An empirical assessment of the effects of
different types of electronic word of
mouth on brand attitude in the pre- and
post-purchase phase
By
Sophie Groeneweg
Master thesis MSc Marketing Management
University of Groningen
Faculty of Economics and Business
First supervisor: Dr. J. Berger
Second supervisor: Dr. A. Schumacher
January 13, 2020
Words: 13878
Hof van Naeltwijck 12
2631 WX Nootdorp
(06) 43216016
An empirical assessment of the effects of different types of electronic word
of mouth on brand attitude in the pre-and post-purchase phases
Abstract:
With the development of the internet and interactive technology, traditional WOM marketing was transferred to the electronic form, which is known as electronic word-of-mouth
(eWOM). It is important for companies to become aware of the influence of reviews on eWOM channels on the opinions and attitudes of consumers towards a product or service. eWOM is considered an important marketing communication tool and must therefore be added to companies’ strategies. The aims of this study are to determine how different types of eWOM influence consumers’ attitudes and to examine if this influence is different between the pre- and post-purchase phase of the customer journey.
To gain a deeper understanding of the influences of different eWOM types on consumers’ attitudes towards restaurants, a quantitative study is conducted. An online survey is created where 158 participants answer a series of questions related to the four types of eWOM. Based on partial least square regression (SmartPLS), it is determined that there is actually a
difference in the effects of the different types of eWOM. The results show that the general eWOM is not composed of four equal parts of the four eWOM types. In addition, based on the results it can be concluded that there is a difference in the extent to which eWOM is used in the pre- and post-purchase phases. Finally, the study showed that there is a difference in the effects of eWOM on people's attitudes towards a hedonic service compared to a hedonic product.
From a managerial perspective, this study offers marketers several practical considerations for selecting the right eWOM channel in the right phase to gain a competitive advantage in the market by attracting the most customers. Furthermore, this research can be a source for further research in this direction.
Table of Content
1. Introduction ………. 3
2. Theoretical background ……….. 7
2.1 Electronic word of mouth ……… 7
2.2 Brand attitude ………... 8
2.3 Customer Journey ……… 9
2.4 Conceptual framework ……… 10
2.4.1 Electronic word of mouth ……….. 10
2.4.1.1 Specialized eWOM ……….. 10 2.4.1.2 Affiliated eWOM ………. 11 2.4.1.3 Social eWOM ……….. 13 2.4.1.4 Miscellaneous eWOM ………. 14 3. Research design ………. 17 3.1 Data collection ……… 17
3.2 Population and sampling method ………... 17
3.3 Operational definitions ……….. 19
3.4 Plan of analysis ……….. 20
4. Analysis of results ………. 22
4.1 Sample ……… 22
4.2 Analysis of the pre-purchase phase ……… 22
4.2.1 Measurement model ……… 23
4.2.2 Structural model ……… 25
4.3 Analysis of the post-purchase phase ………. 27
4.3.1 Measurement model ……….. 27
4.3.2 Structural model ……… 29
5. Discussion and conclusion ………... 32
5.1 Specialized eWOM ……… 33
5.2 Affiliated eWOM ……….. 34
5.3 Social eWOM ……… 35
5.4 Miscellaneous eWOM ……….. 36
6. Theoretical and managerial relevance ……….. 41
7. Limitations and future research ……… 43
References ……….. 45
1. Introduction
In the past years it has become more and more common to give one’s opinion about certain things. However, do people make their decisions based on what other people say, or do people really look at a brand differently based on others’ opinions? This phenomenon is called word-of-mouth (WOM), which occurs when one person shares his or her opinion with another, which may influence the buying behaviour or intention of that other (Kudeshia and Kumar, 2016).
With the development of the internet and interactive technology, traditional WOM marketing has been transferred to the electronic form and is therefore, known as electronic word-of-mouth (eWOM). eWOM marketing is becoming a major source of information for customers and affecting their buying behaviour (Gu et al., 2012). According to Vazquez-Casielles, Suarez-Alvarez and del Rio-Lanza (2013) eWOM is an extremely credible and persuasive force affecting people’s attitudes towards products and services. eWOM includes consumer opinions, user experiences and product reviews and occurs across numerous online channels (Chu and Kim, 2011).
is structured and to what extent its average is influenced by the different types. Examining how different types of eWOM influence consumer’s attitudes is one of the aims of this research.
The customer journey consists of three phases: the pre-purchase phase, the purchase phase and the post-purchase phase (Lemon and Verhoef, 2016). In earlier studies, the
influence of eWOM has mainly been investigated over the entire customer journey, or during an individual phase (Lemon and Verhoef, 2016). It is therefore interesting to investigate whether there are differences between the phases of the customer journey concerning the extent to which eWOM influences the attitudes of consumers. Kristopher et al. (2014) have described the differences between before and during the purchase of a product. However, they have noted that it is more interesting to examine the differences between before and after the purchase so as to compare the pre- and post-purchase phases of the customer journey. Baur and Nystrom (2017) have agreed with this. Their research has shown that there is a gradual transition between the pre-purchase phase and the actual purchase phase with regard to the extent to which people use eWOM. Lemon and Verhoef (2016) have highlighted that during the pre- and post-purchase phases, people talk to each other (online and offline) about the products and services. This makes it interesting to research if there are differences, and what those differences are, between the influences of eWOM on people’s attitudes in the pre- and post-purchase phases. When marketing managers know the differences between those phases, they can adjust their strategies to make sure that people will repurchase their
products. Examining if there are differences between the pre- and post-purchase phases of the customer journey is one of the aims of this research.
information above research is needed to the effects of the different eWOM types on people's attitudes and the extent to which these types are used in the phases before and after the purchase. This research aims to as a contribution and gives more insights to the effects of the eWOM types on people's attitudes and the possible differences between the pre- and post-purchase phases. It is important for a manager to know in which phase which type of eWOM is used to reach the most consumers via the most influential channel per phase. As previously mentioned, the aims of this study are to determine how different types of eWOM influence the attitudes of consumers and to examine if this influence is different between the pre- and post-purchase phases of the customer journey.
The research question for this research is as follows:
To what extent do the different types of eWOM influence the attitudes of consumers, and to what extent does this effect differ between the pre- and post-purchase phases of the customer journey?
To summarize, significant gaps remain open regarding the influence of the different types of eWOM and what their influence on the attitude of consumers is (Kudeshia and Kumar, 2016; Hu and Ha, 2015; Kumar et al., 2016; Liu, Steenkamp and Zhang, 2018). It is hypothesized that all types of eWOM except for miscellaneous eWOM will have a positive influence on people’s attitudes in the pre-purchase phase. In the post-purchase phase, it is proposed that social and specialized eWOM will have a positive influence and that miscellaneous and affiliated eWOM will not have an influence on people’s attitudes. In addition, this study will extend the literature by exploring if there are differences between the pre- and post-purchase phase of the customer journey (Lemon and Verhoef, 2016). It is proposed that there are differences and that most of the influence of the eWOM types will be in the pre-purchase phase.
restaurants. In addition, there is an insignificant relationship between social eWOM channels and the attitude of consumers towards restaurants. Furthermore, the results of the post-purchase phase show that there is a significant effect of miscellaneous and social eWOM channels and an insignificant effect of specialized and affiliated eWOM channels on people's attitudes towards a restaurant. The study reveals that there is a difference in the extent to which eWOM is used in different channels.
The findings contribute to the general eWOM literature by suggesting that it is interesting to examine the differences between the different eWOM types as previously mentioned in studies by Ha and Hu (2015) and Hu et al. (2014). Furthermore, this study provides more integrative and theoretical insights into the extents to which eWOM is used in the pre- and post-purchase phases. Additionally, this study highlights the importance of managers considering which type of eWOM channels they should use in which phase to obtain and retain consumers.
The structure of this thesis is as follows; the first chapter reviews the existing literature and identifies important findings and gaps. This is followed by an overview of the conceptual model of this research. This chapter discusses all the variables and the
2. Theoretical background
This chapter reviews existing literature and identifies important findings and gaps within this research. Afterwards, an overview of the conceptual frameworks is provided. A distinction is made between the pre-and post-purchase phases, because these are the phases in which eWOM is present. Furthermore, it discusses the variables and the proposed hypotheses are identified.
2.1 Electronic word of mouth
eWOM is considered as an electronic version of the traditional word of mouth that is defined as “face-to-face communication between consumers regarding any product, brand or
service”(Arndt, 1967). Traditional WOM is therefore limited to face-to-face oral
communication. eWOM includes several aspects and is defined as “any positive or negative statement made by potential, actual, or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet” (Hennig-Thurau et al., 2004).
There are different communication platforms and functions, the mixed eWOM, and for this reason, Hu and Ha (2013) have separated eWOM into four different types. The first is, specialized eWOM. This refers to customer reviews posted on comparison- or rating websites. These websites do not sell the products themselves (e.g. kieskeurig.nl). The second type is the affiliated eWOM, which refers to customer reviews affiliated with retail websites. These websites provide both services and reviews (e.g. eBay). The third type is social
eWOM, which refers to any information about brands and products exchanged among users of social media networks (e.g. Facebook and Twitter). The last type is miscellaneous eWOM, which includes relevant information about the brand and product on other online media such as discussion boards, forums and emails.
types combined. Hu and Ha (2015) have researched the effect of, social eWOM. They have noted that it might be interesting to research if there are differences between the effect of social eWOM and the other types of eWOM.
The different types of eWOM can be seen as different channels that may influence people’s brand attitudes and purchase intentions. These different channels show different capabilities, influences and characteristics. However, little knowledge exists about the differences between and across these different channels. This makes it important to establish a deeper understanding of differences across these channels (Gvili & Levy, 2016).
2.2 Brand attitude
It is important for firms to differentiate themselves and the brand as an opportunity to
establish their brand reputation. This brand reputation will form people’s attitude towards the brand (Melewar et al., 2017). Attitude is defined as the “overall evaluation of a person against a particular object, persons and issues” (Peter and Olson, 2010; Petty, Cacioppo an
Schumann, 1983). Keller (2003) has defined brand attitude as “the evaluative dimension of brand image, which results from consumer’s beliefs and feelings toward the brand’s attributes and benefits”.
Attitudes towards brands can be thought of as consumers’ liking or lack of liking (Foroudi, 2019). Tariq et al. (2013) have stated that brand attitude is one of the dimensions of the brand upon which purchase intention is based. According to Abzari et al. (2014), brand attitude can be considered the most important determinant of purchase intentions. In addition, Liu et al. (2012) have stated that attitudes can also affect the state of a person in choosing something that he or she considers right when he or she is faced with a choice between right and wrong, because attitude is a person’s emotional state. Keller (2003) agrees with this and has stated that consumers develop their own evaluations and judgments. Keller (2003) has added that brand attitudes represent the synthesis of all relevant brand elements present in consumers’ memory.
the product or service (Liu, Steenkamp and Zhang, 2018; Anaya-Sanchez et al., 2019). When customers have a positive attitude towards a product or service, this is accompanied by more purchases. It is therefore important for a manager to be well aware of which factors have positive influences on this brand attitude in order to increase the positive attitude of consumers (Liu et al., 2012).
2.3 Customer journey
The customer journey is the “journey” that a customer makes to purchase a product or service. The customer journey includes the model in which this “customer journey” is mapped. This model also includes potential customers (Lemon and Verhoef, 2016). According to Richardson (2010), the customer journey can be seen as “a diagram that illustrates the steps a customer goes through in engaging with a company, whether it be a product, an online experience, retail experience, a service or any combination”.
According to Lemon and Verhoef (2016), a customer journey consists of three phases. The first is the pre-purchase phase in which the search for a product takes place. Possible touchpoints during this phase are conversations with others about the product or service and experience with advertising stimuli. The second is the purchase phase in which the actual purchase of the product is made. Possible touchpoints in this phase are using the product, using coupons obtained before the visit and using discounts. Finally, the product is evaluated during the post-purchase phase. Touchpoints in this phase include recommending a particular store, talking to others about purchases and planning a return trip to the store.
Within this research, the pre- and post-purchase phases are compared. As previously mentioned, these are the two phases in which people search for others’ opinions about
products and possibly align their own attitudes towards a certain product (Baur and Nystrom, 2017). During the purchase phase itself, there is little interest in other people’s opinions (Lemon and Verhoef, 2016).
2.4 Conceptual framework
Figure 1: Conceptual model of pre-purchase phase and post-purchase phase.
2.4.1 Electronic word of mouth
When someone reads something positive or negative about a certain product, it makes sense that this will influence how people will look at that product. Gvili and Levy (2016) have suggested that differentiation exists in the delivery of messages across different eWOM types, or communication channels. This can result in different attitudes towards different eWOM types. For each eWOM type, the extent to which they influence the attitudes of consumers in both phases is examined.
2.4.1.1 Specialized eWOM
comparing products is easy and which generally provides a fair assessment of the product. On the other hand, people dislike that they cannot purchase products on these websites (Baur and Nystrom, 2017). Based on the aforementioned knowledge it can be assumed that what is said by an expert on a specialized eWOM channel is positively related to consumers’ attitudes towards a product. This makes possible the formulation of the following hypothesis:
H1a: Specialized eWOM is positively related to consumers’ attitudes in the pre-purchase phase.
Compared to the research of Gvili and Levy (2016), it is proposed that after they have experience with a product, people will still believe what is on the specialized eWOM
channels and let it influence their attitudes towards the product. People are more likely to continue to believe that what the experts say is true and will base their upcoming purchases on this. Liu, Schuckert and Law (2018) agree with this. They have stated that the fact that a review is written by an expert will influence people's opinions in almost every situation. This is also the case after the purchase. People will compare their own experience with what the experts say. People believe that everything the experts say is true and try to determine whether their own experience matches that of the experts (Liu, Schuckert and Law, 2018). Derived from the information mentioned above, it can be expected that people will also examine the reviews of experts on specialized eWOM channels after their purchase and will be influenced by this. This makes possible the formulation of the following hypothesis:
H2a: Specialized eWOM is positively related to consumers’ attitudes in the post-purchase phase.
2.4.1.2 Affiliated eWOM
2017). Based on the results of the research of Petersson and Fjellström (2017) there were just a few people in the sample who concluded that they do not believe what is said in the reviews on the retailer’s website, so this finding of Petersson and Fjellström (2017) is neglected in this study of the pre-purchase phase. In addition, according to Kudeshia and Kumar (2017) and Petersson and Fjellström (2017), most of the respondents indicated that they view the reviews on this type of channel as confidential. Furthermore, Baur and Nystrom (2017) have mentioned that people experience affiliated eWOM channels as easy and quick to use. Based on their findings it can be assumed that there are also people who said that they do not trust the reviews on the affiliated channels since they are perceived as easily manipulated. However, Baur and Nystrom (2017) have come to the conclusion that because of their
convenient placement, many people read the reviews on these channels as a final check when they have already examined the information from other eWOM channels. As previously mentioned people experience a high convenience level at the affiliated eWOM channels. In addition, people indicate that the ratings on the channels have a high confidentiality factor. Based on this, it is expected that affiliated eWOM is positively related to consumers’
attitudes towards a product. This makes possible the formulation of the following hypothesis:
H1b: Affiliated eWOM is positively related to consumers’ attitudes in the pre-purchase phase.
According to earlier studies by Gvili and Levy (2016) and Ha and Hu (2015), people make little use of the affiliated eWOM channels after their purchase. People are more likely to share their own experience on these channels after their purchase in order to influence the opinions of others (Baur and Nystrom, 2017). People will not easily view the reviews on affiliated eWOM channels to compare their own experience with them after making their purchase (Gvili and Levy, 2016). In addition, people feel that the retailer has influenced the reviews on its own site. Peterson and Fjellström (2007) have stated that retailers are less likely to post negative reviews on their sites since this may scare people away. This means that customers believe these reviews to have a low trustworthiness level (Petersson and Fjellström, 2007). Gvili and Levy (2016) agree with this and have stated that the main reason that people see the reviews on affiliated channels after the purchase as unreliable is because they know that these reviews are written or influenced by the retailer. As a result, people do not bother to visit these sites after the purchase to compare their own opinions and
(Gvili and Levy, 2016). People’s attitudes towards a product or service are therefore not influenced by affiliated eWOM in the post-purchase phase. This shows that people are more likely to post their own experiences on the affiliated channels after their purchase and do not typically visit the channels to have their opinion influenced by the reviews. It is proposed that the affiliated eWOM has little or no influence on consumers’ attitudes. This makes possible the formulation of the following hypothesis:
H2b: Affiliated eWOM is not related to consumers’ attitudes in the post-purchase phase. 2.4.1.3 Social eWOM
Social eWOM consists of reviews on social media platforms such as Facebook, Twitter, Instagram and YouTube (Kudeshia and Kumar, 2017). Based on the results of the studies of Hu and Ha (2015), Hu et al. (2014) and Baur and Nystrom (2017) people see social eWOM as a channel with a lack of source trustworthiness. YouTube is used to provide a visual presentation of the product, but the fact that most people get paid to make those videos makes the trustworthiness level of what people say in their videos low (Baur and Nystrom, 2017). On the other hand, Djafarova and Rushworth (2017) have shown that the
trustworthiness of the influencer has a positive influence on people’s attitudes towards the product. The trustworthiness level depends on the degree to which influencers can be considered experts and the number of followers they have on their social media channels. Furthermore, in recent years it has been very common for people to want to belong with others; therefore, they listen to the opinions of other people - mostly celebrities - to ensure that they make the right decisions in their purchase process (Chan, Lee and Wong, 2018; Djafarova and Rushworth, 2017 & Singh and Banerjee, 2018). In this study, it is expected that the social pressure to belong will have a stronger influence than the low trustworthiness level of the influencer, which means that it is expected that social eWOM is positively related to consumer’s attitudes. The following hypothesis is therefore formulated:
H1c: Social eWOM is positively related to consumers’ attitudes in the pre-purchase phase.
experiences and behaviour of influencers. This means that even after the purchase, people are influenced by what influencers and other people post on social media channels, like Facebook and YouTube (Gvili and Levy, 2016). It can be expected that in the post-purchase phase people will still compare their own experiences with those of influencers and that their attitudes towards the product will be influenced by what is said on the social eWOM channels. The following hypothesis is therefore formulated:
H2c: Social eWOM is positively related to consumers’ attitudes in the post-purchase phase. 2.4.1.4 Miscellaneous eWOM
Miscellaneous eWOM refers to brand and product information exchanged on online social media platforms such as blogs, discussion boards and forums (Kudeshia and Kumar, 2017). Gvili and Levy (2017) have demonstrated that many people believe that what is on a blog has been written by professionals and influencers. Therefore, people are more likely to assume that what is contained in a blog post is true, and this will influence people’s attitudes towards a product or service. On the other hand, Kudeshia and Kumar (2017) have
H1d: Miscellaneous eWOM is not related to consumers’ attitudes in the pre-purchase phase.
According to Gvili and Levy (2017), people find it difficult to find miscellaneous eWOM, and searching reviews on miscellaneous eWOM channels takes a great deal of effort. Therefore, people will not use this channel after they have made a certain purchase. On the other hand, Ha and Hu (2015) have demonstrated that there are exceptions in people who are interested in the opinions of others and compare how their own opinions align with these, Ha and Hu (2015) have noted that this is just a small group of their sample so that people
compare their own opinions with that of others can be neglected. Furthermore, following the research by Gvili and Levy (2017), people do not consider the various eWOM channels very reliable. The main reason given here is that it has not been proven that the information and reviews on these channels are not written by people who actually understand the subject. Considering this in combination with the effort it takes to use this type of eWOM channel, it can be expected that during this study the influence of reviews on miscellaneous channels will have little or no influence on people’s attitudes in the post-purchase phase. The following hypothesis is therefore formulated:
H2d: Miscellaneous eWOM is not related to consumers’ attitudes in the post-purchase phase. 2.4.1.5 Customer journey
In their research, Gvili and Levy (2016) have demonstrated that eWOM occurs in both the pre- and post-purchase phases of the customer journey. Based on the aforementioned information, this thesis predicts whether there is a difference between the two phases and if so, what that difference is.
First, the pre-purchase phase. As previously shown, the specialized, affiliated and social eWOM channels influence people’s attitudes in a positive sense. The main reasons given for this are that through these channels, eWOM is provided by experts, influencers or celebrities who have a major influence on people’s behaviour and attitude (Baur and
Second, the post-purchase phase, both specialized and social eWOM influence people's attitudes. According to Gvili and Levy (2016), this is due to the fact that people believe what experts, celebrities and influencers say and adjust their attitudes accordingly. In addition, it is proposed that both affiliated and miscellaneous eWOM do not influence
people’s attitudes. This is confirmed in the research by Gvili and Levy (2016) and Ha and Hu (2015), which states that people in the post-purchase phase mainly share their own
experiences and are not influenced by the opinions and experiences of others on both the affiliated and miscellaneous eWOM channels.
3. Research design
The research design consists of the data collection method, followed by the population and sampling method of this research. Furthermore, the development of the measurement tools is given. Finally, an analysis plan is provided which is used to analyze the collected data. 3.1 Data collection
To investigate the extent to which the different types of eWOM influence people’s attitude towards restaurants, a quantitative study was conducted. The quantitative research design is based on an interpretive philosophy. It is intended to understand socially constructed meanings about the area of interest (Saunders, Lewis and Thornhill, 2012). The aim of this approach is to generate a richer theoretical perspective than already exists in previous literature (Saunders, Lewis and Thornhill, 2012).
A survey was conducted to perform quantitative research. This survey contained several situations from which information could be obtained to assess the effect of the eWOM types on people’s attitudes towards restaurants. Using surveys was the most appropriate way to gain enough data within the limited time given the small sample size of this research (Blumberg, Cooper and Schindler, 2014). The survey consists of three parts: internet usage habits, questions based on situations in the pre-purchase phase and questions based on situations in the post-purchase phase.
During the survey, statements were shown in which respondents had to indicate to what extent they agreed on a 5-point Likert scale. The statements consist of actions that can be done when a restaurant is visited or has already been visited. In addition, statements were displayed stating that a certain event has a major impact on the respondent. Based on this statements the influences of the types of eWOM op people’s attitudes are measured. Respondents were also asked to answer these questions with a 5-point Likert scale. These statements were made regarding both the pre- and post-purchase phases.
3.2 Population and sampling method
eWOM on attitudes towards products like restaurants is a common topic (Liu, Steenkamp and Zhang, 2018; Anaya-Sanchez et al., 2019). In their study, they highlighted that hedonic services like restaurants are higher involved. As a result, consumers conduct research regarding the reviews of other users. Due to the fact that research has already been done on the effect of the general eWOM on people's attitudes towards restaurants, there is still a gap in the effect of different types of eWOM on people’s attitudes towards restaurants (Liu, Steenkamp and Zhang, 2018). For this reason, this study examines the effects of the different types of eWOM on consumers’ attitudes towards restaurants. Therefore, the population of this study contains Dutch people who visit restaurants. The survey was spread through social media and email, so anyone could participate in the research. Control questions were added to the survey to ensure that data is only collected from people who have visited a restaurant.
During this study, partial least squares (PLS) were used. Since indicators are measuring different latent variables in this analysis, PLS is an appropriate technique to use because it combines both multiple regression analysis and factor analysis (Hair, Ringle and Sarstedt, 2011). Furthermore, this type of regression was used because it reduces the
predictors to a smaller set of uncorrelated components to overcome multicollinearity between the four different types of eWOM (Garson, 2016). The PLS method is further explained in section 3.3. There are two conditions for the sample size to use the PLS to analyze the data. First, the sample size should be 10 times the number of indicators of the scale with the largest number of formative indicators. This means that the sample size for the pre-purchase phase should be at least 20 and for the post-purchase phase 40. Furthermore, the sample size should be 10 times the largest number of structural paths directed to the dependent variable. The largest number of structural paths in both the pre- and post-purchase phase is four. This means that the sample size of users should be at least 80 for the combination of both phases (Chin, 1998; Garson, 2016; Hair, Ringle and Sarstedt, 2011).
3.3 Operational definitions
This section provides the definitions of the variables and how they were measured during this research (Table 3). Since this study considers the possible differences between two phases and these are determined with different measuring scales, two models were tested. The scales were adapted from previously validated studies, in order to measure the influences of the four different eWOM types on consumers’ attitudes. Table 3 provides the definitions of the variables and how they are measured per phase (pre- and post-purchase phase). These variables were operationalized based on prior research. The survey questions are listed in Appendix A. The original statements are in English. Since the respondents in this research are Dutch, backward translation was used. The original statements were translated into Dutch by a native Dutch speaker. Accordingly, the Dutch translations were translated into English again by competent English speaker. The original English statements and the translated English statements were compared until both statements were the same (Al-Amer et al., 2015).
Variable Operational definition
Affiliated eWOM (AFF) In the pre-purchase phase the influence of Affiliated eWOM on attitude is measured according to the article of Hu et al., (2014). The influence is measured with one
statement with a 5-point Likert scale. In the post-purchase phase the influence of Affiliated eWOM on attitude is measured according to the articles of Hu et al., (2014) and Baur and Nystrom (2017). The
influence is measured with two statements with a 5-point Likert scale.
Social eWOM (SOC) In the pre-purchase phase the influence of Social eWOM on attitude is measured according to the article of Hu et al., (2014). The influence is measured with one
statement with a 5-point Likert scale. In the post-purchase phase the influence of Social eWOM on attitude is measured according to the articles of Hu et al., (2014) and Baur and Nystrom (2017). The
Specialized eWOM (SPEC) In the pre-purchase phase the influence of Specialized eWOM on attitude is measured according to the article of Hu et al., (2014). The influence is measured with two
statements with a 5-point Likert scale. In the post-purchase phase the influence of Specialized eWOM on attitude is measured according to the articles of Hu et al., (2014) and Baur and Nystrom (2017). The
influence is measured with two statements with a 5-point Likert scale.
Miscellaneous eWOM (MISC) In the pre-purchase phase the influence of Social eWOM on attitude is measured according to the article of Hu et al., (2014). The influence is measured with one
statement with a 5-point Likert scale. In the post-purchase phase the influence of Social eWOM on attitude is measured according to the articles of Hu et al., (2014) and Baur and Nystrom (2017). The
influence is measured with two statements with a 5-point Likert scale.
Brand attitude In both the pre-and post-purchase phase the influence on attitude is measured by letting people indicate to what extent their opinion is influenced in certain situations, based on the article of Erkan and Evans (2016). This will be tested This will be done with a 5-point Likert scale
Table 3: Operational definitions
3.4 Plan of analysis
4. Analysis of results
This chapter describes the analysis and results of this research. The sample characteristics are presented. After this, the analysis for both the pre-purchase and the post-purchase phase is covered separately. Since two different measuring scales are used per phases, two models will be tested and described for each phase. The chapter finishes by analysing the hypotheses of this research.
4.1 Sample
The sample of this research contains 158 respondents, which meets the requirements of 80 when both phases are taken together as mentioned earlier. Of this sample, 43,04 per cent is male, and 56,96 per cent is female. The average age of the sample is 39 years old with a standard deviation of 11.07 (see table 4.1).
Sample size N = 158 Age Mean: 39 SD: 11.07 Min: 18 Max: 63 Gender Males = 68 (43,04%) Females = 90 (56,96 %) Table 4.1: Sample
As mentioned in section 3.4, PLS-SEM analysis exists of two sets of linear equations. Since two phases are being tested with different measuring scales, the two sets of linear equations will be carried out for both phases. In the following sections, these equations will be elaborated in models for both the pre- and post-purchase phases. The algorithm output of both models is given in Appendix B.
4.2 Analysis of the pre-purchase phase
4.2.1 Measurement model
A number of variables in this model have formative scales. Within formative scales, discrepancies in the value of the latent variable are determined by the changes in the value of the indicators. This means that the arrows go from the measured indicators to the latent variable. Furthermore, formative indicators are not related to each other, because they measure different facets of the latent variable (Wong, 2013). In addition, in this model there are also variables with a reflective scale. If the indicators are highly correlated and
interchangeable, they are reflective and their reliability and validity should be thoroughly examined. On a reflective scale, the causality direction goes from the latent variable to the indicators (Hair et al., 2010; Petter, Straub and Rai, 2007).
The measurement model should meet the minimum requirements for reliability and validity. These requirements can be assessed by its convergent and discriminant validity. Convergent validity measures whether a factor is uni dimensional, whereas discriminant variability examines whether each latent variable statistically represents theoretically different concepts. Assessing the validity of the reflective and formative scales requires different approaches (Hair et al., 2010; Hair, Ringle and Sarstedt, 2011).
First, the indicator reliability and convergent validity of the reflective scales are assessed. These can be measured with Cronbach’s alpha and composite reliability. Cronbach’s alpha provides an estimate for construct reliability, based on indicator inter-correlations. Composite reliability includes the factor loading of each indicator and is preferred as a test of convergent validity (Henseler et al., 2009; Götz, Liehr-Gobbers & Krafft, 2010). These two tests are only used for testing the convergent validity in reflective models since the formative scales are not expected to relate to each other (Hair, Ringle and Sarstedt, 2011).
can be concluded that there is no correlation within these indicators. When the convergent validity test was finished, the value of all measurements was 1.000. It can be concluded that all the constructs show high convergent validity and reliability.
For formative scales, convergent validity is measured by assessing the significance (p-values) and the variance inflation factors (VIF’s) of the indicator weights. To assess outer weights’ significance, a bootstrapping technique was applied. Both specialized eWOM and attitude indicators have formative scales since all constructs explains a different part of the indicator. The p-values of the indicators have to be higher than the threshold value of 0.05 to be significant (Wong, 2013). The VIF coefficient values of the indicators should be low because formative latent construct indicators are expected to measure different aspects of the same construct. The VIF values of the indicators have to be lower than the threshold value of 3.3 (Wong, 2013). Based on Table 4.2, all the p-values are below the threshold value of 0.05 and the VIF value are below the threshold value of 3.3. Accordingly, the p-values and the VIF’s show acceptable convergent validity of the formative scales.
Constructs/ indicators Indicator weights p-value VIF Specialized eWOM (SPEC)
PRE-SPEC1 PRE-SPEC2 0.022 0.999 0.043** <0.001*** 1.034 1.002 Attitude (ATT) PRE-ATT-AFF PRE-ATT-MISC PRE-ATT-SOC PRE-ATT-SPEC1 PRE-ATT-SPEC2 0.244 0.052 0.002 -0.059 0.874 <0.001*** <0.001*** 0.005*** 0.037** <0.001*** 1.406 1.364 1.383 1.623 1.280 Note: ***p ≤ 0.01;∗∗ p ≤ 0.05;∗ p ≤ 0.10.
Table 4.2: Convergent validity of formative scales
Furthermore, discriminant validity is examined by comparing “the square root of the AVE values with the latent variable correlations and the other latent constructs”(Fornell and Larcker, 1981). The constructs have to meet the Fornell-Larcker criterion: “the square root of the AVE of each construct should be higher than its highest correlation with any other
AVE are shown on the diagonal, and the off-diagonal coefficients represent the correlation coefficients.
AFF ATT MISC SOC SPEC
AFF 1.000
ATT 0.542 0.564
MISC 0.108 0.473 1.000
SOC 0.102 0.304 0.254 1.000
SPEC 0.129 0.428 0.460 0.421 0.786
Table 4.3: Discriminant Validity
Table 4.3 shows that the square root of each construct’s average variance extracted is larger than any other correlation coefficient. This model reflects the Fornell-Larcker criterion, which reflects discriminant validity.
Since the measurement model is sufficiently strong, the structural model can be assessed. 4.2.2 Structural Model
The second step of a PLS-SEM analysis is assessing the structural model or the so-called inner model. This model estimates the relationship between the latent constructs (Hair et al., 2017). The goodness of fit of the structural model can be checked by utilizing the and multicollinearity and explained variance (R-square and Adjusted R-square).
VIFs MISC → ATT 1.077 SOC → ATT 1.074 SPEC→ ATT 1.243 AFF → ATT 1.259
Table 4.4: Collinearity statistics (VIFs)
Thereafter, the explained variance will be examined. R-square is a statistical measure depicting the percentage of the variance for a dependent variable explained by the
independent variables in a model, whereas adjusted R-square takes the number of predictors in the model into consideration (Garson, 2016). This model shows a high variation of approximately 80 per cent (R-square is 0.799 and the adjusted R-square is 0.791), which means that 80 per cent of the variance of the model is predicted by the independent variables of the model. Independent variable → Dependent variable Beta (β): Sample Mean: SD: T statistic: p-value: MISC → ATT 0.296 0.306 0.088 3.360 <0.001*** SOC → ATT 0.192 0.203 0.093 1.64 0.244 SPEC → ATT 0.170 0.162 0.099 2.307 0.088* AFF → ATT 0.563 0.559 0.079 7.106 <0.001*** Note: ***p ≤ 0.01;∗∗ p ≤ 0.05;∗ p ≤ 0.10.
Table 4.5: Standardized path coefficients
the model can be included in its assessment. Moreover, this part of the research is very important to be able to get an answer to the main question to what extent there is a difference between the four different types of eWOM. Therefore, it is not possible to delete an eWOM type as it means that an important part of the research is missing which results in not being able to measure the main question.
4.3 Analysis of the post-purchase phase
All latent variables of the conceptual model for the post-purchase phase are included with their complementing indicators. The indicators are the statements which have been used in the survey questions. First, the measurement model will be covered. Once the measurement model proves to be sufficiently strong, the structural model will be assessed.
4.3.1 Measurement Model
The variables in this research are formative variables. Because all constructs explain another part of the indicator. Within formative models, discrepancies in the value of the latent variable are determined by the changes in the value of the indicators. This means that the arrows go from the measured indicators to the latent variable. Furthermore, formative indicators are not related to each other, because they measure different facets of the latent variable (Wong, 2013).
For formative scales, convergent validity is measured by assessing the significance (p-values) and the variance inflation factors (VIF’s) of the indicator weights. To assess outer weights’ significance bootstrapping technique was applied. The p-values of the indicators have to be higher than the threshold value of 0.05 to be significant (Wong, 2013). VIF coefficient values of the indicators should be low because formative latent construct
indicators are expected to measure different aspects of the same construct. VIF values of the indicators have to be lower than the threshold value of 3.3 (Wong, 2013) or 4.0 (Malhotra, 2009). Based on Table 4.6, the p-values of four indicators (i.e. ATT_AFF; ATT_SPEC1 and ATT_SPEC2) are above the threshold value of 0.05 which means that they are not
significant.
attitude. This might lead to contradicting subjective interpretations, hence explaining the insignificant results. Therefore, deleting the item does not lead to any complications in the theoretical framework. Furthermore, the influence of affiliated eWOM on attitude is only measured with one indicator (ATT_AFF). The p-value is not significant (0.372), but due to the fact that this indicator is theoretically significant for the latent variable and without this indicator the factor might not yield meaningful information, this variable was retained. Finally, ATT_SPEC2 has a p-value of 0.277, which indicates insignificance. Because all constructs have formative scales and therefore measure an important part of the indicator, this construct, despite its insignificance, cannot be removed from the model.
Despite the non-significance of the aforementioned indicators, the VIF coefficient of almost all the indicators is below the threshold value of 3.3 (Wong, 2013) and the VIF of POST-ATT-SPEC2 is below the threshold value of 4.0 (Malhotra, 2009), which illustrates that indicators measure different facets of the same latent variable. Therefore, even though indicators are not significant they adequately capture the theoretical dimensions.
Constructs/ indicators Indicator weights p-value VIF Specialized eWOM (SPEC)
POST-SPEC1 POST-SPEC2 0.836 0.291 <0.001*** 0.079* 1.246 1.314 Social eWOM (SOC)
POST-SOC1 POST-SOC2 0.312 0.809 0.024** <0.001*** 1.322 1.254 Affiliated eWOM (AFF)
POST-AFF1 POST-AFF2 0.569 0.621 0.001*** <0.001*** 1.203 1.156 Miscellaneous eWOM (MISC)
POST-MISC1 POST-MISC2 0.792 0.382 <0.001*** 0.020** 1.165 1.373 Attitude (ATT) POST-ATT-AFF POST-ATT-MISC POST-ATT-SOC POST-ATT-SPEC1 POST-ATT-SPEC2 0.214 0.341 0.364 0.319 -0.145 0.372 0.078* 0.098* 0.113 0.277 2.462 2.974 2.766 3.164 3.936 Note: ***p ≤ 0.01;∗∗ p ≤ 0.05;∗ p ≤ 0.10.
Furthermore, discriminant validity utilizes the average variance extracted (AVE) scores by the Fornell-Larcker criterion. For each latent variable, the square roots of AVE should be greater than correlation coefficients of those latent variables with other constructs (Fornell and Larcker, 1981). Table 4.7 shows the discriminant validity results. The squared roots of the AVE are shown on the diagonal, and the off-diagonal coefficients represent the correlation coefficients. Table 4.7 shows that the square root of each construct’s average variance extracted is larger than any other correlation coefficient. This model reflects the Fornell-Larcker criterion, which reflects discriminant validity.
AFF ATT MISC SOC SPEC
AFF 0.746
ATT 0.523 0.844
MISC 0.578 0.638 0.790
SOC 0.690 0.684 0.623 0.753
SPEC 0.714 0.581 0.755 0.747 0.771
Table 4.7: Discriminant validity
Since the measurement model is satisfactory, the structural model can be assessed. 4.3.2 Structural Model
The second step of a PLS-SEM analysis is assessing the structural model or the so-called inner model. This model estimates the relationship between the latent constructs (Hair et al., 2017). The goodness of fit of the structural model will be checked by utilizing the and multicollinearity and explained variance (R-square and Adjusted R-square).
VIFs
MISC → ATT 2.386 SOC → ATT 2.599 SPEC→ ATT 3.421 AFF → ATT 2.301
Table 4.8: Collinearity statistics (VIFs)
Thereafter, the explained variance will be examined. R-square is a statistical measure depicting the percentage of the variance for a dependent variable explained by the
independent variables in a model, whereas adjusted R-square takes the number of predictors in the model into consideration (Garson, 2016). This model shows a high variation of approximately 55 per cent (R-square is 0.541 and the adjusted R-square is 0.527). Which means that 55 per cent of the variation of the model is predicted by the independent variables in the model. Independent variable → Dependent variable Beta (β): Sample Mean: SD: T statistic: p-value: MISC → ATT 0.381 0.392 0.172 2.215 0.027** SOC → ATT 0.494 0.515 0.154 3.213 0.001*** SPEC → ATT -0.098 -0.124 0.183 0.536 0.592 AFF → ATT 0.032 0.050 0.095 0.338 0.736 Note: ***p ≤ 0.01;∗∗ p ≤ 0.05;∗ p ≤ 0.10.
Table 4.9: Standardized path coefficients
The significance of the factor loading is measured using the bootstrapping technique. Table 4.9 presents the structural path coefficients, which can be defined as weights
requirement (the p-values are below the threshold value of 0.05 and the t-statistics are above the threshold value of 1.96). Furthermore, Table 4.9 shows that the effects of both SPEC to ATT and AFF to ATT are insignificant (the p-values are above the threshold value of 0.05 and the t-statistics are below the threshold value of 1.96). Because a moderate adjusted R-square has been established in this model, it can be said that despite the insignificance of the effect of specialized eWOM (p = 0.592) and affiliated eWOM (p = 0.736) on attitude, all parts of the model can be included in its assessment. In addition, this is a part of the research that is very important to be able to get an answer to the main question of to what extent there is a difference between the four different types of eWOM; this makes it impossible to remove an eWOM type.
4.4 Hypotheses testing
The hypotheses of both the pre- and post-purchase phases were tested by examining the results of Tables 4.5 and 4.9 in which the overall values of the effects are given.
Hypothesis 1a predicted that specialized eWOM would positively influence consumers’ attitudes in the purchase phase. The effect of SPEC on ATT in the pre-purchase phase is positive and significant at a 10% significance level (β = 0.170, p = 0.088). This means that Hypothesis 1a is supported. The direct effect of AFF on ATT is positive and significant at a 1% significance level (β = 0.563, p = <0.001). This supports hypothesis 1b, so the affiliated eWOM has a positive influence on consumers’ attitudes in the pre-purchase phase. Regarding the influence of social eWOM on consumers’ attitudes towards restaurants in the pre-purchase phase, we determined that this effect is insignificant (β = 0.192, p = 0.244). This means that there is no effect of SOC on ATT, so Hypothesis 1c is rejected. The direct effect of MISC on ATT is positive and significant at a 1% significance level (β = 0.296, p = <0.001). Thus, Hypothesis 1d is rejected.
Based on the results, we ascertain that MISC has a significant effect of a 5% significance level on ATT (β = 0.381 and p = 0.027). So miscellaneous eWOM is related to people’s attitudes towards restaurants in the post-purchase phase. Thus, Hypothesis 2d is rejected. An overview of the tested hypotheses is given in Table 4.10.
Independent → Dependent Hypothesis effect Effect
H1a: Specialized eWOM → Attitude Positive effect Supported
H1b: Affiliated eWOM → Attitude Positive effect Supported
H1c: Social eWOM → Attitude Positive effect Rejected
H1d: Miscellaneous eWOM → Attitude No effect Rejected
H2a: Specialized eWOM → Attitude Positive effect Rejected
H2b: Affiliated eWOM → Attitude No effect Supported
H2c: Social eWOM → Attitude Positive effect Supported
H2d: Miscellaneous eWOM → Attitude No effect Rejected
5. Discussion and conclusion
This chapter covers the discussion of the results. The analysis is highlighted and discussed, reflecting the findings from Table 4.10.
5.1 Specialized eWOM
Regarding our results, we determined that specialized eWOM have a positive and significant effect on people’s attitudes towards restaurants in the pre-purchase phase. If the specialized eWOM increased by one, the attitude would increase by 0.170 (Table 4.10). When people see reviews of experts on rating websites or in online videos before they have visited a restaurant, they are be positively influenced. This effect is in line with prior studies by Baur and
Nystrom (2017) and Kudeshia and Kumar (2017). These researchers have explained that people see specialized eWOM channels as a platform where experts with a great deal of knowledge post their reviews. According to their research, people believe what is said on these platforms. Furthermore, Baur and Nystrom (2017) and Kudeshia and Kumar (2017) emphasize that comparison sites are considered sites where comparing products is easy and which generally provides a fair representation of the product. Therefore, it can be concluded that what is said on specialized eWOM channels is positively related to consumers’ attitudes towards restaurants.
Furthermore, we examined the results of the effect of specialized eWOM on
product like a car, they are more inclined to compare what possible alternatives have been posted on comparison sites. In addition, when making a more expensive purchase, people prefer to compare whether their expectations and experience match those of others by looking at online platforms where these opinions are shared (Lin and Lekhawipat, 2016). Lin and Lekhawipat (2016) have researched how people’s expectations after making the purchase can be influenced by what other people say. Furthermore, Dabholkar (2014) has indicated that people are more likely to visit a rating website when they are able to return the product. This is not possible when visiting a restaurant and can therefore result in very few visits to these websites and therefore no influence of this type of eWOM.
5.2 Affiliated eWOM
As expected, affiliated eWOM has a positive and significant effect on consumers’ attitudes towards restaurants in the pre-purchase phase. According to Table 4.10, if the use of affiliated eWOM increased by one unit, consumers’ attitudes towards restaurants would improve by 0.563. When people search for a product on the website itself, they are influenced by the reviews posted on that same website. This is in line with an earlier study by Kudeshia and Kumar (2017), who have stated that people indicate that what is said on the affiliated channels has a high confidentially factor with a high convenience level. This positive and significant effect is probably caused by the fact that before visiting a restaurant, people first visit the restaurant’s website to search the menu for something they would like for dinner (Johns and Kivela, 2008). In this way, people will quickly see the reviews of previous guests, and this influences their attitudes towards the restaurant. Furthermore, in line with the results of Baur and Nystrom (2017), people make extensive use of affiliated eWOM channels before going to a restaurant. According to Baur and Nystrom (2017), this may be due to the fact that people experience a high level of user-friendliness and therefore use affiliated eWOM
affiliated eWOM channels after their purchase. These researchers have stated that people are more likely to share their own experience on these channels to influence the opinions of others instead of letting these channels influence their own opinions. According to the researchers mentioned above, the main reason of having no effect of affiliated eWOM channels in the post-purchase phase is that people have the idea that what has been said on these channels is influenced by the restaurant itself. Research by Petersson and Fjellström (2007) has shown that retailers mainly post positive reviews on their sites to create a positive image of the restaurant. This means that people assign these reviews a low trustworthiness level and therefore do not use them much. It can be concluded that people make little or no use of affiliated eWOM in the post-purchase phase, so it has no influence on consumers’ attitudes towards restaurants.
5.3 Social eWOM
We determined that social eWOM has an insignificant effect (p = 0.244) on people’s attitudes towards restaurants in the pre-purchase phase. If the social eWOM increased by one, the attitude would increase by 0.192 (Table 4.10). This is not in line with the proposed hypothesis and earlier studies in which Baur and Nystrom (2017) and Djafarova and Rushworth (2017) have demonstrated that people look at social media channels before purchasing a product, mainly because people want to belong with others (i.e., influencers) and therefore listen to their opinions. The results of this study indicate that there is little or no influence of social eWOM on people’s attitudes towards restaurants in the pre-purchase phase. This could be due to the fact that people know that influencers get paid to give their opinion in their videos on social media channels. This results in people feeling that
influencers do not express their own opinions, which reduces the trustworthiness level of these opinions. This reduced trustworthiness level corresponds to what is said in the studies by Baur and Nystrom (2017), Ha and Hu (2015) and Hu et al., (2014). When drawing up the hypothesis, we stated that the extent to which an influencer is seen as an expert would
increase the reliability factor and that this would be more important than that these people get paid, which influences their opinions. Based on the results of this study, it appears that this is not the case and that people are most likely to let the fact that people get paid play an
Furthermore, we examined the results of the effect of social eWOM on the attitude of consumers towards restaurants in the post-purchase phase. Table 4.10 shows that social eWOM has a positive and significant direct effect on consumers’ attitudes towards
restaurants with a magnitude of 0.494. This is in line with previous studies. Djafarova and Rushworth (2017) and Singh and Banerjee (2018) have demonstrated that despite their own experiences with a product or service, people will, still listen to the opinions of others – especially well known-people - on social media channels and adjust their decision accordingly. Based on the results of this study, after visiting a restaurant, people look at social media channels for the opinions of other people and influencers to compare their own experience and possibly have their own opinion influenced by it accordingly. In addition, some people post their own opinion on social media channels after their experience with the restaurant in order to persuade others regarding their opinions or to stop them from visiting the restaurant (Boo and Kim, 2013).
5.4 Miscellaneous eWOM
Based on our results, we determined that miscellaneous eWOM has a positive and significant effect on people’s attitudes towards restaurants in the pre-purchase phase. If the
paid to create an attractive image of a product or service to increase purchases (Lim et al., 2017). The study by Lim et al., (2017) focuses on influencers’ impact on the purchase of hedonic products. Based on the results of this research, it can be concluded that when people are going to visit a restaurant they are influenced by what influencers or other well-known people have said on miscellaneous eWOM channels.
Furthermore, we examined the results of the effect of miscellaneous eWOM on consumers’ attitudes towards restaurants in the post-purchase phase. Table 4.10 shows that miscellaneous eWOM has a positive and significant direct effect on consumers’ attitudes towards restaurants with a magnitude of 0.381. This is not in line with previous studies. In their research, Gvili and Levy (2017) have discussed that based on the fact that they are difficult to find, people do not use these channels after they have made a purchase. The results of this study show that people use miscellaneous eWOM channels after they have visited a restaurant. Previous research has mainly focused on the influence of miscellaneous eWOM channels for hedonic products, such as cars and watches. These studies have
concluded that miscellaneous eWOM channels have little or no influence on people's attitudes. The results of this research indicate that when visiting a restaurant - a hedonic product - there is some influence of miscellaneous eWOM channels on people’s attitudes towards the restaurant. The difference between the results of this study and those of earlier studies is therefore probably due to the type of product with which the study was conducted. Miscellaneous eWOM channels therefore have an influence on hedonic services and not on hedonic products. Ha and Hu (2015) have concluded that only a small group is interested in the opinions of others after buying a product. The results of this study indicate that this group is actually not so small in the sample used and that there is much interest in the opinions of other well-known people after visiting a restaurant. It can be concluded that miscellaneous eWOM channels are positively related to consumers’ attitudes towards restaurants in the post-purchase phase.
5.5 Customer journey
Furthermore, it is predicted that miscellaneous eWOM have little or no influence on people’s attitude in the pre-purchase phase. This study has shown that specialized, affiliated and miscellaneous eWOM channels are positively related to people’s attitudes towards
restaurants. In addition, social eWOM has little or no influence, which is different from what was predicted. The prediction about the miscellaneous eWOM channel was incorrect. It was predicted that there would be little or no influence on people’s attitudes towards restaurants, but based on the results of this study, the miscellaneous eWOM channel is positively related to consumers’ attitudes towards restaurants in the pre-purchase phase.
For the post-purchase phase, it was predicted that both specialized and social eWOM have a positive influence on people's attitudes (section 2.4.1.5). According to Gvili and Levy (2016), this is due to the fact that people believe what experts, celebrities and influencers say and adjust their attitudes accordingly. In addition, it was predicted that affiliated and
miscellaneous eWOM channels would have little or no influence. The results indicate that two predictions were incorrect. Specialized eWOM has no influence on people’s attitudes towards a restaurant in the post-purchase phase, and miscellaneous eWOM does have an influence. The predictions about social and affiliated WOM were correct.
Overall, it can be concluded that three types of eWOM are used in the pre-purchase phase and two in the post-purchase phase. This is in line with what has been predicted, because there is more influence from eWOM in the pre-purchase phase. In the pre-purchase phase, use is primarily made of the specialized, miscellaneous and affiliated eWOM
channels. Before purchasing, people are more likely to look at channels such as comparison sites, blogs and retailer websites. These channels influence people’s attitudes towards restaurants in the pre-purchase phase. In the post-purchase phase, the miscellaneous and social eWOM channels are mainly used. After visiting a restaurant, people visit blogs and social media channels to share their opinions and may have their own opinions influenced by these sources.
5.6 Conclusion
First, the results of this study indicate that there are real differences between the influences of the different types of eWOM on people’s attitudes towards a restaurant. The results demonstrate that miscellaneous eWOM channels are most frequently used in the pre- and post-purchase phases. People therefore make the most use of blogs on which influencers post their opinions of products and services. Furthermore, specialized and affiliated eWOM channels are only used in the pre-purchase phase, and the social eWOM channel is used in the post-purchase phase. It can be concluded that the average eWOM is not composed of four equal parts of the four eWOM types; one type has more influence than the others and
therefore comprises a larger part of the total eWOM.
Second, when the two different phases of the customer journey are compared, the results show that three types of eWOM are used in the pre-purchase phase and two in the post-purchase phase. In the pre-purchase phase, specialized eWOM channels (comparison sites and videos from experts) miscellaneous eWOM channels (blogs written by influencers) and affiliated eWOM channels (retailer websites) are used. In the post-purchase phase, miscellaneous eWOM channels (blogs written by influencers) and social eWOM channels (social media channels such as Facebook and Twitter) are used. The difference may lie in the fact that before visiting a restaurant, people determine exactly what the options are and can find this on comparison sites, blogs and the retailer’s site. In addition, only the post-purchase phase uses social media channels. This may be due to the fact that after a purchase, people want to compare their personal experiences with those of others and this can influence their opinions about and attitudes towards the restaurant.
Furthermore, the effect of the different types of eWOM on consumer’s attitudes towards a restaurant - a hedonic service - is different from this effect on, for example, a hedonic product, as has been investigated in earlier studies by Baur and Nystrom (2017), Hu and Ha (2015), Hu et al., (2014) Lim et al., (2017) and Kudeshia and Kumar (2017). This is based on the fact that the hypotheses are derived from information of studies on the effect of eWOM on people's attitudes towards a hedonic product. The results of this study showed that a number of predictions about the influences of the types of eWOM on people’s attitudes were incorrect. One of the causes is the difference between the extent to which people use eWOM for a hedonic service compared to a hedonic product.
6. Theoretical and managerial relevance
There are several implications of this study. The following section first describes the theoretical relevance, which is followed by the managerial relevance.
First, compared to other research related to eWOM (Hu and Ha, 2015; Hu et al., 2014), this conceptual model includes the four different types of eWOM. Therefore, this study expands the research by exploring new variables affecting consumers’ attitudes. These variables consist of the four different types of eWOM: social eWOM (SOC), affiliated eWOM (AFF), specialized eWOM (SPEC) and miscellaneous eWOM (MISC). The findings illustrate that the average eWOM is not composed of four equal parts of the eWOM types. It appears that one type of eWOM has more influence than another type and therefore
comprises a larger part of the total eWOM. The results of this study provide valuable information about how the eWOM is structured and that it therefore does not consist of the four equal parts that were assumed in previous studies. Moreover, in line with Coker (2012) and Erkan and Evans (2018), we highlight the importance for future research to not only examine eWOM as a whole or just one type of eWOM but to consider the effects of the different types of eWOM since there is a difference in their effects on people’s attitudes.
Secondly, compared to other research related to eWOM, this study has added an extra component by comparing the use of eWOM in the pre-purchase phase and the post-purchase phase. This provides added value to the previously conducted studies where just the entire consumer journey or only one phase was considered. The findings illustrate that there is a difference in the extent to which eWOM is used in the two different phases. For example, it appears that there is more usage of eWOM in the pre-purchase phase than in the
Third, this study examines the effect of eWOM on consumers’ attitudes towards restaurants - a hedonic service - instead of the influence on a hedonic product, which previous studies have mainly focused on. The research by Abzari, Ghassemi and Vosta (2014) has shown that in addition to the many studies of the influences of eWOM on attitudes towards hedonic products, there is also a need for knowledge about the influences on hedonic services. The findings of this study indicate that there is actually a difference between the influences of eWOM on hedonic services compared to hedonic products. In line with Abzari, Ghassemi and Vosta (2014), this research highlights the importance of future research
regarding the impact of eWOM on a hedonic service and therefore demonstrates that there is actually a difference.
7. Limitations and future research
Although this study provides significant insights into the effects of the different types of eWOM on consumers’ attitudes towards restaurants in both the pre- and post-purchase phases of the customer journey, there are several limitations that need to be addressed, which
provides scope for future research.
First, when testing the influence of the different types of eWOM on people’s attitudes towards a restaurant, no control variable was used. This is an important limitation of this study. This means that it is now difficult to indicate the extent to which there is a cause effect relationship between the eWOM types and the consumers’ attitude. Adding a control variable would ensure better accuracy of the research. A possible control variable that could be used in future research is health. People’s health can influence how often they visit a restaurant and search for reviews (eWOM) about the restaurant. In addition, income could also be used as a control variable. Having a higher income is related to the ability to visit a restaurant more often and investigate which restaurant has the best scores according to the reviews of others (eWOM).
Second, this study used quantitative methods. In future studies it could be interesting to use qualitative methods in the form of interviews to find the underlying reasons for certain answers. This is a better way to determine the reasons that one type of eWOM has more influence than the others. In this way, managers can better respond to consumers’ needs and possibly attract and retain more people.
Third, this study made no distinction between different ages and the difference in their experiences with the internet. Therefore, in further research, it would be interesting to
consider possible differences between, for example, Generation X (born between 1960 and 1980 and therefore had to learn how to use the internet) and Generation Y (born between 1981 and 2000 and therefore having grown up with the internet) (Erkan and Elwalda, 2018; Baur and Nystrom, 2017). This provides a better insight into how different generations use eWOM and to what extent this is due to their experience with the internet.
