usually used for particular tasks for specific occupations, Platform-related software is required for running an operating system, and Entertainment-related software is used in daily life. There are significant differences among the product demands, resulting in different demands for corresponding samples. H2 is supported.

Moreover, the coefficients of free licences and free-to-try licences are significantly positive, suggesting that compared with purchase licences, free licences and free-to-try licences lead to more sample downloads. As a sample discloses more information, consumers are more willing to download it to learn more about the product. The desktop dummy coefficient is significantly positive, implying that software samples for Windows and Mac systems lead to more downloads than do samples for Android and iOS systems. Desktop software (Windows and Mac systems) is generally more complicated than similar mobile software (Android and iOS systems); thus, samples of the former are more necessary for consumers to acquire further information. The coefficients of price and file size are insignificant, implying that software price and size do not influence consumers’ sample downloading behaviours.

The moderating effect of product categories on the relationship between software ratings and sample downloads

H3a and H3b, which investigate how product categories moderate the impact of software ratings on consumers’ sample downloads, are tested separately using equation (2) (Table 4). The coefficient of $Userratin g_{i}$ is significantly positive, suggesting that user ratings positively affect the sample downloads of Platform-related software. The coefficient of the interaction term $Userratin g_{i} \times Entertainmen t_{i}$ is insignificant, indicating that such rating effects are not different between Platform-related software and Entertainment-related software. In contrast, the coefficients of the interaction terms $Userratin g_{i} \times Decoratio n_{i}$ and $Userratin g_{i} \times Wor k_{i}$ are significantly negative, suggesting that the effects of user ratings on sample downloads are weaker for Decoration-related/Worker-related software than for Platform-related software.

Table 4.

Regression with the moderating effect.

Variables	Downloads
Editor rating	−48.459
	(88.241)
User rating	229.39***
	(86.177)
Decoration	459.66
	(438.44)
Entertainment	−721.22
	(689.77)
Work	588.18
	(452.13)
Editor rating × decoration	23.458
	(97.347)
Editor rating × entertainment	274.57*
	(160.73)
Editor rating × work	−48.153
	(92.746)
User rating × decoration	−253.02**
	(108.11)
User rating × entertainment	−138.37
	(109.12)
User rating × work	−257.02**
	(108.25)
Free licence	729.97***
	(227.94)
Free-to-try licence	390.48*
	(215.41)
File size	−0.14741
	(0.43036)
Price	−0.067726
	(0.16256)
Desktop	694.39***
	(236.39)
Released year dummies	Yes
Observations	5454
R-squared	0.02813

^*p < 0.1, **p < 0.05 and ***p < 0.01.

In accounting for the moderating effect of product categories, we glean further insight into how consumers process product information. Although user ratings can provide credible information based on previous consumers’ experiences, they may lose their influence if the current consumers rarely rely on other consumer evaluations to make their own purchase decisions. Specifically, Decoration-related software is more subjectively evaluated by consumers than is Platform-related software because consumers can have distinct preferences, tastes and styles. Moreover, Work-related software is mainly used in organisations to accomplish job tasks. This software category features strong network externalities, leading to weaker influences of user ratings on consumers’ purchase decisions. As long as previous users’ evaluations of Decoration-related/Work-related software are less effective than are those of Platform-related software in terms of affecting current consumers’ own purchase decisions, the effect of user ratings on sample downloads would be weakened.

In addition, Table 4 shows that the coefficient of $Editorratin g_{i}$ is insignificant, suggesting that editor ratings fail to affect sample downloads for the Platform-related software. The coefficients of the interaction terms $Editorratin g_{i} \times Decoratio n_{i}$ and $Editorratin g_{i} \times Wor k_{i}$ are also insignificant, indicating that, similarly to Platform-related software, editor ratings do not significantly influence the sample downloads for Decoration-related/Work-related software. However, it is interesting to notice the significantly positive coefficient of the interaction term $Editorratin g_{i} \times Entertainmen t_{i}$ , indicating that compared with Platform-related software, editor ratings play a significantly positive role in the sample downloads for Entertainment-related software.

Unlike other categories (Platform-related, Decoration-related and Work-related software), Entertainment-related software turns out to be the only one whose sample downloads are affected by editor ratings. This moderating effect is driven by how editor ratings interact with product characteristics to affect consumers’ decision-making. Entertainment-related software features a highly vibrant market in which numerous software products are released with innovative designs and the latest technologies. Thus, experts can use the advantages that are related to lab tests and professional knowledge to deliver timely, relevant and systematic reviews that help consumers learn about the advanced technology and complex operations of newly released entertainment software. Accordingly, although editor ratings have no significant overall effect on sample downloads, they do affect sample downloads for Entertainment-related software.

Discussion

Our research leads to several important findings. We first find that user ratings significantly increase consumers’ sample downloads, indicating that consumers would first resort to other users’ evaluations of software and then decide whether to download the sample and try it themselves. The reason might be attributed to the fact that referring to other users’ ratings requires less effort than sampling software, which would include time and effort to download, instal and use the samples. As we have hypothesised, the influence of editor ratings is not significant in the basic regression model. The previous literature on the influence of editor ratings has mixed results. On one hand, some scholars found the influence of editor ratings to be insignificant [30,31]. On the other hand, Moon et al. [26] found that professional critics may have significant influences on movie revenues. In particular, Zhou et al. [27] found that professional reviews affect downloads positively, based on a data set drawn from CNET.com. One needs to note that, although our results from the basic regression indicate that the aggregate effect of editor ratings on sample downloading is insignificant with the effects of software category and licence type controlled (Table 3), we do find that editor ratings have a significantly positive influence on downloads of Entertainment-related software (Table 4). In other words, our result is consistent with Zhou et al. [27] for specific software categories. This result also demonstrates the importance of exploring the categorization effect on the relationship between ratings and sample downloads (H3).

Another key finding is that consumers would download more samples for some categories than for others. Specifically, consumers download more samples for Platform-related and Entertainment-related software than for the other two categories. The reason might be because consumers have more demands for the first two categories of software and thus are more incentivized to sample these types of software.

More importantly, we investigated the moderating effect of software categories. Moderation is commonly used in the management literature to evaluate the effect of one variable on the relationship between two other variables [43]. In particular, we interact the software category with online ratings to explore how software product characteristics, including externalities, customization and innovation, affect consumer acquisition of product information. The results suggest that user ratings have stronger effects on consumers’ sample downloads for Platform-related and Entertainment-related software than for Work-related and Decoration-related software. One plausible explanation is that Work-related software has stronger network externalities and that the purchase decisions for this software are more influenced by factors such as compatibility, leading to a weakened effect of user ratings. For Decoration-related software, consumers usually have their unique tastes when choosing software; thus, others’ ratings are insignificant. Furthermore, the previous literature shows that the influence of editor ratings on consumer downloading of samples is insignificant, despite the positive effect of user ratings [44]. However, although its overall effect is not significant, we find that editor ratings have a significantly positive effect on sample downloads for Entertainment-related software. The intuition behind this effect might be that Entertainment-related software features extensive innovation and frequent updates, which increases the challenge for consumers to follow the latest developments by themselves. Therefore, editors’ opinions are deemed an important reference.

Our research results offer important implications for relevant business practitioners. It is a common practice for firms to provide product information through user and editor ratings. Our findings first suggest that, in general, user ratings are important in influencing consumer downloads. Therefore, it would be a good strategy for software firms with high user satisfaction to allow previous users to post their first-hand experiences to induce more downloads. Furthermore, software firms should also identify in which category their products fit, which is equivalent to determining the intrinsic characteristics of the software, since software characteristics may interact with user ratings and affect sample downloads. In particular, user ratings’ influence is more effective for Platform-related and Entertainment-related software such as operating systems, security software and games.

Second, software firms should give extra attention to editor ratings. We find that the overall effect of editor ratings on sample downloads is not significant, which agrees with previous findings [44, 45]. However, concerning different categories of software, the story might be different. In particular, editor ratings have a significantly positive influence on downloads of Entertainment-related software. That is, software firms, such as game developers, might recruit experts to post reviews on their products to prompt consumer downloads.

Although an increase in consumers’ downloading of samples will expand the software market, such downloading might also lead to cannibalization of sales [10]. Therefore, to achieve optimal profits, software firms also need to balance the trade-off between more sample downloads and a stronger cannibalization effect, as pointed out in the previous literature [44].

Moreover, software category may affect sample downloads in a more complicated way than what this article presents. For example, consumers may not frequently change Platform-related software, in comparison with Decoration- or Work-related software. This behaviour might lead to two opposite effects on consumers’ sampling downloading behaviour. On one hand, consumers are more serious and are more probably to download samples before purchasing. On the other hand, reduced demand for new platform-related software may lead to fewer sample downloads. A more detailed analysis on this issue would be an interesting topic for future research.

Finally, without dynamic data, this article does not consider whether sample downloads in turn affect ratings and how user and expert ratings interact with one another. Relevant research issues are well studied in the previous literature [27]. However, these issues are thus beyond the scope of this article. Our research instead attempts to address the moderating role of software category in the relationship between online reviews and consumers’ sample downloads.

Conclusion

Prior to purchase, consumers are uncertain about product qualities for experience goods such as software. Online reviews and free samples serve as widely used information channels to provide consumers with product information. Numerous studies have examined the effects of both channels on consumers’ purchase behaviour. However, few studies have looked into how one information channel influences the other or how online reviews affect consumers’ sample downloading behaviours.

In addition, several scholars have examined the different effects of online reviews for different types of goods (e.g. search versus experience goods). Although software is considered a typical type of experience good, it can be further divided into different categories based on the heterogeneous characteristics of different software. Thus, it is also interesting to explore how product characteristics interact with online ratings to affect consumers’ sample downloads by incorporating software categories.

To answer these research questions, we draw on a data set from CNET.com and identified four software categories: Platform-related, Entertainment-related, Work-related and Decoration-related software. We then conduct an empirical analysis based on this data set. Major findings include that user ratings have significant effects on consumer downloads of software samples and that software categories moderate such effects.

Our findings make an important theoretical contribution to the literature related to online ratings and free sampling. By further dividing software, a typical type of experience good, into four categories based on the heterogeneous characteristics of different software, this research takes one step further than existing research on moderating the effects of product type (search or experience) on user ratings. Our research also provides important managerial implications to relevant practitioners, including software firms and consumers.

This article is not without limitations. Without software sales data on CNET.com, we are unable to examine how product characteristics moderate the relationship between information disclosure and product sales. Further study may be extended via merging sales data from different data sets.

Footnotes

Acknowledgements

We thank the editors and the anonymous referees for their constructive suggestions and comments through the revision process.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work is supported by the National Natural Science Foundation of China (71702109,71972004, 71731009, 71832011, 2018WZDXM020), the Research Foundation of Department of Education of Guangdong Province (2018WTSCX124) and the Philosophy and Social Science Foundation of Shenzhen (SZ2020C009).

ORCID iD

Shengli Li

Notes

References

Nelson

Advertising as information. J Polit Econ 1974; 82(4): 729–754.

Piramuthu

Kapoor

Zhou

, et al. Input online review data and related bias in recommender systems. Deci Support Syst 2012; 53(3): 418–424.

Banerjee

Chua

AYK

. Authentic versus fictitious online reviews: a textual analysis across luxury, budget, and mid-range hotels. J Inf Sci 2017; 43(1): 122–134.

Rana

Cheah

Sequential patterns rule-based approach for opinion target extraction from customer reviews. J Inf Sci 2019; 45(5): 643–655.

Morrison

How app store ratings impact downloads (Infographic), https://www.adweek.com/digital/how-app-store-ratings-impact-downloads-infographic/.(accessed 15 January 2016).

Dou

Niculescu

DJ.

Engineering optimal network effects via social media features and seeding in markets for digital goods and services. Inf Syst Res 2013; 24(1): 164–185.

Liu

Cheng

Tang

, et al. Optimal software pricing in the presence of piracy and word-of-mouth effect. Deci Support Syst 2011; 51(1): 99–107.

Niculescu

DJ.

Economics of free under perpetual licensing: Implications for the software industry. Inf Syst Res 2014; 25(1): 173–199.

Zhou

Duan

Online user reviews, product variety, and the long tail: an empirical investigation on online software downloads. Electron Commer Res Appl 2012; 11(3): 275–289.

10.

Cheng

Liu

Optimal software free trial strategy: the impact of network externalities and consumer uncertainty. Inf Syst Res 2012; 23(2): 488–504.

11.

Dey

Lahiri

Liu

Consumer learning and time-locked trials of software products. J Manag Inf Syst 2013; 30(2): 239–268.

12.

Cheng

Tang

QC.

Free trial or no free trial: optimal software product design with network effects. Eur J Oper Res 2010; 205(2): 437–447.

13.

Nan

Shi

Dou

, et al. Duopoly pricing of software products under free strategy: limited-feature vs seeding. Comput Indus Eng 2016; 100: 13–23.

14.

Cheng

Liu

Optimal software free trial strategy: limited version time-locked or hybrid?

Prod Oper Manag 2015; 24(3): 504–517.

15.

Hsu

Chang

KC.

Exploring the effects of online customer reviews, regulatory focus, and product type on purchase intention: perceived justice as a moderator. Comput Human Behav 2017; 69: 335–346.

16.

Lee

Shin

SY.

When do consumers buy online product reviews? Effects of review quality, product type, and reviewer’s photo. Comput Human Behav 2014; 31(1): 356–366.

17.

Park

Lee

TM.

Information direction, website reputation and eWOM effect: a moderating role of product type. J Bus Res 2009; 62(1): 61–67.

18.

Dellarocas

Zhang

Awad

NF.

Exploring the value of online product reviews in forecasting sales: the case of motion pictures. J Interact Mark 2007; 21(4): 23–45.

19.

Cui

Lui

H-K

Guo

The effect of online consumer reviews on new product sales. Int J Electron Commer 2012; 17(1): 39–57.

20.

Mai

The effect of online reviews on product sales: a joint sentiment-topic analysis. Inf Manag 2019; 56(2): 172–184.

21.

Ham

Lee

Kim

, et al. Subjective perception patterns of online reviews: a comparison of utilitarian and hedonic values. Inf Process Manag 2019; 56(4): 1439–1456.

22.

Nelson

Information and consumer behavior. J Polit Econ 1970; 78(2): 311–329.

23.

Cao

Duan

Gan

Exploring determinants of voting for the ‘helpfulness’ of online user reviews: a text mining approach. Deci Support Syst 2011; 50(2): 511–521.

24.

Zhou

Liu

Online roduct rating manipulation and market performance. IEEE Comput 2015; 48(5): 72–75.

25.

Chen

Duan

Zhou

The interplay between free sampling and word of mouth in the online software market. Deci Support Syst 2017; 95: 82–90.

26.

Moon

Bergey

Iacobucci

Dynamic effects among movie ratings, movie revenues, and viewer satisfaction. J Mark 2010; 74(1): 108–121.

27.

Zhou

Duan

Do professional reviews affect online user choices through user reviews? An empirical study. J Manag Inf Syst 2016; 33(1): 202–228.

28.

Awad

Ragowsky

Establishing trust in electronic commerce through online word of mouth: an examination across genders. J Manag Inf Syst 2008; 24(4): 101–121.

29.

Clemons

EK.

An empirical investigation of third-party seller rating systems in e-commerce: the case of buy SAFE. J Manag Inf Syst 2007; 24(2): 43–71.

30.

Brown

Broderick

Lee

Word of mouth communication within online communities: conceptualizing the online social network. J Inter Mark 2007; 21(3): 2–20.

31.

Lee

Tan

Effects of different types of free trials and ratings in sampling of consumer software: an empirical study. J Manag Inf Syst 2013; 30(3): 213–246.

32.

Lian

Lin

TM.

Effects of consumer characteristics on their acceptance of online shopping: comparisons among different product types. Comput Human Behav 2008; 24(1): 48–65.

33.

Pascual-Miguel

Agudo-Peregrina

ÁF

Chaparro-Peláez

Influences of gender and product type on online purchasing. J Bus Res 2015; 68(7): 1550–1556.

34.

Cheema

Papatla

Relative importance of online versus offline information for Internet purchases: product category and Internet experience effects. J Bus Res 2010; 63(9–10): 979–985.

35.

Zhu

Zhang

Impact of online consumer reviews on sales: the moderating role of product and consumer characteristics. J Mark 2010; 74(2): 133–148.

36.

Katz

Shapiro

Network externalities, competition, and compatibility. Am Econ Rev 1985; 75(3): 424–440.

37.

Cheng

Duan

, et al. A study of enterprise software licensing models. J Manag Inf Syst 2017; 34(1): 177–205.

38.

Nan

Zhang

, et al. Optimal pricing for new product entry under free strategy. Inf Technol Manag 2018; 19(1): 1–19.

39.

Basu

Mazumdar

Raj

SP.

Indirect network externality effects on product attribute. Mark Sci 2003; 22(2): 209–221.

40.

Brynjolfsson

Kemerer

CF.

Network externalities in microcomputer software: an econometric analysis of the spreadsheet market. Manag Sci 1996; 42(12): 1627–1647.

41.

Cheng

Jin

. Optimal distribution strategy for enterprise software: retail, SaaS, or dual channel? Prod Oper Manag 2018; 27(11): 1928–1939.

42.

Ellison

Fudenberg

The Neo-Luddite’s Lament : excessive upgrades in the software industry. RAND J Econ 2000; 31(2): 253–272.

43.

Dawson

Moderation in management research: what, why, when, and how. J Bus Phycol 2014; 29(1): 1–19.

44.

The interaction effects of online reviews and free samples on consumers’ downloads: an empirical analysis. Inf Process Manag 2019; 56(6): 102071.

45.

Huang

J-H

Chen

Y-F.

Herding in online product choice. Psychol Mark 2006; 23(5): 413–428.