“Hi there, how can I help you? J”
In a digital world is being more human necessarily more trustworthy?
Experimental research investigating the influence of a brand’s communication
style on perceived brand trust in providing online customer care.
Name: Fleur Bourgonje
Student number: 11398485
Institution: Amsterdam Business School
Master Programme: MSc Business Administration: Marketing Supervisor: Mr. Abhishek Nayak
2
Statement of originality
This document is written by Student Fleur Bourgonje, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have
been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Table of Contents
... 2
Index of Tables and Figures ... 4
Abstract ... 5 1. Introduction ... 6 1.1 Problem Statement ... 8 1.2 Contributions ... 9 2. Literature review ... 11 2.1 Conversational Human Voice ... 12 2.2 Social presence in online service encounters & developing brand trust ... 14 2.3 Conversational norms & the impact of context ... 17 3. Method ... 20 3.1 Design ... 21 3.2 Pre-test ... 22 3.3 Measures ... 24 3.4 Sample ... 26 3.5 Computing Dummy Variables ... 27 4. Results ... 27 4.1 Testing for normality ... 28 4.2 Randomization & Perceived realism check ... 29 4.3 Correlations ... 30 4.4 Hypothesis Testing ... 31 5. Discussion of results ... 37 5.1 Conclusion & Discussion ... 37 5.2 Managerial Implications ... 42 5.3 Limitations & Future research ... 43 Appendices ... 46 Appendix 1: Survey design ... 46
4
Appendix 2: Real life customer-brand interactions on Facebook Messenger ... 58
Appendix 3: Survey pre-test ... 59
References ... 64
Index of Tables and Figures
Tables
1 Items conversational human voice scale 2 Items measures
3 Reliability analysis
4 Demographics of the respondents 5 Testing for normality
6 Correlation matrix
7 Results hypothesis 1 using PROCESS model 4 (I) 8 Results hypothesis 1 using PROCESS model 4 (II) 9 Results hypothesis 3 using PROCESS model 5 10 Results hypothesis 4 using PROCESS model 3 11 Summary results hypotheses
12 Results additional analysis using PROCESS model 7 Figures
1 List of the linguistic features used in an informal communication style 2 Conceptual model
5
Abstract
This study aims at understanding how a communication style of a customer service agent contributes to the development of brand trust of consumers, and particularly how the
consumption context influences this relationship. By means of a vignette study, the effect of brand communication style was analysed using a fictitious Facebook chat conversation between a customer and a customer service agent of a fictitious airline company. The participants of the study were positioned in a hypothetical situation, simulating either a hedonic or a utilitarian consumption context. Results revealed no direct effect of brand communication style on brand trust, but showed that the entire effect was mediated through social presence. Meaning that using a conversational tone of voice will only foster brand trust if it is perceived as a cue of social presence. Moreover, consumption context did not have an impact on the perceived appropriateness of the communication style nor did it influence the relationship between brand trust and communication style. This study expanded on previous research by exploring the relationship between brand communication style and brand trust within the scientifically rather unexplored communication medium of instant messaging. Moreover, it provides explanation for the relation between communication style and brand trust by introducing the mediating role of social presence. Organizations are recommended to use a conversational human voice in their online customer care as a cue to convey social presence, in order to ultimately enhance the perceived trustworthiness of the brand. To maximize the impact of the communication strategy organizations are advised to test for appropriateness of the communication style, as this influences the perceived social presence of the interaction.
Keywords: brand communication style, conversational human voice, instant
6
1. Introduction
A shift is taking place in the way customers are interacting with companies. In the last decade companies predominantly interacted with consumers publicly on their Facebook walls or on Twitter. However, over the past three years the majority of “social customer care” has moved to private, instant-messaging channels such as Facebook Messenger and Whatsapp. The public posts on Facebook are in slow decline, and there has been a major rise in private instant messaging between customers and brands (Edwards, 2017; D’Silva, 2017). According to a study conducted by SocialBakers based on 3.1 million private messages and 1.5 million public wall posts, collected from 256 companies, the scale of private messaging is now five times the volume of public wall posts (Edwards, 2017). For example, in Q1 2016 KLM handled 5,000 wall posts and 35,000 private messages, and this proportion is also representative for other industries (Edwards, 2017). Although consumers still prefer to publicly express their frustrations or to share their best customer experiences with others on social media, when concerns asking brands simple questions they tend to prefer private messaging (Fabrizio, 2016). These simple questions generally concern store opening hours, inventory and current orders (Nielsen, 2016). As a result companies are increasingly
integrating their customer care with social messaging platforms, such as WeChat, Whatsapp and Facebook messenger. Through these messaging applications brands aim to provide personal service by providing solutions to problems, or sought-for information in a both effective and efficient manner (Verhagen, van Nes, Feldberg & van Dolen, 2014). This service is either provided by a human customer service representative or by an artificial intelligent chatbot. For example, KLM uses Facebook messenger to communicate flight information, share boarding passes and answer travellers’ questions (Frankwatching.com, 2017). Whereas ING is engaging with customers via Whatsapp to respond to their inquiries
7
(ING Website, 2017), and Suitsupply uses Whatsapp to assist their customers in finding a shirt and tie that looks well with a specific suit (Manen, 2017). Instead of attracting
consumers to their traditional touch points such as their own website, mobile applications, or even retail stores, tapping into social messaging platforms allows businesses to interact with consumers where they already are (Accenture, 2017).
In the past these social platforms were primarily used for exchanging messages between friends and acquaintances. However, nowadays more than one billion messages are exchanged between consumers and businesses on Facebook Messenger per month, and this number is expected to continue to grow (D'Silva, 2017). When companies entered these platforms a shift from a formal, corporate communication style towards a more informal, and humanlike style became apparent (Kelleher, 2009). On corporate websites, email
communications, and telephony businesses tend to communicate formally towards their customers. In contrary, on social media and instant messaging applications predominantly an informal and conversational style is used (Beukeboom, Kerkhof & de Vries, 2015), that simulates interpersonal communications (Kelleher, 2009). Kelleher (2009) was the first to refer to this style as the conversational human voice. Although the concept of private chat is not something new within the domain of customer service, the rise of social media has altered the way consumers and organizations interact (Dijkmans, Kerkhof, Buyukcan-Tetik &
Beukeboom, 2015). Therefore, it is critical to gain proper understanding of how customers wish to be addressed by brands in the specific setting of instant messaging applications.
The main motives for businesses to communicate in an informal manner are generally to establish a personal customer-brand interaction and convey perceptions of closeness (Delin, 2005). Although the usage of a conversational human voice was proven to be effective to foster trust, commitment, satisfaction, and control mutuality in (corporate) blogs (Kelleher & Miller, 2006; Kelleher, 2009; Sweetser & Metzgar, 2007; Dijkmans et al., 2015) , and
8
webcare communication (Van Noort et al, 2014) it remains unclear whether these results also apply in a private, and highly interactive customer care setting. Xu and Wyer (2010) indicated that message effectiveness is dependent up on the fit with the type of medium of
communication. Therefore, it is critical to evaluate whether engaging in a conversational human voice is also effective in the private messaging context. Moreover, Gretry, Horvath, Belei & van Riel (2017) indicated that in certain situations adopting an informal
communication style can also negatively influence the brand. For instance, when the consumer is unfamiliar to the brand, addressing consumers in an informal manner on Facebook can be considered inappropriate, and decreases the trustworthiness of the brand (Gretry et al, 2017). Previous academics have not yet investigated how and to what extent the adoption of a conversational human voice can be employed to shape and improve online service encounters.
1.1 Problem Statement
Although the provision of customer service in a social messaging context is being adopted at a rapid pace, and seems to be taking a prominent role in a companies’ customer care provision the concept is still new and scientifically rather unexplored. As there is no prior research on how brands should interact with their customers on these messaging platforms, and since previous research has indicated that adoption of a conversational human voice does not always lead to positive outcomes, it is important for marketers to gain a thorough
understanding of the implications of adopting a certain communication style. Especially considering that for younger generations of consumers recognized as “digital natives” (Prensky, 2005), information exchange, brand perceptions, and purchase intentions are primarily evoked online on these social networking platforms (Köhler, Rohm, de Ruyter & Wetzels, 2011). The immense number of daily consumer-brand interactions on these channels
9
indicates the need for specific insights on the way companies should interact with these consumers to build relationships, and more specifically, to gain their trust. Therefore, the following research question will be investigated in this study:
To answer this research question an experiment will be carried out, by means of a vignette study. Respondents will be asked to first read a fictitious customer brand interaction on Facebook Messenger. Afterwards they are requested to evaluate the perceived
trustworthiness of the brand and social presence of the interaction. The customer will either be positioned in a hedonic or utilitarian consumption context, and the customer service agent will either use a corporate tone of voice or a conversational human voice. By positioning participants in hypothetical situations participants responses should reveal their perceptions and impression of the given interaction.
This study does not consider whether the service is provided by a human being, an artificially intelligent chatbot, or some hybrid form. Due to the rapid development of chatbots it is becoming more difficult for consumers to detect the difference. Moreover, it is predicted that in the future it will be rather impossible for consumers to detect whether they are
interacting with a human agent or chatbot (Messina, 2016), which makes it irrelevant to investigate this factor.
1.2 Contributions
This paper aims to make three empirical contributions. First, it aims to shed light on how brand communication style in a private customer care context influences perceived brand
RQ: How is the employed communication style in private online customer care provision related to the development of brand trust, and what role does the perceived social presence and
10
trust. Although previous research has shown that using a conversational human voice in contrary to a corporate tone of voice is effective in fostering trust, this has only been
investigated in public contexts such as blogs and on social media. The sheer volume of daily consumer-brand interactions through private messaging applications highlights the need for guidance on how brands should communicate on these platforms to gain the trust of the consumer. This study is different from prior research as it is the first to investigate the impact of communication style in consumer-brand interactions in a private, instant messaging
context. It will contribute to existing literature, by investigating the impact of communication style in a different and unexplored type of communication medium.
Secondly, this study is the first to investigate the role of a conversational human voice as a cue to convey social presence. Prior research has not studied the impact of
communication style yet in simulating social presence. As social presence was proven to be effective in fostering brand trust, this study will add to prior literature by providing
explanation for the relation between communication style and brand trust through social presence.
Thirdly, this study aims to identify to what extent conversational norms differ across consumption contexts, and how this interacts with the perceived appropriateness of a certain communication style in fostering brand trust. As prior research has shown that using a
conversational human voice can also lead to negative associations if considered inappropriate it is critical to gain insights on the impact of context. Although prior research has proved that the effectiveness of linguistic features can vary across hedonic and utilitarian consumption contexts (Kronrod, Grinstein & Wathieu, 2012; Kronrod & Danziger, 2013), the effectiveness of a communication style across these contexts is unexplored. Therefore, study will add to prior literature by investigating the effectiveness of a conversational versus a corporate tone of voice across a hedonic and utilitarian consumption context.
11
With regard to practical contributions, this study will provide companies insights on the impact of the style of a message, beyond its content, specifically within the context of online customer care through private messaging channels. Moreover, it will provide
companies guidance on the suitability of a certain communication style in social messaging environments, and whether this is dependent upon the consumption context of the consumer.
The remainder of this document is structured as follows; at first the current literature on the matter is reviewed and detailed descriptions of the main concepts of the study such as conversational human voice, social presence and brand trust are presented. Accordingly, hypothesis are formulated and subsequently the conceptual framework is introduced. The conceptual framework provides a clear overview of the study by illustrating the predicted relations amongst the main constructs. Hereafter, the research methodology of the study is presented. This includes an elaborate description of the design of the vignette study, the pre-test, measures and collected sample. Next, the result section presents findings from both preliminary analysis and hypothesis testing. At last, the conclusions and an extensive
discussion of the findings is presented. In this section, also the managerial implications of the results are discussed, as well as the limitations of the study and potential areas for future research.
2. Literature review
In this section, several key concepts will be explained to clarify the context of this research study. First of all, the concept of conversational human voice will be introduced and the operationalization of the concept will be clarified. Secondly, in order to understand the relation between communication style and brand trust the key aspects of realizing successful service encounters are discussed. Social presence is introduced as one of these key constructs and proposed as mediating variable. Thirdly, to understand the effectiveness of a
12
communication style the role of conversational norms is evaluated. Consumption context is introduced here as a potential moderator to provide explanation for the variance in
conversational norms and perceived appropriateness.
2.1 Conversational Human Voice
Social media has altered the way in which companies and consumers interact
(Dijkmans et al., 2015). In contrary to the traditional non-interactive media, characterized by one-way communications, according to Men and Tsai (2015) the “the interactive, personal
and communal characteristics of social media communication allow organizations to engage publics in conversations, supportive behaviours and meaningful relationships” (p.395). By
engaging in public conversations companies started to use a more human voice in their online communications (Locke, Searls, Weinberger & Levine, 2001). Kelleher (2009) defined this communication style as the conversational human voice; “an engaging and natural style of
organizational communication as perceived by an organizations public based on interactions between individuals in the organization and individuals in the publics.” (p.177).
Conversational human voice is the opposite of the formal communication style known as the “corporate voice”, by which a company tends to speak with one standardized tone (Kramer, 2001). A high degree of conversational human voice is seen when a company is open to dialogue, provides prompt feedback, not afraid to admit mistakes, treats others as human and incorporates a sense of humor in communication (Kelleher, 2009). Specifically the latter two elements are generally not associated with corporate communications (Kelleher, 2009). Message personalization and informal speech are the two main tactics mentioned in literature that allow organizations to establish this conversational human voice in their communications (Van Noort et al., 2014).
13
Message personalization concerns the extent to which a message is addressed
specifically to a certain individual (Walther, 2011). Both the recipient (i.e. the consumer) and the sender (i.e. the company) of a message can be subject to personalization. The objective of message personalization of the receiver is to invite consumers to a conversation (Kwon & Sung, 2011; Pollach, 2005; Willemsen, Neijens & Bronner, 2013). Examples of tactics to realize this include the usage of second-person pronouns like “you” and “your”, and addressing the receiver directly by their first name (e.g. “Hi Tom”). Personalizing of the organization refers to the usage of first-person pronouns like “I” and “we”. Moreover, companies identify their human representatives by name or photo to give the impression that the message is derived from individuals in the organization, rather than the organization that the individuals must represent (Pollach, 2005). By making use of individualized
communication rather than depersonalized communication, a perception is created of conversing with real people rather than a faceless organization (Kruikemeier, van Noort, Vliegenhart & van de Vreese, 2013), making it easier for customers to relate to (Pollach, 2005).
Informal speech is described as “casual and expressive language that is exchanged in everyday conversations and contrasts in this sense with the declarative and neutralized language that is often used in formal corporate communications” (Van Noort et al., 2014,
p.91). Whereas formal language tends to avoid slang terminology, and is applied in conformance with previously established professional rules and standards (Kramer, 2001; Pollach, 2005), informal language generally includes elements such as contradictions and abbreviations that make communication more efficient (Van Noort et al., 2014; Gretry et al., 2017). For instance, in informal speech one would rather say “thanks” instead of “thank you”, which is more typical for formal communications. Moreover, interjections like “oh” and “wow” which are generally used in spoken language, are also seen in informal
computer-14
mediated communications (Pearce, 2005; Pollach, 2005). Apart from that, in informal speech emoticons are often used to make up for the emotions and verbal nuances that cannot be conveyed online. According to Huang, Yen and Zhang (2008) usage of emoticons has a positive effect “on enjoyment, personal interaction, perceived information richness, and
perceived usefulness” (p.466). Based from findings in previous literature and an experiment
amongst undergraduate students in a social media context Gretry et al. (2007) identified 14 important linguistic features that characterize an informal communication style. Please refer to figure 1 for an overview of these linguistic features.
2.2 Social presence in online service encounters & developing brand trust
The nature of customer support services has drastically changed over the past decade due to the infusion of technology and the rise of Web 2.0 (Bitner, Brown & Meuter, 2000; Eisenfeld & Fluss, 2009). Services encounters’ generally entail a dyadic interaction in which sought-for information, or solutions to problems is provided to customers (Turel, Connelly & Fisk, 2013; Verhagen et al., 2014). It has moved from physical (face-to-face) and telephonic
Figure 1 - List of the linguistic features used in an informal
communication style (Gretry et al., 2007)
15
service encounters towards online service encounters through e-mail, mobile applications, online helpdesks, social media networks and live chat facilities with customer service agents. These interactions are critical in shaping a customer’s impression of a certain company or brand, and therefore key in developing trust and ultimately determining the success of the organization (Grönroos, 2000; Crosby, Evans & Cowles, 1990; Darian, Wiman & Tucci, 2005; Williams & Spiro, 1985).
Establishing brand trust is paramount to building relationships and developing
customer loyalty (Morgan & Hunt, 1994). Trust has been defined as “a willingness to rely on
an exchange partner in whom one has confidence” (Moorman , Deshpande & Zaltman, 1993,
p. 82). It is based on the perceived predictability of the brand’s performance, meaning that brand trust refers to the consumer’s trust that the brand will behave as expected (Rempel, Holmes & Zanna, 1985; Scanzoni, 1979). The development of trust and commitment are determined by the establishment of shared values and effective communication (Cyr, Hassanein, Head & Ivanov, 2007). When communicating with consumers companies are expected to be polite, responsive, helpful and understanding (Verhagen et al., 2014; Price, Arnould & Deibler, 1995). Apart from that, service encounters are not solely being evaluated by customers based on its outcomes (i.e. whether the requested information was given in a timely manner, or whether a solution to the problem is provided and compensation is offered), but also by the perceived interpersonal contact (Van Noort et al, 2014). Social presence and personalization have traditionally been considered fundamental in establishing interpersonal interactions, and thus also in creating successful service encounters (Bitner, Booms & Tetreault, 1990; Suprenant & Solomon, 1987; Czepiel, 1990).
Social presence entails the perceived feeling of human contact, warmth and sociability
(Yoo & Alavi, 2001), and is positively related to foster brand trust and ultimately loyalty (Cyr et al., 2007). In traditional physical and telephonic service encounters, social presence is
16
conveyed naturally. However, in computer-mediated encounters establishing the sense of human contact is more complicated due to the more impersonal and automated context (Head, Yuan and Archer, 2001), and greater physical and social distance (Schijns, 2003; Cyr et al., 2007). Therefore, it is critical for organizations to infuse the medium of communication with social cues that evoke feelings of social presence and simulate face-to-face communications (Qui & Benbasat, 2005; Gefen & Straub, 2004; Cyr et al., 2007). Examples of such social cues that have proven to be effective in evoking perceptions of high social presence in
computer-mediated communications are high interactivity (Keeling et al., 2010), personalized greetings, and inclusion of socially rich text context (Gefen & Straub, 2003). Conversing in a human voice online is suggested to resemble real-life encounters with employees from the organization (Kwon & Sung, 2011; Park, 2008). Moreover, Park and Sundar (2015) showed that emoticons in a business-to-consumer context served “to provide users a window into the
customer service agent’s emotions, leading them to believe that the agent understands their feelings and that the site is sensitive to customers’ needs” (p. 127). Therefore, interacting with
a conversational human voice with consumers could function as a social cue, and approximate a more real-life interaction (Dijkmans et al., 2015). Accordingly, this source of social
presence should provide assurance of the trustworthiness organization (Holzwarth,
Janiszewski & Neumann, 2006; Wood, Solomon & Englis, 2005), therefore the following hypotheses are proposed:
H1a: Customer service agents communicating in a conversational human voice create a
higher perception social presence, in comparison to customer service agents who use a
H1: Customer service agents communicating in a conversational human voice evoke higher
levels of perceived brand trust through the perceived level of social presence, as compared to customer service agents employing a corporate tone of voice
H1a: Customer service agents communicating in a conversational human voice create a
higher perception social presence, in comparison to customer service agents who use a corporate tone of voice.
17
Given the adoption of a conversational human voice was proven to be effective in fostering brand trust in corporate blogs (Kelleher & Miller, 2006; Kelleher, 2009) and public webcare (Van Noort et al, 2014), and the projected mediation effect of social presence on this relationship the following is hypothesized:
2.3 Conversational norms & the impact of context
Prior research has shown that interacting in a conversational human voice with consumers does not necessarily improve consumer-brand relationships. When considered inappropriate it can also mitigate the perceived trustworthiness of a brand (Gretry et al., 2017). Gretry et al. (2017) showed that in a social media context conversing with customers in an informal manner only leads to an increase in brand trust if the customer considers the communication style to be appropriate. For instance, customers who are unfamiliar to the brand generally consider an informal communication style to be inappropriate, leading to a decrease in brand trust (Gretry et al., 2017). These results can be explained by role theory, which substantiates that the realization of a successful interaction is dependent upon whether the participants behave appropriately according to conversational norms (Sarbin & Allen, 1968; Schewe, 1973; Solomon, Suprenant, Czepiel & Gutman, 1985).
H1b: High levels of perceived social presence in online customer service, positively
influences the perceived brand trust, in comparison to online customer service with low perceived social presence.
H2: Customer service agents communicating in a conversational human voice evoke higher
levels of perceived brand trust, as compared to customer service agents using a corporate tone of voice.
18
Conversational norms are “a set of nondeclared rules that lay out the optimal way to
communicate” (Kronrod & Danziger, 2013, p.727). These norms may vary depending on
social distance and context. For instance, people generally talk more formally to strangers than they would to their friends. By using a polite, formal way of communicating a certain distance is maintained (Brown and Levinson, 1987). The proper distance between the
participants is key in shaping the quality of the interaction, and sustaining a level of intimacy that is appropriate and comfortable (Kaitz, Bar-Heim, Lehrer & Grossman, 2004).
Furthermore, although it is inappropriate to exaggerate in formal documentation, in rhetorical speeches this is actually the norm, reflecting the variation of conversational norms across different contexts (Kronrod & Danziger, 2013). As predictability and consistency with expectations are the fundamentals of brand trust, a communication style that is congruent (versus incongruent) with one’s expectations should increase (versus decrease) perceived trustworthiness of a brand. Therefore, the following hypothesis is proposed:
Prior research has shown that relation between language use and perceived
appropriateness, and thus ultimately consumer behavior can be moderated by the consumption context. For instance, both figurative and assertive language were proven to be perceived as more appropriate and thus also effective in hedonic consumption contexts, as opposed to utilitarian contexts (Kronrod et al., 2012; Kronrod & Danziger, 2013). The difference in perceived appropriateness of the communication style indicates different conversational norms across utilitarian and hedonic consumption contexts. Hedonic experiences are characterized in prior literature to be more emotional in nature, whereas utilitarian
H3: The relationship between the communication style of the customer service agent and
perceived brand trust is moderated by the perceived appropriateness of the communication style. So that this relationship is stronger (weaker) for higher (lower) values of perceived appropriateness.
19
experiences are generally shaped by reason and logic (Alba & Williams 2013; Babin, Darden, & Griffin 1994;Dhar & Wertenbroch 2000; Drolet, Williams & Lau-Gesk 2007). In a
utilitarian consumption context consumers aim to make purchases in “an efficient and timely
manner to achieve their goals with a minimum of irritation” (Childers et al., 2001, p. 513). In
contrary, “hedonic value focuses on fun and playfulness, reflecting shopping’s potential
entertainment and emotional worth rather than the achievement of any prespecified end goal” (Babin, Darden & Griffin, 1994, p. 646). As a result, hedonic consumption generally tends to
trigger positive mood (e.g., Chaudhuri & Holbrook, 2001). As figurative and assertive language are associated with spoken language and conversation, and thus used as tactics to convey the conversational human voice (Gretry et al., 2017; Van Noort et al., 2014) it is therefore expected that similar results apply to this communication style. Moreover, as the conversational human voice is about conveying emotion and providing a pleasant experience by means of incorporating a sense of humor and emoticons in their communication this is expected to be more consistent with hedonic consumption in contrary to utilitarian
consumption. As in a utilitarian context efficiency is key, socialization is probably not desired and may be considered inappropriate. Therefore, a corporate tone of voice is expected to be more congruent with consumers expectations and needs in a utilitarian context. We propose interacting in a conversational human voice is more consistent with customers’ expectations - and therefore more effective and thus appropriate– in hedonic consumption than in utilitarian contexts. The following hypothesis is proposed:
H4: A conversational human voice (vs corporate tone of voice) is perceived to be more
appropriate in hedonic consumption contexts in comparison to utilitarian contexts. In turn, this will lead to higher levels of brand trust in hedonic consumption contexts as opposed to utilitarian contexts.
20
The hypotheses are compiled together in a conceptual framework presented below in figure 2. The model visualizes the proposed direct and in-direct relation between a brand’s communication style and brand trust. With regards to the in-direct relation, the impact of brand communication style on brand trust is predicted to be mediated by the perceived social presence. Moreover, looking at the direct relation between communication style and brand trust the conceptual model shows that this relationship is expected to be moderated by
perceived appropriateness. In addition, it illustrates that the perceived appropriateness in turn is predicted to be influenced by the consumption context.
Figure 2 – Conceptual Model
3. Method
In this section, the methodology of the research study will be discussed. First, the design of the experiment is presented. Secondly, a pre-test is introduced and evaluated in order to ensure successful execution and manipulation of the stimuli material used in the experiment. This is followed by introduction of the measures used to analyze the main constructs of the conceptual model, which are also tested on reliability. Next, the sampling
21
method and demographics of the sample are discussed. At last, the creation of the dummy variables for the different conditions is explained.
3.1 Design
In order to answer the proposed research question and test hypothesis an experiment was carried out, by means of a vignette study. A between-subject factorial research design was used to minimize the potential carryover effect between the different conditions. This entailed the creation of four surveys; 2 (conversational human voice/corporate tone of voice) x 2(hedonic/utilitarian consumption context), which were randomly distributed across the respondents.
Via an online survey developed with Qualtrics, respondents were given hypothetical situations positioning them in either a hedonic or an utilitarian consumption context.
Participants were asked to evaluate a customer service interaction on Facebook messenger with a fictitious airline company. The airline company either served a utilitarian goal in which the participant was assigned the role as business traveller, or a hedonic goal in which the participant was asked to imagine going on holiday. In order to prevent bias from prior experiences and brand associations a fictitious company was used. Depending on the condition the respondent was allocated to, the respondents were exposed to a chat in which the customer service agent either used a conversational human voice or a corporate tone of voice. Throughout the chat the company provides the customer with flight status updates and responds to customer’s inquiries. Respondents were then questioned on the perceived
appropriateness of the communication style, to what extend they experienced social presence and to evaluate the trustworthiness of the brand. By positioning participants in hypothetical situations participants responses should reveal their perceptions and impression of the given interaction. To ensure the participants fully understood the given chat conversation, and were
22
capable of identifying the minor language variations the survey was conducted in Dutch, the native language of the participants. Please refer to appendix 1, for the survey design that was used in this study.
The airline industry was chosen as a research setting for two main reasons. First of all, it was declared as the most “socially devoted” business sector by Socialbakers (2014). This indicates that airline companies most actively use social media in comparison to businesses from other sectors (Hvass & Munar, 2012). This demonstrates significant relevancy for this sector to gain more insights on the impact of adopting a certain communication style. Secondly, because the product of the airline industry could serve both a hedonic as well as a utilitarian consumption goal. Using a single product (but varying the consumption context in which it was taking place by means of message framing) granted the opportunity to remove any effects resulting from distinctive differences between a hedonic and utilitarian product (Kronrod & Danziger, 2013). Botti and McGill (2011) have shown that hedonic and utilitarian framing of identical consumption goods can be significant in conveying a different
consumption goal of one consumption good.
3.2 Pre-test
In order to test whether the manipulation of the communication styles was successful, and to assure internal validity a pre-test was conducted. A fictitious Facebook chat
conversation between a customer service agent of a fictitious airline company and customer was presented to the participants. Respondents were either presented the conversational human voice condition or the corporate tone of voice condition. Only the language use of the customer service agent was different across the two conditions, the design and information presented was kept identical in order to prevent these factors influencing the results. The list of linguistic features of informal communication developed by Gretry et al. (2017) was used
23
to formulate the speech of the customer service agent. Moreover, through social networks people were asked to share their past instant messaging conversations with airline companies, these were used as inspiration for the creation of the stimuli material. Please refer to appendix 2, for a few examples of these real life interactions with airline companies on Facebook Messenger. In appendix 3 the survey used in this pre-test is presented.
The participants were asked to rate 6 items selected from the validated conversational human voice scale developed by Kelleher and Miller (2006) on a 7-point Likert scale. A few examples of items from this scale are: “uses a conversational communication style”, “tries to
communicate in a human voice”, and “uses a sense of humor in communication” (Kelleher
and Miller, 2006), please refer to table for an overview of all the items used. Item
CHV-1 Air.EU tries to communicate in an interesting manner CHV-2 Air.EU uses a conversational communication style CHV-3 Air.EU communicates with a human voice CHV-4 Air.EU treats the customer in a human way
CHV-5 Air.EU tries to communicate in an entertaining manner CHV-6 Air.EU uses a sense of humor in communication
Table 1 – Items conversational human voice scale (Kelleher and Miller, 2006)
In total 60 participants took part in the pre-test, of which 26 were randomly allocated to the conversational human voice condition and 34 were allocated to the corporate tone of voice condition. At first a reliability analysis was conducted on the items of the
conversational human voice scale. A Cronbach’s alpha of .870 was reported, thus exceeding the minimum criteria of .70. As the corrected item-total correlation of all items were above .30, and deleting any scale item would not result in a higher Cronbach’s Alpha, the scale was considered to be reliable. For further analysis a scale mean was computed and coded as CHV_TOT. The results of the Shapiro-Wilk test of normality indicated that the scores of the scale mean were normally distributed (p=.505), and therefore the data was suitable for running parametric tests. An independent sample t-test showed that there was a significant
24
difference between the customer service agent using a conversational human voice (M=4.84, SD=1.15) and a corporate tone of voice (M=3.85, SD=1.04), t(58)=3.48, p=0.00. From this may be concluded that the manipulation of the communication styles was successful, and that respondents were able to distinguish between a conversational human voice and a corporate tone of voice.
3.3 Measures
In order to test the relationships amongst the different constructs of the conceptual model multiple validated measures were used. Some of the items of the original measures were slightly altered to fit the context of the research study. An overview of the items used in this study are presented in table 2. Moreover, the reliability of the measures was assessed by means of calculating the Cronbach’s alpha and analyzing the item-total correlation.
Brand Trust. The dependent variable Brand Trust was measured by using the validated three-item scale developed by Morgan and Hunt (1994). The items were scored by participants on a 7-point Likert scale (1=strongly disagree, 7=strongly agree). The scale was measured using the following items; “I feel I can trust brand X”, “I feel that brand x can be
counted on to help me and others”, and “Brand x appears to be reliable” (Morgan & Hunt,
1994). The cronbach’s alpha of .885 indicates high construct reliability.
Social Presence. To measure the proposed mediating variable Social Presence the validated five-item measure from Yoo and Alavi (2001) was used. Respondents were asked to rate the perceived social presence by scoring the experience with the customer service agent on five items using a bipolar 7-point Likert scale. Examples of these items are
“1=impersonal, 7=personal”, “1=corporate, 7=human”, and “1=cold, 7=human warmth”
25
Perceived Appropriateness. For the proposed moderating variable Perceived
Appropriateness, the three-item scale introduced by Gretry et al. (2017) was used.
Participants scored the appropriateness of the customer service agents’ communication style on the following three items: “is appropriate”/”meets my expectations”/”corresponds to how
I wish to communicate with a customer service agent” (Gretry et al, 2017). The items were
slightly altered from the original scale to fit the context of the research study. The participants rated the three items by using a 7-point Likert scale (1=strongly disagree, 7=strongly agree). The Cronbach’s alpha of .874 indicates a reliable scale.
As the item-total correlation of all the three measures reported above .30, and deleting any scale item would not result in a higher Cronbach’s alpha it can be concluded that all items and scales are good. Please refer to table 3 for an overview outcomes of the reliability
analysis. Scale means have been computed for Social Presence, Brand Trust, and Perceived
Appropriateness and were coded as SP_TOT , TRUSTTOT, and APPRTOT.
Table 2 – Items measures Cronbach’s alpha Brand Trust .885 Social Presence .917 Perceived Appropriateness .874 Table 3 – Reliability analysis Item
Brand Trust BT-1 I feel I can trust Air.EU
BT-2 I feel that Air.EU can be counted on to help me and others
BT-3 Air.EU appears to be reliable Social Presence SP-1 Impersonal – Personal
SP-2 Not social – social
SP-3 Insensitive – Sensitive
SP-4 Cold – Warm
SP-5 Corporate – Human
Perceived Appropriateness PA-1 The communication style of Air.EU is appropriate
PA-2 The communication style of Air.EU meets my expectations
PA-3
The communication style of Air.EU corresponds with how I wish to communicate with a customer service agent.
26
3.4 Sample
Since communication norms vary across countries and cultures (Ambady, Koo, Lee and Rosenthal, 1996; Yin, 2009; Shuter and Chattopadhyay, 2010), the population scope of this study is limited to one nationality, namely the Dutch consumers. A non-probability convenience based sampling method was used, as the sampling frame was unknown in advance. Respondents were recruited via social networks (e.g. Facebook, Linkedin) to collect as many responses as possible. In total 278 respondents took part in the vignette study. First, of all data cleaning was done to ensure the quality of the data. A check question was included in the survey to test whether the participants were actually paying attention. In between the questions regarding the social presence construct participants were presented the following question “A quick check if you are paying attention, please tick the box extremely agree”. The 10 participants who did not answer this question correctly were disregarded from the study. Moreover, the respondents were asked to what extend they perceived the given scenario to be realistic and whether they had difficulties imaging that they were the consumer in the given situation. The 37 respondent who indicated to have had difficulties with placing themselves in the given scenario were disregarded from the study. At last, since the population of this research study is the Dutch people who make use of instant messaging applications, the respondents who indicated that they do not make use of this type of communication medium were removed from the dataset. After data cleaning, there were 226 responses left for
analysis, and the minimum requirement of 50 respondents per condition was met.
Out of these 226 respondents 49.1% was male and 50.9% was female. They ranged in age from 19 to 74 years old (M=35.28, SD=13.78). Overall, the participants were highly educated, since 39.4% obtained a degree from an university of applied sciences, and 54% from a research university. For a complete overview of the demographic details of the respondents please refer to Table 4.
27 Frequency Percentage Gender Male 111 (49.1%) Female 115 (50.9%) Age 18-29 years 127 (56.2%) 30-44 years 30 (13.3%) 45+ 69 (30.5%) Education High School 4 (1.8%)
Vocational Education (MBO) 11 (4.8%)
University of Applied Sciences (HBO) 89 (39.4%)
Research University (WO) 122 (54.0%)
Condition
CTV x Utilitarian 55 (24.3%)
CHV x Utilitarian 59 (26.1%0
CTV x Hedonic 55 (24.3%)
CHV x Hedonic 57 (25.2%)
Table 4 – Demographics of the respondents 3.5 Computing Dummy Variables
All respondents were randomly assigned to one of the four conditions. For the analysis of the hypotheses two dummy variables were created. Communication style with the values “conversational human voice” (1) and “corporate tone of voice” (0). Consumption context with the values “hedonic” (1) and “utilitarian” (0).
4. Results
This section presents the results of the experimental study. First, the data was tested and reported on normality. Secondly, a randomization and perceived realism check was performed to ensure validity of the data and check for randomization of the control variables. Next, a correlation analysis was conducted to identify relationships across the different constructs and identify control variables. This was followed by the testing of hypotheses.
28
4.1 Testing for normality
First, the data was tested on normality. The items of Brand Trust, Social Presence and
Perceived Appropriateness had no normally distributed items as shown in table 5. All items
from Brand Trust, Social Presence and Perceived Appropriateness showed negative skewness between -1.15 and -0.5. The negative skewness indicates a clustering of scores at the high end. According to Tabachnick and Fidell (2014) skewness will not make a substantive
difference in the analysis when dealing with reasonably large samples. Looking at the kurtosis for all items from the Perceived Appropriateness construct, and the items BT1, BT3, SP1 and SP2, revealed positive kurtosis between 0 and 0.6. This indicates that the distribution is slightly peaked, with long thin tails. For the items SP3, SP4, SP5, and BT3 negative kurtosis is seen between -0.6 and 0, indicating a relatively flat distribution. As all the kurtosis and skewness values lie between the acceptable values of -2 and +2 (George & Mallery, 2010), and the sample consist of 200+ cases the risks from kurtosis and skewness are reduced, and therefore should not influence further analysis.
Shapiro Wilk-test M SD Skewness Kurtosis Statistic Sig. Brand Trust BT-1 5.10 1.243 -0.704 0.083 0.824 0.000 BT-2 5.18 1.315 -1.018 0.394 0.832 0.000 BT-3 5.15 1.205 -0.615 -0.056 0.847 0.000 Social Presence SP-1 5.05 1.588 -1.008 0.384 0.869 0.000 SP-2 5.07 1.561 -0.883 0.248 0.889 0.000 SP-3 4.48 1.449 -0.371 -0.204 0.941 0.000 SP-4 4.55 1.491 -0.469 -0.209 0.936 0.000 SP-5 4.58 1.685 -0.512 -0.582 0.922 0.000 Perceived Appropriateness PA-1 5.23 1.393 -1.143 0.585 0.897 0.000 PA-2 5.15 1.420 -1.057 0.311 0.849 0.000 PA-3 5.10 1.477 -1.037 0.286 0.896 0.000
29
4.2 Randomization & Perceived realism check
In order to check whether the randomization was successful across the four conditions, the distribution of gender, age and education was investigated. The distribution of gender and education amongst the different conditions was checked by conducting Chi-square tests, as this test is robust for non-normal distributed datasets. No significant difference between the conditions was found with regards to gender (χ2 (3)=1.424, p=.70) and education (χ2
(9)=8.538, p=.48). From which can be concluded that the randomization was successful for gender and education.
As the data is not normally distributed and therefore the assumptions of one-way ANOVA are not met, the non-parametric alternative Kruskal-Wallis H test was used to test the variation of age amongst the different conditions. The Kruskal-Wallis test checks for significant differences on a continuous dependent variable, in this case age, by a categorical independent variable that consists of at least two groups. In this case the independent variable was the condition the respondent was allocated to. The Kruskal-Wallis H test showed that there was a statistically significant difference in age across the four different conditions, χ2(3)=10.777, p=.013. Meaning that the distribution of age was not similar across the four conditions, from which can be concluded that the randomization was not successful for age. Therefore, age must be kept as control variable during further data analysis.
To verify that the manipulations of the communication styles and consumption contexts did not differ in terms of perceived realism, a realism check was included in the study. The participants were asked to rate the perceived realism of the given scenario on a seven-point scale anchored by 1=”Not realistic at all”, 7=”Very realistic”. The non-parametric Kruskal-Wallis H test was conducted to test whether the perceived realism did not differ across the different conditions. There was no statistical difference between the four different
30
conditions, χ2(3)=1.548, p=.671.The total mean score was 5.56, which verifies that the manipulations were successful in terms of being perceived as realistic.
4.3 Correlations
A Spearman rho test was used to test the correlations amongst all the variables since the data was not normally distributed. The mean, standard deviations and correlations of the variables are presented in the correlation matrix in table 6. A weak negative relation was seen between the variables age and social presence (r= -.155, p<.05), as well as for gender and social presence (r=.133, p<.05). Although the effect size of these correlations for both gender and age account for only 2% (r2socialpresence-age=.024, r2socialpresence-gender=.017) of the variation in mean social presence scores among the different respondents, to prevent this relation
influencing further analysis age and gender must be used as a control variables.
A moderate positive relation between communication style and social presence was reported, r=.383, p<.01 with a r2=.146. This suggests that the perceived social presence is generally higher for a conversational human voice (value=1), in comparison to a corporate tone of voice (value=0), however there is need for further analysis to confirm this. Moreover, a correlation of r=.555 (p<.01) with a r2=.308 is reported between brand trust and social presence. This indicates a rather strong positive relation between the two concepts, as
approximately 30% of the variance in mean score of brand trust amongst the respondents can be explained by the relation between brand trust and social presence. A weak negative relation between communication style and perceived appropriateness was revealed, r=-.144, p<.05 with a r2=.021. Moreover, a rather strong positive relation is seen between the
perceived appropriateness and brand trust, r=.524, p<.01 with a r2=.275. At last, a moderate positive relation between perceived appropriateness and social presence is identified, r=.293, p<.01 with a r2=.086. Although a relation between perceived appropriateness and social
31
presence was not previously hypothesized, further investigation will be done to further examine this relationship.
Variables M SD 1 2 3 4 5 6 7 8 1. Gender .51 .50 - 2. Age 35.28 13.78 -.205** - 3. Education 4.46 .67 -.231** -.071 - 4. Communication Style .51 .50 .035 -.097 .029 - 5. Consumption context .50 .50 .000 .061 .061 -.009 - 6. Brand Trust 5.14 1.14 .073 -.024 -.064 .012 .016 (.885) 7. Social Presence 4.75 1.35 .133* -.155* .008 .383** .074 .555** (.917) 8. Perceived Appropriateness 5.16 1.28 -.045 -.067 -.026 -.144* .068 .524** .293** (.874) * Correlation is significant at the 0.05 level (2-tailed)
** Correlation is significant at the 0.01 level (2-tailed) Table 6 – Correlation matrix
4.4 Hypothesis Testing
Hypothesis 1. In order to examine hypothesis 1, a mediation analysis was conducted by means of the Hayes (2012) PROCESS model 4. In this model, Brand Trust (Y) was inserted as the dependent variable, Communication Style (X) the independent variable, and
Social Presence (M) as proposed mediating variable. Age and gender were used as control
variables. All assumptions to carry out the analysis were met except the assumption of normality. However, the robustness of the test and large sample sized allowed to run the regression. The results of the mediation analysis are presented in table 7 & 8.
Consequent
Social Presence (M) Brand Trust (Y)
Antecedent Coeff. SE p Coeff. SE p
Communication Style (X) α1 .957 .1686 .0000 c1’ -.4799 .1274 .0002 Social Presence (M) --- --- --- b1 .5552 .0475 .0000 Gender (control) .2080 .1699 .2220 -.0685 .1203 .5698 Age (control) -.0097 .0062 0.1203 .0026 .0044 .5543 constant i1 4.6342 .2548 0.000 i2 2.6187 .2893 0.000 R2=.1524 R2=.3848 F(3,221) = 13.2490, p<.001 F(4,220) = 34.3950, p<.001 Table 7 – Results hypothesis 1 using PROCESS model 4 (I)
32
Effect SE p LLCI ULCI
Direct effect c1’ -0.4799 0.1274 0.0002 -0.731 -0.2289
Total effect c1 0.0516 0.1512 0.7334 -0.2464 0.3495
Boot SE Boot LLC1 Boot ULCI
Indirect effect a1b1 0.5315 0.1244 0.3121 0.7948
Table 8 – Results hypothesis 1 using PROCESS model 4 (II)
The effect of the communication style on social presence a1=.957, indicates that a customer exposed to a customer service agent using a conversational human voice in
comparison to a customer who is exposed to a customer service agent using a corporate tone of voice are expected to differ by a=.957 units in perceived Social Presence. The sign is positive, meaning that those exposed to a conversational human voice are estimated to perceive higher levels of social presence; thus greater feelings of human contact, warmth and sociability. This effect is statistically different from zero, t=5.6790, p=.000, with a 95% confidence interval from .6251 to 1.2895, providing support for hypothesis 1a.
The effect b1=.555 indicates that a customer who experiences one unit level of Social Presence higher than another customer (regardless of the communication style they were exposed to) are estimated to differ by b1=.555 units in their perceived brand trust. The sign of b is positive, meaning that those who perceive relatively higher levels of social presence are expected to perceive the brand as more trustworthy. This effect is statistically different from zero, t=11.6904, p=.000, with a 95% confidence interval from .4616 to .6488, providing support for hypothesis 1b.
The indirect effect of a1b1=.532 indicates that a customer who is exposed to a customer service agent using a conversational human voice in comparison to a customer exposed to corporate tone of voice are estimated to differ .532 units in brand trust. This as a result of those exposed to a conversational human voice tend to perceive greater feelings of social presence, which in turn generates more brand trust. This indirect effect is statistically
33
different from zero, as revealed by a 95% BC bootstrap confidence interval that is entirely above zero (.3121 to .7948). Therefore, hypothesis 1 is accepted.
The direct effect of communication style on brand trust, c’=-.48, indicates that a customer who is exposed to a customer service agent using a conversational human voice but that experiences the same level of social presence as a customer exposed to a corporate tone of voice is estimated to differ by c1’=-.48 units in their perceived brand trust. The negative sign indicates that a customer exposed to a conversational human voice perceives a brand to be less trustworthy in comparison to a customer exposed to a corporate tone of voice with similar levels of social presence. This direct effect is statistically different from zero, t=-3.7672, p=0.000, with a 95% confidence interval from -.7310 to -.2289.
Hypothesis 2. The results from the mediation analysis also show the total effect of communication style on brand trust, c1=.0516, p=.7334. This indicates that there is no significant effect from communication style on brand trust, but that the entire effect is mediated through social presence. Therefore, hypothesis 2 is rejected. This is also supported by the non-parametric Mann-Whitney Test that was carried out to determine whether there is a significant difference in perceived brand trust between people exposed to a customer service agent using a conversational human voice or a corporate tone of voice. From this test can be concluded that the brand trust of the participants exposed to the conversational human voice was not significantly different from the participants exposed to the corporate tone of voice (U=6290, p=.854). Therefore, hypothesis 2 is rejected; exposure to online customer service agents using a conversational human voice has not led to significantly different levels of perceived brand trust, thus also not significantly higher levels in comparison to a corporate tone of voice.
Hypothesis 3. In order to examine hypothesis 3, a moderation analysis was conducted by means of the Hayes (2012) PROCESS model 5. In this model, Brand Trust (Y) was
34
inserted as the dependent variable, Communication Style (X) the independent variable, Social
Presence (M) as proposed mediating variable, and Perceived Appropriateness (W) as
proposed moderating variable. Again, gender and age were included in the analysis as control variables. The results from this analysis are presented in table 9. Consistent with previous findings, the results from the PROCESS model 5 test confirm that there is no significant relation between communication style and brand trust, cx=-.1384, p=.2489. A significant relationship is identified between perceived appropriateness and brand trust, cw=.4884, p=.0000. This indicates that someone who experiences one unit level of perceived
appropriateness higher than someone else (regardless of which one of the two communication style they were exposed to) are estimated to differ by cw=.4884 units in their brand trust. The sign of the coefficient is positive, meaning that those who perceive relatively higher levels of appropriateness are expected to perceive the brand as more trustworthy. However, there is no interaction taking place between the communication style and perceived appropriateness, c=.0151, p=.8916. Thus, there is no moderation taking place of the perceived appropriateness on the relation between communication style and brand trust, and therefore hypothesis 3 is rejected.
Social Presence (M) Brand Trust (Y)
Antecedent Coeff. SE p Coeff. SE p
Communication Style (X) .9573 .1686 .0000 -.1384 .1197 .2489 Social Presence (M) --- --- --- .5552 .0475 .0000 Appropriateness(W) --- --- --- .4884 .0939 .0000 Comm.styl. * appropr. .0151 .1105 .8916 Gender (control) .2080 .1699 .2220 .0527 .1060 .6197 Age (control) -.0097 .0062 0.1203 .0056 .0039 .6197 constant 4.2858 .3858 0.000 3.1264 .3013 0.000
Table 9 - Results hypothesis 3 using PROCESS model 5
Hypothesis 4. In order to examine hypothesis 4, a moderation analysis was conducted by means of the Hayes (2012) PROCESS model 3. In this model, Brand Trust (Y) was
35
Appropriateness (M) as proposed moderating variable, and Consumption Context (W) as
proposed second moderator. Again gender and age were included in the analysis as control variables. The results from this analysis are presented in table 10. There is no significant direct relation between consumption context and brand trust, cw=.1721, p=.3242. Moreover, the consumption context did not interact directly with the communication style, cint_2=-.4261, p=.0824. Nor did it interact indirectly with the communication style through the perceived appropriateness, cint_4=-.0384, p=.8784. Therefore hypothesis 4 can be rejected.
Brand Trust (Y)
Antecedent Coeff. SE t p
Consumption context (W) .1721 .1741 .9882 .3242
Int_1 Comm.styl.*appropr. -.1905 .1913 -.9954 .3207
Int_2 Comm.styl.*consump. -.4261 .2442 -1.7451 .0824
Int_3 Appropr.* consump. -.0726 .1p63 -.3700 .7118
Int_4 Comm.styl.*Appropr.*Consump -.0384 .2508 -.1532 .8784
Gender (control) .1837 .1199 1.5322 .1269
Age (control) .0033 .0044 .7415 .4592
constant 4.4941 .2887 15.5688 0.000 Table 10 – Results hypothesis 4 using PROCESS model 3
Please refer to table 11, for a clear overview of all the outcomes from the hypothesis testing.
Hypothesis Result
H1: Customer service agents communicating in a conversational human voice evoke higher levels of perceived brand trust through the perceived level of social presence, as compared to customer service agents employing a corporate tone of voice
☑
H1a: Customer service agents communicating in a conversational human voice create a higher perception social presence, in comparison to customer service agents who use a corporate tone of voice.
☑ H1b: High levels of perceived social presence in online customer service, positively influences the
perceived brand trust, in comparison to online customer service with low perceived social presence.
☑ H2: Customer service agents communicating in a conversational human voice evoke higher levels of
perceived brand trust, as compared to customer service agents using a corporate tone of voice. ☒ H3: The relationship between the communication style of the customer service agent and perceived
brand trust is moderated by the perceived appropriateness of the communication style. So that this relationship is stronger (weaker) for higher (lower) values of perceived appropriateness.
☒ H4: A conversational human voice (vs corporate tone of voice) is perceived to be more appropriate in
hedonic consumption contexts in comparison to utilitarian contexts. In turn, this will lead to higher levels of brand trust in hedonic consumption contexts as opposed to utilitarian contexts.
☒
36
Additional Analysis: The interaction effect of communication style and perceived appropriateness on social presence. As the correlation analysis indicated a moderate
positive relation between the constructs of perceived appropriateness and social presence, this relation was further analyzed by means of a moderated mediation analysis using the Hayes (2012) PROCESS model 7. In this model, Brand Trust (Y) was inserted as the dependent variable, Communication Style (X) the independent variable, Social Presence (M) as
mediating variable, and Perceived Appropriateness (W) as proposed moderating variable. Age and gender were again used as control variables. There is evidence of an moderated mediation taking place, the regression coefficient for XW is a3i=-.5841 and is statistically different from zero, t(219)=-3.8574, p<0.01. Please refer to table 12 for the results of the proces model 7 test. Moreover, this model accounts for 36% of the variance in perceived social presence. The results indicate that the size of the indirect effect, the effect of communication style on the perceived social presence, is moderated by the perceived appropriateness. However as visualized in figure 3, although the perceived social presence is significantly lower when using a conversational human voice is considered not appropriate, this will always be higher in than when using a corporate tone of voice.
Consequent
Social Presence (M) Brand Trust (Y)
Antecedent Coeff. SE p Coeff. SE p
Communication Style (X) α1i 1.1825 .1495 .0000 c1’ -.4799 .1274 .0002 Social Presence (M) --- --- --- bi .5552 .0475 .0000 Appropriateness (W) a2i .9151 .1177 .0000 Comm.stye * approp. (XW) a3i -.5841 .1514 .0002 Gender (control) .2983 .1487 .0461 -.0685 .1203 .5698 Age (control) -.0038 .0055 -.6959 .0026 .0044 .5543 constant i1 3.7789 .3418 .0000 i2 2.7764 .3400 0.000 R2=.3614 R2=.3848 F(5,219) = 24.7907, p<.001 F(4,220) = 34.3950, p<.001 Table 12 – Results additional analysis using PROCESS model 7
37
Figure 3 – Visualization moderated mediation effect
5. Discussion of results
In this section findings are discussed and answer is given to the main research question of this study. First, the results from the experiment are reviewed per hypothesis and the
outcomes are discussed and compared to findings from prior literature. Thereafter, a discussion on the managerial implications of these findings is presented. At last, the limitations of study are discussed and suggestions for future research are given.
5.1 Conclusion & Discussion
The objective study of this study was to show the potential effects of a brand’s
communication style in providing online customer care on brand trust by means of conducting an experiment to answer the following research question; How is the employed
communication style in private online customer care provision related to the development of brand trust, and what role does the perceived social presence and consumption context play
38
in this relationship? Specifically, this study aimed at understanding the mediating role of
social presence and the moderating role of perceived appropriateness of the communication style, and to what extent the perceived appropriateness of the communication style was influenced by the consumption context. This study aimed to fill the research gap by exploring the effectiveness of communication strategies used to foster brand trust in the unexplored but yet massively used context of instant messaging customer care, and across different
consumption contexts. In that sense, this research study to contributed to the scarce literature on instant messaging channels and expanded the understanding of the concept of
conversational human voice in a different context and medium of communications. The communication style used on these instant messaging channels resembles the informal, conversational communication style used on social media (Beukeboom et al., 2015). Businesses generally communicate in a more human and informal manner to establish
personal customer-brand interactions and convey perceptions of closeness (Delin, 2005). As this was proven to be effective in public wall posts on social media (Van Noort et al, 2014) and corporate blogs (Kelleher and Miller, 2006; Kelleher, 2009) it was hypothesized that the use of a conversational human voice in a private messaging context would influence the trust a consumer has in a brand or organization. Contrary to initial expectations, the study did not show significant difference in perceived brand trust between customers exposed to a customer service agent using a conversational human voice or a corporate tone of voice. Indicating that the results found by Kelleher (2009), Kelleher and Miller (2006), and Van Noort et al. (2014) did not seem to apply in the private instant messaging context. However, this result could possibly be explained by the trend that engaging in a conversational human voice has become more normative over the last couple of years in customer-brand interactions on social media and within the social messaging environment (Dijkmans et al., 2015). As a result, the presence of a conversational human voice in itself might not be sufficient anymore to foster
