WHO DECIDES WHERE TO GO FOR A COFFEE? e-WOM AND CONSUMERS’ PURCHASE INTENTION

The primary purpose of this study is to examine the impact of eWOM on consumers' selection intention of a café in New Zealand. The factors of eWOM that may affect customers' selection intention have been identified, focusing on cafes in New Zealand. A quantitative research approach was employed, and a survey was conducted to collect data from 234 New Zealand consumers through an online questionnaire. The findings of this study indicate that the quantity, quality, valence, and consistency of eWOM positively impact consumers' purchase intentions. The perceived credibility and the perceived usefulness have indirect effects on eWOM and consumers' purchase intention. Among them, the valence of eWOM is fully intervened by the perceptual credibility, and the quantity, quality and consistency of eWOM are partly influenced by perceptual credibility. The quality and valence of eWOM are partially mediated by the perceived usefulness, while the quantity and consistency are not mediated by perceived trust. TAM to study online consumers' purchase intention in the New Zealand café industry. In addition, the modified model of this study explains the mediating effect of customer perceived usefulness and perceived credibility in the sphere of eWOM. Future research may use this theoretical model to investigate other restaurants, and comparison studies can be explored in different types of restaurants. The study also provides practical implications for café managers, especially in electronic marketing. The results of this study show that the volume, quality, valence and consistency of eWOM have a significant positive impact on consumers' purchase intention. Furthermore, managers should consider eWOM as one of the essential marketing tools to promote their businesses.


INTRODUCTION
With the development of the internet and smartphones, consumers started to share their personal views and feedback about products or services on the internet. Different scholars call for this online feedback in different ways, such as the internet word-of-mouth (Gelb & Johnson, 1995), word-ofmouth (Helm, 2000), internet communication (Putrevu & Lord, 2003), electronic referrals (De Bruyn & Lilien, 2004), viral marketing (Palka et al., 2009), and others. These different concepts overlap each other in connotation and have different emphases. It is found that consumers can channel word of mouth through public comments, emails, comment areas, distributed mail systems, one-on-one mail, chat rooms and instant messaging (Hennig-Thurau et al., 2004). eWOM can improve consumers' attention and acceptance and deepen their impression and memory (Mican & Sitar-Taut, 2020;Hu & Huang, 2015). Electronic word of mouth (eWOM) about products has changed the traditional communication mode and become a more influential online source. The electronic economy has made the information channels more diversified and given more space to consumers to make choices. In addition, eWOM is considered a better tool for promotion and advertising than traditional approaches. The eWOM has a high degree of transparency, broad spectrum, and significant time effectiveness and is more acceptable to consumers (Bruhn et al., 2012;Cheung & Thadani, 2012;Trusov et al., 2009). Therefore, eWOM has attracted more attention from enterprises where they have taken the initiative to use eWOM as a marketing tool to enhance their brand image, especially in the restaurant industry. Restaurant marketing managers are increasingly concerned about online reviews or posting about their food and services from customers (Al-Azzam, 2016;Arif, 2019;Litvin et al., 2008).
In 2018, the gross sales of cafes and restaurants in New Zealand were more than $ 5.6 billion, which has increased by 2.5% compared with gross sales in 2017. In the meantime, compared with 2017, the cafe and restaurant industry has employed 72,000 workers, and the employment rate in this group increased by 7.4% in 2018 (scoop.co.nz, 2019). In addition to the significant economic impact of the coffee industry, many cities in New Zealand, such as Auckland and Wellington, have established a unique identity based on their coffee culture (Weaver, 2010). Undoubtedly, the café índustry has already been the main contributor to the New Zealand economy and culture, especially in the hospitality and tourism industry (Zhang et al., 2019). Therefore, the impact of eWOM communications in the New Zealand café industry is a meaningful research topic.
Customers now can acquire knowledge or comments through the eWOM generated by people who have visited the place before. The online customer feedback system, also known as a reputation system, utilizes the internet's two-way communication ability to allow individuals to share experiences and opinions online about a company, product, service, or even an event (Vilpponen et al., 2006). These eWOMs enable them to gain a better and more comprehensive understanding of the restaurant ahead, such as food quality, service, price, environment, and other related aspects (Jeong & Jang, 2011;Levy & Gvili, 2019;Tan et al., 2016). The new information gained by customers from the eWOM could be helpful for them in reducing the potential risk of making wrong decisions about the selection of a restaurant (Huifeng & Ha, 2021;Serra-Cantallops et al., 2020;Vajjhala & Ghosh, 2021). The eWOM is communicated via the internet, which happens among anonymous consumers.
The eWOM used to be a helpful approach for people to gain valuable and essential information before purchasing or making decisions (Lerrthaitrakul & Panjakajornsak, 2014;Stauss, 1997). According to a New Zealand Asia Institute survey, more than 80% of consumers may believe or trust the recommendations from family and friends (The University of Auckland, 2019). Even though most of the tools for advertising purposes utilized by companies are effective, the impact of exchanging information via interpersonal communication has increasingly become more significant during customers' decision-making process. Customers are more likely to pay attention to products that have gained heated discussions on the internet, such as forums or websites of companies or products (Pai et al., 2013). It is valuable and interesting to choose the café industry in New Zealand as one specific research topic because cafes are not only an essential part of city dwellers' life as well as a way to present the local culture to tourists but also have made significant contributions to New Zealand economy and culture (Zhang, 2017). Although the effectiveness of eWOM impacting consumers' views and behaviors has been observed, it still needs to be explored by sufficient research in the New Zealand café industry.

RESEARCH OBJECTIVES
This study aims to achieve the following objectives: 1. To assess whether the eWOM affects the selections of café shops.
2. To explore the impact of the four dimensions of eWOM (volume, quality, valence and consistency) on café shop selections.
3. To find the relationship between eWOM (perceived credibility and usefulness) and consumers' preferences of café shops. 4. To examine the factors of perceived credibility and perceived usefulness that affect the eWOM for consumers' selections of café shops.
This study focuses on the café industry in New Zealand, especially how consumers' selection intention is impacted by eWOM. The study addresses a relatively unexplored area in New Zealand.
The study aims to fill in an important gap in the existing literature and provides more insights for future research. Further, it is a worth exploring topic to understand why consumers are full of enthusiasm about eWOM and why they seek help and suggestion through a network environment.

LITERATURE REVIEW
EWOM has become one of the critical sources of consumer information search and impacts purchase decisions (Chu & Kim, 2011). The present study draws on the Technology Acceptance Model (TAM) to detail consumers' purchase intentions affected by eWOM. It explains people's behavior in terms of their acceptance of Information Technology (IT) and has been commonly used in research on the eWOM as an effective mechanism for customers' online transaction decisions (Erkan & Evans, 2016;Tseng & Hsu, 2010;Wu & Lin, 2017;Yang & Zhou, 2011).
Based on TAM, Sussman and Siegal (2003) put forward an information adoption model, explaining the factors influencing people's information adoption behaviors in the context of the internet and emphasizing the mediating role of perceived usefulness in the process of information adoption. Wong (2018) asserted that perceived usefulness, perceived risk, trust, and electronic word of mouth are four predicting factors of customers' purchase decisions; these factors are integrated and proposed in the model based on the TAM to evaluate the role of social media in customer's buying decision. eWOM has been identified as an intermediate variable related to the effective mechanism of customers' online shopping decisions. The present study applies the research models of both Sussman and Siegal (2003) and See-To and Ho (2014) by positing online review as an independent variable and positing perceived usefulness and perceived credibility as dependent variables to examine the relationship between eWOM and online purchase decision.

Dimensions of eWOM and Consumers' Perceptions
Since there are many forms of eWOM, in order to facilitate unification and measurement, scholars usually investigate the influence of eWOM on consumers' purchasing decisions from the following four dimensions: Volume. The volume of eWOM refers to the scale and amount of eWOM on a particular product or service across the Internet (Cheung & Thadani, 2012). The more customers discuss a product and the higher they pay attention to a brand or product, the more the customers intend to purchase and the more a retailer sells (Chevalier & Mayzlin, 2006;Duan et al., 2008;Godes & Mayzlin;Moon et al., 2010). Zhu and Zhang (2010) studied how sales volume and consumer reviews affect consumers' online purchasing behavior and also examined the relative effectiveness of two kinds of information sources (expert reviews and consumer reviews). It turns out that people like to make decisions based on the choices and evaluations of others. In addition, other consumer reviews have more impact than expert reviews. However, these studies do not research the characteristics of the audience. The number of comments is positively correlated with product sales: the greater the number of online reviews, the greater the impact on customers' purchasing decisions (Hu et al., 2011;Mudambi & Schuff, 2010). Likewise, the volume of eWOM, score, and rate of negative comments significantly correlate with sales revenue, intermediated by product price as a variable (Liao et al., 2021).
However, the quantity of information could not be measured by the number of comments but from the perspective of quality. The quantity of information refers to the number of comments and includes the amount of information contained in the content itself . In addition, more information does not promote consumers' purchase behavior, and information overload can lead to the reduction of consumers' perception of information. For consumers with high involvement degree, their purchase intention increases first and then decreases with the increase of information quantity.
For consumers with low involvement, purchase intention rises with increased information quantity.
Quality. The quality of eWOM refers to the authenticity, reliability, relevance, and adequacy of comments (Mtchedlidze, 2019). Compared to plain and insipid information about eWOM, robust and distinctive information can garner more consumers' attention and interests, enabling consumers to invoke association of thought that may generate more significant influence on their memories and judgment affected by eWOM. On the other hand, the enjoyment of information content has been identified to directly impact communication effects and speed of information (Liu et al., 2012). There is a positive correlation between review quality and rating and sales volume and consumers paid attention to the content of reviews rather than just statistical data (Casaló et al., 2010;Chen et al., 2008;Chevalier & Mayzlin, 2006;Mtchedlidze, 2019).
Valence. The valence of eWOM refers to consumers' evaluation of a certain product. Positive word of mouth can increase purchase intention, while negative word of mouth can lower purchase intention, as well as the latter degree is more substantial (Brown et al., 2005;Litvin et al., 2008). Willemsen et al. (2011) found that consumers pay more attention to two-sided information than to pure positive or negative information because two-sided information is more interesting, novel and credible. Two-sided word of mouth information means that word of the mouth contains both positive and supportive information and negative and opposing information . The purchase intention and sales volume of consumers are also positively correlated with the comprehensive valence of eWOM (Hu et al., 2011). Positive word of mouth has a positive valence, while negative word of mouth has a negative valence, as well as neither positive nor negative word of mouth, is neutral. In most cases, the higher the score of eWOM, the stronger its persuasion (Boerman et al., 2017). The impression formation of psychology research has shown that people respond differently to positive and negative information (or events) and that negative information elicits stronger psychological arousal, attention, emotion, attribution, and social action responses than positive information.
Many studies in the marketing field support this view because consumers tend to view negative information as more judgmental than positive information and thus rely more on negative information. Negative eWOM had a negative impact on consumers' trust level, and purchase motivation in both online and physical stores, as well as the lower the price of goods, the greater the negative impact and customers tend to attribute the motivation of negative reviewers to subjective rather than objective product factors for experiential goods (Chatterjee, 2001;Sen & Lerman, 2007).
However, Chevalier and Mayzlin (2006) argue that positive word of mouth positively impacts customer purchasing behavior, while negative word of mouth has no significant impact. Consumers who received positive recommendation information were much more likely to buy a product than consumers who did not receive positive recommendation information (Garbarino & Strahilevitz, 2004;Lee et al., 2008).

Consistency.
Consistency of eWOM refers to the level to which different reviewers agree with the eWOM of the same product or brand (Cheung et al., 2009;Cheung & Thadani, 2012). If everyone has similar views, people believe that eWOM is highly consistent. Clemons et al. (2006) showed that the variance of online scoring and the mean of higher scoring were positively correlated with the sales growth of new products through their research on craft beer. The consistency of eWOM indicates that most people in the consumer group agree with the same view. Under the influence of social norms, customers will accept the word of mouth information. Chang et al. (2015) studied the influence of negative eWOM on consumers' brand switching behavior. The results showed that the stronger the consistency of negative eWOM was, the easier it was for consumers to make external attributions and vice versa.
To sum up, it is found that the four dimensions of eWOM have a significant influence on customers' purchasing decisions. The higher the volume of eWOM, the greater the influence on consumers' buying decisions. While the higher the quality of eWOM, the stronger the persuasiveness of consumers' purchases online. When the overall eWOM on a product is consistently high, consumers would easily accept the product.

Perceived Usefulness and Perceived Credibility
Electronic word of mouth not only has a direct impact on customers' purchasing intentions but also has an indirect impact with perceived usefulness and perceived credibility as mediating variables (Sussman & Siegal, 2003;Van der Heijden et al., 2003;Ha & Stoel, 2009;Wu, 2013;See-To & Ho, 2014;Wong, 2018). Ghose and Ipeirotis (2007) found that positive and negative word of mouth affects the quality and usefulness of online reviews through variance analysis. The quality of eWOM has a highly significant impact on customers' perception of usefulness and credibility. High-quality eWOM has higher perceived credibility and perceived usefulness. On the contrary, low-quality word of mouth has lower perceived credibility and perceived usefulness (Cheung et al., 2009;Lee et al., 2008).
Similarly, eWOM, which is highly logical and can reflect the real attributes of a product, is more likely to be believed by consumers to influence decisions ultimately (Wu & Wang, 2011). However, disorganized, and baseless eWOM is rarely trusted. Therefore, eWOM is significantly correlated with customers' perceived usefulness and credibility. Furthermore, the credibility of information perceived by receivers is the key to the information persuasion process. If people believe that the information is credible, they are more likely to adopt the eWOM to make decisions, and there is a positive correlation between the recipient's perception of the credibility of the virtual network and the recipient's intention to accept the information (Flanagin & Metzger, 2007;Wasko & Faraj, 2005).
According to the TAM proposed by Davis (1986), people's perceived usefulness to a certain technology system will significantly influence their attitude and final behavior toward adopting the technology system. Cheung et al. (2008) found a significant positive correlation between customers' perceived usefulness to eWOM and their catering decisions through their research on catering review websites (openrice.com). Therefore, it can be inferred that perceived usefulness will significantly affect customers' purchase intention.
Trust and perceived usefulness would positively influence purchase intention (Farzin & Fattahi, 2018;Gefen et al., 2003). Based on the trust theory, Yang and Zhou (2011) used reputation and eWOM as moderating variables to analyze the mediating role of trust in the influence of reputation and eWOM on consumers' purchasing intentions. The perceived usefulness and perceived credibility of eWOM play an intermediary role in the influence of eWOM on purchase intention (Cheung & Lee, 2012;Wong, 2018).

RESEARCH MODEL AND HYPOTHESES DEVELOPMENT
Based on the literature review, this study proposed the research model in Figure 1. below. The model consists of seven factors, including volume, quality, valence, consistency, perceived usefulness, perceived credibility, and purchase intention. The model also anticipates a relationship between four independent variables related to eWOM (volume, quality, valence, and consistency) as well as two dependent variables of perceived usefulness and perceived credibility. Furthermore, this study believes that different dimensions of eWOM have a significant impact on customer intention. The presentation of information has a significant effect on consumers' judgment. Different volumes, quality, valence and consistency of the same eWOM lead to different degrees of purchase intention.
According to research using TAM and the study of Erkan and Evans (2016), eWOM influences purchase intention through two mediating variables: perceived usefulness and perceived credibility (Daowd et al., 2020). Figure 1. demonstrates the research model formulated for this study.

Figure 1. Conceptual Model for this Study
Based on the literature review and operationalization of variables, the hypotheses have been formalized for this study and presented in Table 1 below:

Ethical Considerations
Because this research involves a data collection focusing on the customer's purchase intention related to eWOM, a consent form was shown to the participants to protect them from subsequent effects; the researchers expect to build trust with participants and promote the research completeness through this. On the other hand, the participants can understand the purpose of the study and the direction of data usage, and they can also know that the data collecting process would be rigorous, and identities would keep secret and not be disclosed. Moreover, each participant was required to sign the consent form. Those who did not wish to participate were entitled to withdraw it at any time.

RESEARCH METHODOLOGY
Quantitative research is applied in this study, and the data was collected through an online questionnaire which was distributed via a URL link. The primary data of this study is mainly collected in Auckland, which is the largest city in New Zealand. Auckland attracts more than 38% of companies in the whole country (Zhang, 2017). The online questionnaire was distributed at the ICL school's website and coffee shops in Auckland. At the ICL campus, the URL link of the questionnaire was submitted to the students via Canvas. The managers in Aoraki café and Coffee Plenty café were contacted and requested to send the survey URL link to participants via their Facebook messaging systems.

Questionnaire Design
The questionnaire was adopted from previous studies questionnaires by Al-Azzam (2016)

Sample Size
The target investigated group in this survey are the ordinary consumers above 18 years of age who shared their recent electronic word of mouth experience about coffee shop services or product purchases to obtain critical data. The convenience sampling technique was used in this research to collect data. Moreover, using the convenience sampling method, a higher response rate can be obtained to satisfy the desired target population (Kelley et al., 2003;Palinkas et al., 2015). The sample size of this study was calculated using the sample size calculator of creative research systems.
According to the New Zealand government, the resident population was 4,929,700 in 2018 (Stats NZ, n.d.). The confidence level and confidence intervals are 95% and 6%, respectively, which most academic studies use to calculate the sample size (Bonett & Wright, 2015). In total, 234 questionnaires were gathered.

Data Analysis
This research uses the SPSS software to carry out the descriptive statistical analysis, reliability and validity analysis, correlation analysis and multiple linear regression, testing hypothesis and the proposed research model. At first, the research uses descriptive statistical analysis to describe the characteristics of the subject by frequency and percentage. Then an analysis of reliability and test on validity were performed. For describing the measurement framework, confirmatory factor analysis was used. Next, Spearman's rank correlations analysis was applied to find relationships between the study structures. The exploratory factor analysis was used to examine whether the basic measurements for each study structure were correct. Finally, the mediation effect test method was applied to study the function of regression analysis variables in the Internet word of mouth effect.
The research data analysis had concerns about the framework, overall fit, predictability, and the importance of the path. Thus, in order to perform data analysis, we grouped questions to their respective variable.

Participants' Demographic Information
In this study, different members of the New Zealand population, including males and females, were invited to answer the online questionnaire. The invitees were coffee consumers in New Zealand.
Members of the public and professionals are also contacted and invited to participate through Facebook and the ICL school website. However, they were asked those with online purchasing experience to participate. This study collected 234 samples, and the sample efficiency was 100% (see Table 2).   The education level of the participants is divided into four levels, including below undergraduate degree, bachelor's degree, master's degree, and Ph.D. degree (see Figure 4).
56.4% of participants had completed their bachelor's degree, and 29.9% had a master's degree. The remaining participants comprised 10.7% having an educational level below an undergraduate degree, while 7% were PhDs. who used the internet more than 21 hours per week. The remaining 17.5%, 17.1%, 10.7% and 3% used the internet between 11 and 20 hours, between 6 and 10 hours, as well as between 1 and 5 hours and less than one hour per week, respectively. Table 3 shows the number, maximum, minimum, means and standard deviations of research factors.

Descriptive Statistics
Among these independent variables, the perceived usefulness score was the highest, with a mean score of 2.861 and a standard deviation of .9810. At the same time, the valence demonstrated the lowest mean score of 2.485 with a standard deviation of 1.1805.

Reliability and Validity Analysis
Reliability refers to the consistency and stability of measurement results when repeated measurements are made at different times (Melchers & Beck, 2018). The reliability analysis of the questionnaire in this study uses the Cronbach's α (Alpha), and the factor analysis method carried out the validity analysis.

Reliability Analysis
The reliability analysis results investigated in this study are shown in Table 4 below: As can be seen from the table, the α coefficient of 7 variables is higher than 0.9, and the Cronbach's Alpha Based on Standardized Items is 0.934. Therefore, the question items in the questionnaire have α certain degree of internal consistency and stability, which is credible and acceptable.

Validity Analysis
In this study, structural validity analysis is carried out. Factor analysis is one of the structural analysis methods, which extracts common factors to replace original variables for analysis and is the most used validity analysis method. Before extracting common factors, KMO (Kaiser-Meyer-Olkin) and Bartlett's test are required to determine whether factor analysis is suitable. In general, KMO above 0.70 means it is suitable for factor analysis, while below 0.60 means it is not suitable for factor analysis, and between 0.60 and 0.70 means it is barely suitable (Frohlich & Westbrook, 2001  .000

Source: Study Analysis
As indicated in Table 5, the KMO of the sample survey is 0.928, close to 1, and its criterion is beyond compare. Moreover, the significance probability of the approximate Chi-Square of Bartlett's Test of Sphericity is 0.000 (Sig.<0.05), so it can be judged that the sample is suitable for factor analysis.

Correlation Analysis
In this study, Spearman's rank Correlation Coefficient was mainly adopted to investigate the correlation between the seven variables and their significance (two-sided test). The analysis results are shown in Table 6. It can be seen from Table 6 above that the four dimensions (quality, volume, valence, and consistency) and two mediating variables (perceived usefulness and perceived credibility) of eWOM were all significantly correlated with purchase intention. More specifically, perceived usefulness had the strongest correlations with purchase intention (r = .749) compared to the other variables, followed by valence (r = .741). Consistency had the lowest correlation with purchase intention, which was .627.

Hypotheses Testing
In this study, regression analysis was conducted according to the mediating effect analysis method, as well as to test the proposed hypothesis. Meanwhile, gender, age, educational level, and network usage duration were taken as control variables in all models. This section hence presents the mediating effect analysis results of the perceived credibility and usefulness of eWOM.

Mediating Effect Analysis Method
The analysis carried out as simultaneously evaluating the relationship between independent factors from aspects like the single, interval scaled or ratio scaled is called the multiple regression (Hair et al., 2007). The significance of factors will be justified by the multiple regression and the relationship between independent factors, and the multiple regression will determine the dependent variable.

Journal of Marketing Strategies (JMS)
Volume 4, Issue 1, January 2022 40 In this study, regression analysis was conducted according to the mediating effect analysis method, as well as testing the proposed hypothesis. Meanwhile, gender, age, educational level, and network usage duration were taken as control variables in all models. The study conducted by Chatterjee et al. (2021) found that eWOM has no effect on the UK male and female customers; however, it has an impact on Indian male and female customers, which means the gender has varied impact on the customers' purchase intentions of different countries. This section presented the mediating effect analysis results of the perceived credibility and usefulness of eWOM.
This study uses the mediating effect analysis method proposed by Kenny (2008) (see Figure 6) when investigating the mediating role of perceived credibility and perceived usefulness in the effect of eWOM. In particular, to test the mediating effect between the independent variable (X) and dependent variable (Y) of mediating variable (M), three conditions (1) to (3) should be met before (4) can be carried out. There are four conditions listed below.

i. The independent variable (X) was used for regression of the dependent variable (Y),
and the regression coefficient (c) was significant.

ii. The independent variable (X) was used for regression of the intermediary variable (M), and the regression coefficient (a) was significant.
iii. The dependent variable (Y) was regressed with the intermediary variable, and the regression coefficient (b) was significant.

iv. The independent variable (X) and the intermediate variable (M) were used to conduct the regression of the dependent variable (Y).
If the regression coefficient c' of X obtained from (4) is significantly reduced (but still significant) relative to c, it indicates that M plays a partial mediating role. If c' is not significant, then M is fully mediating.

Figure 6. Schematic Diagram of Mediating Effect
Source: Baron & Kenny, 1986 The following four sections will use the above mediating effect analysis method to test the eight hypotheses of this study. The four dimensions of eWOM will be tested step by step for conditions 1 to 4 through multiple regression analysis. The independent variable (X) is volume, quality, valence, and consistency in this study.  Table 8 shows the regression results of the mediating effect of perceived credibility and usefulness on the volume of eWOM.

Source: Study Analysis
As shown in Table 8 above, Model 1 presents the regression of volume to purchase intention, and the adjusted R square of this model is 0.048, indicating that this model can explain a total variation of 4.8% of purchase intention; its F-statistic is 4.370, and its significance coefficient is 0.000, which shows this model is significant on the whole. The regression coefficient of volume is positive (0.112) and significant at the level of 0.01, indicating that the volume positively impacts the purchase intention, meeting the condition 1.
Model 2 is the regression of volume to perceived credibility. The significance coefficient and regression coefficient of the model are both 0.000, significant at the level of 0.01. This indicates that quantity has a significant positive effect on perceived credibility and satisfies condition 2.
Model 3 is the regression of volume to perceived usefulness. The significance coefficient of the model is 0.125, which is not significant at the level of 0.05. Similarly, the regression coefficient of quantity is 0.013, which is not significant at the level of 0.01. It indicates that volume does not significantly influence perceived usefulness and does not meet condition 2. Therefore, mediating effect analysis cannot be carried out. Similarly, perceived usefulness has no mediating effect on the volume of eWOM, thus accepting H20 and rejecting H2A Model 4 is the regression of perceived credibility and usefulness to purchase intention. The adjusted R square of this model is 0.089, indicating that this model can explain 8.9% of the total variation of purchase intention, its F-statistic is 5.723, and its significance coefficient is 0.000, which shows this model is significant on the whole. The regression coefficients of perceived credibility and usefulness are both significant at the level of 0.05, indicating that perceived credibility and usefulness have a significant positive impact on the purchase intention, meeting the conditions 3.
The volume and perceived credibility conform to the analysis conditions of mediating effect, so model 5 is further analyzed. Model 5 is the regression of volume and perceived trust on purchase intention, and the adjustment R square of this model is 0.068, which shows this model is significant on the whole. By comparing model 5 with model 1, it can be found that the coefficient of volume decreases, but it is still significant (p=0.017). Therefore, the following outcome can be drawn that perceived credibility plays a partial mediating role in volume, rejecting H10 and accepting H1A. Table 9 shows the regression results of the mediating effect of perceived credibility and usefulness on the quality of eWOM.

Source: Study Analysis
As shown in the above table 9, model 6 presents the regression of quantity to purchase intention, and the adjusted R square of this model is 0.063, indicating that this model can explain a total variation of 19.5% of purchase intention; its F-statistic is 3.537, and the significance coefficient is 0.000, which shows this model is significant on the whole. The regression coefficient of quality is positive (0.063) and significant at the level of 0.01, indicating that quality positively impacts the purchase intention, meeting the condition 1.
Model 7 and Model 8 present the regression of quality to perceived credibility and perceived usefulness, respectively. The significance of the model is 0.000, which is significant at the level of 0.01. Similarly, the regression coefficients of perceived credibility and perceived usefulness are 0.254 and 0.059, respectively, both significant at the level of 0.01. This shows that quality has a significant positive impact on perceived credibility and perceived usefulness, satisfying the condition (2).
The regression results of Model 4 in Table 4 indicate that both perceived credibility and usefulness have significant positive effects on the purchase intention, satisfying the condition 3. The above regression results indicate that the above variables all meet the analysis conditions of mediating effect, so Model 9 and Model 10 can be further analyzed.
Model 9 presents the regression of quality and perceived credibility to purchase intention. The adjusted R square of the model is 0.311, indicating that this model can explain a total variation of 31.1% of purchase intention. The F-statistic is 4.231, and the significance coefficient is 0.000, which shows this model is significant on the whole. Through the comparison between Model 9 and Model 6, it can be found that the coefficient and significance of quantity are reduced but still significant (p=0.002). Therefore, the following outcome can be drawn that perceived credibility partially mediates quantity, hence rejecting H30 and accepting H3A.

Journal of Marketing Strategies (JMS)
Volume 4, Issue 1, January 2022 44 Model 10 presents the regression of quality and perceived usefulness to purchase intention. The adjustment R of this model is 0.252, indicating that this model can explain a total variation of 25.2% of purchase intention. The F-statistic is 3.812, and the significance coefficient is 0.000, which shows this model is significant on the whole. Comparing model 10 with model 7 shows that both coefficient and significance of quantity are reduced but still significant (p=0.010). Therefore, the following outcome can be drawn that perceived usefulness partially mediates quantity, rejecting H40 and accepting H4A.  can be found that the coefficient and significance of valence are reduced, which is not significant at the level of 0.05 (p=0.063). Therefore, the following outcome can be drawn that perceived credibility has a full mediating effect on quantity, rejecting H50 and accepting H5A.

Multiple Regression Analysis for Valence
Model 15 presents the regression of valence and perceived usefulness to purchase intention. The adjusted R square of this model is 0.092, indicating that this model can explain 9.2% of the total variation of purchase intention. The F-statistic is 4.621, and the significance coefficient is 0.000, which shows this model is significant on the whole. By comparing Model 15 with Model 12, it can be found that the coefficient and significance of titer are reduced but still significant (p=0.093).
Therefore, the following outcome can be drawn that perceived usefulness has a partial mediating effect on valence, rejecting H60 and accepting H6A. Table 11 presents the regression results of the mediating effect of perceived credibility and usefulness on the consistency of eWOM.

Source: Study Analysis
As seen from the above table, Model 16 presents the regression of consistency to purchase intention.
The adjusted R square of this model is 0.056, indicating that the model can explain a total variation 46 of 5.6% of purchase intention. The F-statistic is 3.365, and the significance coefficient is 0.000, which shows this model is significant on the whole. The regression coefficient of consistency is positive (0.176) and significant at the level of 0.01, indicating that consistency has a positive impact on purchase intention, meeting condition 1.
Model 17 presents the regression of consistency to perceived credibility. The significance coefficient and regression coefficient of the model are both 0.000, significant at the level of 0.01. It indicates that consistency has a significant positive effect on perceived credibility and satisfies condition 2.
Model 18 presents the regression of consistency to perceived usefulness. The significance coefficient and regression coefficient of the model are both 0.081, which is not significant at the level of 0.05, indicating that consistency has no significant impact on perceived usefulness and does not meet the condition 2. Therefore, mediating effect test cannot be carried out. Similarly, perceived usefulness has no mediating effect on online word-of-mouth consistency, accepting H80 and rejecting H8A.

Hypothesis Test Results and Model Modification
According to the data analysis results, the hypothesis test results in this study are shown in Table 12. intention. Inconsistent with the other eWOM research taken in the volume and consistency of eWOM conducted by Erkan and Evans (2016) as well as Tseng and Hsu (2010) showed that perceived usefulness is the mediating effect that influences the eWOM on purchase intention. The reason might be that consumers' demographic information (e.g., age, gender, habit, culture) has different views on eWOM's usefulness. On the other hand, with the growth of the valence of eWOM, the customers would have a higher trust in eWOM which may lead to a more apparent intention to buy (Rahman et al., 2020;Saleem & Ellahi, 2017;Yusuf et al., 2018;Kim & Park, 2013).
Based on the eWOM of the New Zealand coffee industry, this study reveals the impact of eWOM on consumers' purchase intention from the perspectives of the volume, quality, valence, and consistency of the eWOM. This study found that all the four dimensions of eWOM will significantly impact consumers' purchase intention. Some of the dimensions are affected by the perceived credibility and usefulness of the intermediary role. The results show that the volume, quality, valence, and consistency of eWOM significantly positively impact consumers' purchase intention. Particularly, the result is consistent with previous studies (Al-Azzam, 2016;Boerman et al., 2017;Chang et al., 2015;Mtchedlidze, 2019;Saremi, 2014), which suggest that service and environmental elements as value-added attributes that contribute to consumers' overall consumption experience.

CONCLUSION
The study investigated the relationship between the main research variables of eWOM, perceived credibility, perceived usefulness and purchase intention in the New Zealand Cafe industry based on the research model of TAM. This research shows that the four dimensions of eWOM, i.e., volume, quality, valence, and consistency, have significant positive impacts on consumers' purchase intention.
This purchasing intention leads to generating purchase decisions and post-purchase behaviors (Cantallops & Salvi, 2014). Besides, some dimensions will be mediated by perceived credibility and perceived usefulness. The empirical results of this research help improve the TAM theoretical model to verify the role of eWOM in the coffee industry. This research provides a better way for managers in New Zealand cafes to understand how to operate eWOM. The knowledge about operating eWOM can replace traditional marketing methods, improve the industry marketing methods, and raise business performance (Seraj, 2012;Tidd & Bessant, 2018). The marketing of eWOM can enhance the market competitiveness of cafes. The New Zealand coffee industry should continually increase the volume of eWOM and the overall valence, quality, and consistency of eWOM. At last, future research could focus on the different sources of eWOM to increase consumers' purchase intention and could use more diverse measurement methods to increase the comprehensiveness of the research model.

THEORETICAL CONTRIBUTIONS
The main objective of this research is to examine the influence of electronic word of mouth (eWOM) on consumers' purchase intention in the New Zealand Cafe industry. The model of this study is based on the Technology Acceptance Model (TAM) used by Sussman and Siegal (2003), which intends to explore customer behavior in terms of their acceptance of information technology (IT). One of the most important theoretical contributions of this study was to apply TAM to studying online consumers' purchase intention in the New Zealand café industry. In addition, the modified model of this study explains the mediating effect of customer perceived usefulness and perceived credibility in the sphere of eWOM. Future research may use this theoretical model to investigate other restaurants, and comparison studies can be explored in different types of restaurants. Moreover， reliability and validity analysis has been performed to support the validity of the survey sample and data. Accordingly, future research on eWOM and purchase intention can replicate this model and develop more diverse aspects to measure.

PRACTICAL CONTRIBUTIONS
The study also provides practical implications for café managers, especially in electronic marketing.
The results of this study show that the volume, quality, valence, and consistency of eWOM have a significant positive impact on consumers' purchase intention. Furthermore, the modified research model has explained 29.7% of the total variance in the valence of eWOM, which implies that the café managers need to invest more attention to the positive and negative of online comments. In order to get more customers' positive online comments on products and services, timely interaction and communication with consumers are crucial in the online platform. For example, café staffs ask customers how they feel about services and products and guide them to post positive and truthful online reviews. On the order hand, consumers' perception of eWOM will indirectly influence their consumption choices in the New Zealand café industry. Firstly, perceived credibility has an intermediary effect on the volume, quality, valence, and consistency of eWOM. Therefore, consumer trust in eWOM may need cafe managers' more attention and investment. For instance, café managers launch reward campaigns to encourage consumers to post more positive online reviews. Secondly, perceived usefulness has an intermediary effect on the quality and valence of eWOM. This outcome suggests that the usefulness of eWOM to consumers is another concern of café managers in electronic marketing. For example, café managers present and update merchant information (e.g., opening hours, phone number, address), comprehensive menu, food pictures, dining environment pictures, parking information and other services provided on their online review website or social media.

LIMITATIONS AND DIRECTIONS FOR FUTURE STUDY
The limitations of this research have been organized and described as follows.
a) There were limitations in terms of data collection. The two primary sources of data collection in this research were students from the ICL school and consumers of Café shops. Participants from these websites are convenience samples, which means those samples could fail to represent a larger number of people. Therefore, the results of the survey in this research may be more intended to represent people who have received a higher level of education, students from the North Island of New Zealand and the Internet users rather than the whole population group of New Zealand. The research in the future could apply larger sample sizes and more complicated sampling strategies to improve the universality of the results. b) From the perspective of data analysis, the entire analysis process is relatively complex, which may easily lead to mistakes. This research tests step-by-step assumptions that may not systematically reflect the entire model. Future research could try to apply the SEM (Structural equation modeling) to test the mediation effect of eWOM. On the one hand, the operating procedure could be simplified, and on the other hand, the goodness of fit of the entire model could be examined.

RECOMMENDATIONS
With the rapid development of network information, new forms of eWOM have been continuously emerging (Tidd & Bessant, 2018). The organization's marketing concepts and strategies are constantly innovating in response to the changes in consumption patterns. eWOM marketing is based on the internet and takes advantage of interactivity between digital information and online media to assist in achieving the marketing objectives. It has been adopted by many organizations and has become an essential strategy for brand promotion and product promotion. Although there is not much eWOM marketing applied in the coffee industry in New Zealand currently, eWOM marketing will be widely applied in the future due to its features like across time and space, multimedia, interaction and personalization.
In order to make better take advantage of the effects of eWOM, according to the results from this research, in future eWOM marketing, companies should continuously increase the influence of eWOM and make the recipients trust and willing to adopt them through various channels. Companies should increase the eWOM marketing efforts, widely spread the positive eWOM and display eWOM, which has a high degree of relevance, comprehensiveness, and timeliness in the most noticeable positions. For example, companies can collaborate with famous review sites or search engines and publish large numbers of positive reviews of products through third-party organizations.