People talk with people, not with brands : the effectiveness of employees as brand ambassadors on social networking sites

People talk with people, not with brands.

The effectiveness of employees as brand ambassadors on social networking sites

Research Master’s thesis

Graduate School of Communication

Research Master’s program Communication Science

Dr. Guda van Noort

Tessa van Dijk

6170455

June 27, 2014

Table of contents

Acknowledgements...3

Abstract...3

Introduction...4

Study 1... ...6

Communication of Employees on Facebook...6

Content or topics on Facebook...7

Communication style on Facebook...8

Media richness in Facebook...8

Personalization in Facebook…...9 Formality in Facebook...9 Emotion in Facebook...10 Method... ...10 Research units...11 Procedure...11 Coder training...11 Codebook... .12 Results...13

Conclusion and discussion...14

Study 2... ...15

Effectiveness of Brand Ambassadors versus Brand Communication...15

Strength of tie within SNS...15

Persuasive Knowledge...17 Presence on SNS...18 Method... ...19 Pretest...19 Participants... ...20 Stimuli materials...20 Measures... ...21 Control variables...22 Manipulation check... ...23 Procedure...23 Results... ...24 Manipulation check...24 Control variables...24 Testing hypothesis...24 Hypothesis 1 and 3...24 Hypothesis 2 and 4...24 Conclusion... ...27 General discussion... ...28

Theoretical implications and future research...29

Practical implications... ...30

References...32

Appendices...42

Appendix A - Tables...45

Appendix B – Figure...46

Appendix A – Codebook Study 1... ...47

Appendix C – Manipulated Facebook messages...52

Appendix D - Questionnaire Study 2... ...53

Acknowledgements

This master’s thesis was written on behalf of the Research Master of Communication Science at the University of Amsterdam. A content analysis and an experiment were conducted within this thesis. Both studies aimed to shed light on the effectiveness of employees as brand ambassadors on social networking sites, in terms of affective and behavioral reactions towards a brand. In particular, this study shed light on the influence of tie strength and perceived persuasive intent on this relationship. In this report, the development, implementation and results of this study are described in detail.

I would like to thank my supervisor Guda van Noort. It was due to her critical view that I was able to write this thesis. Furthermore, I want to thank my colleagues from Red Bull and Steven Marges, who is a social media manager at ClubJudge, for their support during the development of this study. My final thanks goes to all participants who have cooperated in this research.

Abstract

Social networking sites are increasingly used as marketing tool to reach possible consumers. But social platforms were made for people, not for brands. Marketers are starting to notice that they are not welcome on these platforms. Employees may be successful as brand ambassadors to reach and persuade target audiences within social networking sites. However, the role of employees as brand ambassadors is not studied yet. This study is an important step towards a better understanding of employees’ behavior on Facebook, in terms of content and style features of their communication. Results are used in a follow-up experiment to explain that employees’ communication result in better affective and behavioral responses towards the brand than other sources of communication. More specifically, the source of communication is expected to make a difference in effectiveness. This study shows that employees’ self-created content is more effective than content that they share from brands, than content created by brands, than no content at all. Perceived persuasive knowledge has a mediating role within this relationship. This current study contributes to our understanding of effective marketing activities on social networking sites. Implications and ideas for future research of marketing strategies in social networks are discussed.

Introduction

Social media is a serious business for marketers. Social networking sites (SNS) especially gain important place in brand communication and marketing (Mairinger, 2008; Social Embassy, 2013). Dutch brands expect a growth of 50% in their social media spending in 2014, while they adopted SNS as effective medium to retain and maintain attention from possible consumers (Social Embassy, 2013; Bansal & Rani, 2014). This is hardly surprising considering that 86 percent of consumers have an account on at least one social platform (Newcom, 2014). Facebook is by far the most popular network among Dutch consumers (50,5%). Consequently, brands are also largely represented here (97%; Verhoeven, 2012). The question that remains is if the large investments pay off for marketers. It is not clear whether a brands’ presence on Facebook leads to more effective reactions by consumers than not being present, especially since users do not always appreciate the presence of brands within their personal social space (Clemons, Barnett & Appadurai, 2007; Joinson, 2008; Verhoeven, 2012).

SNS were made for people, not for brands (Fournier & Avery, 2011). Pursuing new relationships and maintaining existing ones are important motivations for people to spend time on SNS (Joinson, 2008; Urista, Dong & Day, 2009). Users are often annoyed by brand intrusions and may be resentful of what they perceive to be an encroachment in their personal space (Croft, 2008; Nitins & Burgess, 2014; Yang & Brown, 2013; Boyd & Ellison, 2013). SNS lose their appeal to users as they become dominated by commercial brands (Boyd & Ellison, 2007). Utilizing employees as brand ambassadors on SNS appears to be a logical and effective strategic solution to get around the resistance of users (Verhoeven, 2012). Brand ambassadors act in support and on behalf of a particular brand (Fisher-Buttinger & Vallaster, 2008). Employees may be experienced as more trust-worthy and less persuasive than traditional advertisements when they act as ambassadors (Wright, 2002).

The main aim of this study is therefore to investigate if employees are indeed effective as brand ambassadors on Facebook. It is important to understand employees’ brand communication on Facebook in order to study its effectiveness. The first aim of this study is to create an overview of the brand content or topics and style features that employees post on Facebook, since communication within SNS can differ in terms of content and style (Verhoeven, 2012; McCallister, 1997). So far, only Verhoeven (2012) has studied employees’ use of topics within Twitter. This study showed that topics about daily

work experiences was mostly used, followed by professional content and news about the brand. It is not clear whether the same categorization applies to Facebook, partly due to the fact that Facebook is a more social medium than Twitter (Huberman, Romero & Wu, 2009). This study will extend the study of Verhoeven (2012) by using his topics to identify brand-related communication on Facebook. Furthermore, communication style is identified among employees since prior research in the domain of computer-mediated communication showed that style features such as formality, emotion, personalization and communication richness are critical in the appreciation of messages (Hiltz & Turoff, 1978; Broadbent, 1977; Tang, 2006; Berger, 2013). With a better understanding of employees’ communication, the effectiveness of this communication can be studied.

The second aim of this study is therefore to show that employees’ communication is indeed more effective than ‘direct’ marketing communication from the brand itself. So far, literature demonstrated that employees can be effective in terms of reputation and corporate communications (Harris & De Chernatony, 2001; Deuze, 2007; Verhoeven, 2012), but no effect studies within the sales field are yet conducted. This is needed since the aim of companies is not only to improve their reputation, sales are just as important (Social Embassy, 2013). In examining these effects, this study adopts the framework of Persuasion Knowlegde (Friestad & Wright, 1994). This framework explains that communication is more persuasive when receivers are not aware it is aimed to persuade them. Employees may be able to get around the activation of persuasion knowledge since they could be recognized as ‘friends’ and not as ambassadors by connections within their social network (Granovetter, 1973). Moreover, employees’ communication may be more credible than brand communication since they have a stronger connection with the receivers of their messages (Gotlieb & Sarel, 1992; Jansen, Zhang, Sobel & Showdury, 2009).

Furthermore, it is plausible that people link brands to commercial purposes when they are confronted with brand communication in SNS. This may activate their persuasion knowledge (Friestad & Wright, 1944). This study suggests that a commercial link may also be made when employees share messages of brands. Persuasion knowledge is expected to be less activated for self-created posts, and more likely for shared posts of the brand. The third aim is therefore to examine whether posts created by a brand ambassador his or herself are more effective than when a brand ambassador shares content

that was originally created by the brand. It is important to make this distinction since marketers need to know how to implement employees effectively in their marketing strategies.

Lastly, it is important to know whether presence of employees’ brand communication in SNS is more effective than no presence at all. Employees may be more effective than brands and their personal posts may be more effective than shared ones. But most importantly, presence on Facebook needs to have more effect than no brands presence at all. Otherwise large amounts of money that marketers invest are wasted. Besides, this will contribute to the understanding of Facebook marketing since no comparison has yet been made between different sources of brand content and no brand presence at all. Previous studies only compared the effectiveness of Internet platforms like blogs, websites and social media platforms (Herring, Scheidt, Bonus & Wright, 2004; Kelleher & Miller, 2006). The fourth aim of this study is therefore to show that employees’ communication on Facebook is more effective than no presence on Facebook at all. The following research question was formulated to achieve all aims:

How do employees communicate as brand ambassadors on Facebook, in terms of content and communication style, and to what extent can effectiveness be explained by persuasive knowledge?

The current study attempts to answer this question. First, a content analysis is conducted of 912 Facebook messages to get a better understanding of the content and features of communication style used by employees on Facebook. Results of this analysis serve as starting point for an experiment in which the effectiveness of employees’ self-created and shared messages, brand initiated messages and no brand presence on Facebook are compared. In the next section, the first part of the research question is answered. The second part of the research question will be answered in Study 2 (Figure 1).

----insert Figure 1 about here---- Study 1

Communication of Employees on Facebook

Employees seem to have a powerful impact on consumer’s perceptions since they represent their organization with every move, especially on social networks (Harris & De Chernatorny, 2001; Schade, 2007). Employees form a personal and professional identity on SNS, to control the impressions that

others have of them (Daft & Lengel, 1987). Verhoeven (2012) demonstrated that one out of five tweets contain professional content within his Twitter analysis of employees. However, it remains ambiguous whether this content is effective within other social platforms. In order to contribute in the direction of understanding the effectiveness of employees’ communication, it is important to know how they actually represent themselves on SNS. A better understanding of their content and communication style on SNS is needed since these features have been demonstrated to be important and effective by previous studies (Palmer & Koenig-Lewis, 2009; Homans, 1958; Peters & Kashima, 2007). Therefore, content and communication style features are discussed in the next sections.

Content or topics on Facebook

Content can be described as the topics or theme of a message (Anderson, 1972). It is the most explicit element of persuasive communication (Williams & Spiro, 1985). Although a lot of studies focused on topics within SNS (Hughes, Rowe, Batey & Lee, 2012; Heinonen, 2011; Forrester, 2010), no studies shed light on employees’ topics within these networks. Only one prior research conducted by Verhoeven (2012) studied the topics that employees share about their company. The research context for this study was Twitter. Verhoeven (2012) categorized the 140 character messages into nine topics. Daily work (31,1%), in particular, was a frequent posted topic on Twitter. This was followed by professional content (28,1%), news about the brand (17,5%), internal contacts (6,1%), product/service (6%), persuasive message (4,5%), human resources (3,1%), costumer communication (1,8%) and external professional contacts (2,9%).

However, this current study will focus on Facebook, which is a different kind of network than Twitter. The focus within Twitter is on brand-related material (Diffen, 2014). Moreover, Twitter is mainly used to communicate to individuals with similar interests, regardless of whether users know one another offline (Kwak, Changhyun & Moon, 2010). Facebook is used to share stories with family and friends for entertainment purposes (Huberman, et al. 2009; Hughes, et al. 2012). Therefore, employees may communicate about different topics on Facebook than on Twitter. The topics of Verhoeven (2012) are distinguished in this study to have a better understanding of employees’ brand communication on

Facebook. However, other topics may be identified through open coding. Together, this will answer the following research question:

RQ1: How do employees post about their brand on Facebook, in terms of topics? Communication style on Facebook

Features of communication within messages may also differ in effectiveness, since the way that people put words together reflects a great deal about the writer and its influence (Tausczik & Pennebaker, 2010). Communication style can be defined as an individual’s particular pattern of communication (Barber, 1978). The way people write is key in understanding its intention, influences its effectiveness and the intention to forward this message (Chung & Pennebaker, 2007). This study looks more closely at features of communication styles within a computer-mediated environment. Previous literature distinguishes four important features within communication style that affects its effectiveness; media richness, formality, personalization and emotion (Mosquera & Moreda, 2014; Daft & Lengel, 1986; Kelleher, 2008; Ittelson, 1973). These features are used to shed light on the way employees talk about their company, they are discussed here one by one (Table 1).

----insert Table 1 about here---- Media richness in Facebook

Rich, personal communication means a generally more effective way of communication than less richer media (Daft & Lengel, 1986). Media Richness Theory explains that richer media, with more information, content or visuals are transferred in a clearer way to receivers than less rich media. Furthermore, rich media is proven to evoke more favourable attitudes and forwarding intentions by viewers and bring users of social networks more pleasure and positive satisfaction (Tang, 2006; Simon & Peppas, 2004). This theory can be applied to Facebook, since it offers the possibility to share a status, a photo, a video and a link within a message. These types of media represent different levels of media richness or vividness of an online post, since users can add up to four different types of media within their messages (Daft & Lengel, 1986). The richness of media could eventually determine differences in effectiveness. Therefore it is important to shed light on the way employees use media within their Facebook posts.

Personalization in Facebook

A better understanding of personalization within messages is also needed since this might increase the effectiveness of Facebook messages (Pollach, 2006). Personalization can be described as personal stories about the writer, using personal pronouns (e.g, we, me, you, us; Davis & Brock, 1975). People feel the need to share personal information on Facebook, Baron, Squires, Tench and Thompson (2005) demonstrated that Facebook users expressed personal pronouns when communicating about personal opinions, judgements or beliefs within their messages. Adopting a conversational human voice or communications based on one’s own experiences can be an essential element of building and maintaining relationships in a networked social platform (Kelleher & Miller, 2006). Self-expression has a positive effect on the forwarding intention of messages (Cheng, 2012). People are more likely to engage in word of mouth activities when they read personal stories from friends or people with whom they can identify themselves (Sun, Youn, Wu & Kuntaraporn, 2006). The question that remains is if employees use personal stories and personal pronouns on Facebook within brand communication. This study will therefore look at the usage of personal pronouns within posts of employees on Facebook.

Formality in Facebook

Formality is exists of a formal and informal style. A formal style is a bookish and proper communication style and an informal style which is more like spoken language, using short sentences (Maeda, Shimada & Endo, 2012). Moreover, informal communication is perceived as face-to-face communication. Facebook is an informal platform, people talk with friends or family they know offline (Diffen, 2014). Therefore employees may talk in an informal matter. Mosquera and Moreda (2014) studied formality by measuring if texts were similar to everyday language. They demonstrated that comments and greetings within messages on Facebook are mainly written in an informal matter. Computer-mediated communication is often written as if it was spoken (Raacke & Bonds-Raacke, 2008). Therefore people respond to this in a more spontaneous matter (Fish, Kraut & Chalfonte, 1999). Informality is demonstrated to have a positive effect on buyer-supplier relationships (Vijayasarathy & Rohey, 1997; Carr & Pearson, 1999). Formal communication, however, negatively influences distortion

and withholding of information (Mohr and Sohi, 1995). This study will look into employees’ use of formality within brand-related messages on Facebook, since no research studied this yet.

Emotion in Facebook

Emotion tends to play a prominent role in models of attention and perception. This study adopts a scale of 16 emotions from Chew & Eysenbach (2010) to identify emotions in Facebook messages. It is important to identify emotions within messages since such messages are perceived more readily (Ittelson, 1973; Broadbent, 1977). People are more likely to talk about emotional experiences on Facebook (Dobele, Lindgreen, Beverland, Vanhamme & Van Wijk, 2007). Emotion within ads allow consumers to experience sympathy (Batra & Stayman, 1990). Advertisements are more convincing and better liked when they contain emotion (Bagozzi & Moore, 1994). Consequently, these emotions spill over to the brand that is embedded in the message (Ahluwalia, Unnava & Burnkrant, 2001). Content that contains emotions is often shared, by both its senders and receivers (Berger, 2013). Stories that contain a surprise element are remarkable, so they make the sender look ‘cool’ for passing them on (Broadbent, 1977). This study wanted to shed light on the usage of emotions within brand-related Facebook posts, since this increases its effectiveness.

Thus, these four features (media richness, formality, personalization and emotion) of communication are used differently and cause different effects in SNS. The first study is therefore aimed to indicate which types of communication features employees use, while Study 2, later on in this research, will use these findings to test its effectiveness. The following research question is formulated:

RQ2: What communication style features do employees adopt in terms of media richness, formality, personalization and emotions?

Method

A content analysis of employees’ Facebook messages was conducted to shed light on the brand-related topics and the style features that employees use on Facebook. Employees from four different companies were approached. The three most successful social media companies of 2013 were included, namely, KLM (airline), ING (finance), and Cool Blue (IT) (Social Embassy, 2013). The brand, Red Bull (food and beverages) was added to this design since this company was also used in the follow up study. It is

useful to compare the first and second study for the same brand to avoid bias. Within this analysis it is important to compare brands from different sectors though, since this may cause differences in employees’ communication (Verhoeven, 2012).

Research units

A total of 20 active Facebook users who currently work for one of the companies were approached. Their Facebook messages are the units of this analysis. Messages of employees within the sample were collected from the 1th January until the 31th March 2014. A total of 912 Facebook messages were posted by the 20 active Facebook users. The final data included the name of the ambassador and the Facebook posts that the ambassador created, liked, shared or commented on. All messages that were not related to professional content, the job of its sender, or their company have been removed from the sample based on criteria developed by Verhoeven (2012). A total of 330 brand-related Facebook posts remained, this was 36,2% of the total sample.

Procedure

The employees were selected through the search option within LinkedIn (using the terms; KLM, ING, Cool Blue and Red Bull). In this way, employees within all positions were included to avoid bias in this study. LinkedIn profiles were used to find matching Facebook profiles, since 70% of the people who are present on LinkedIn are also active on Facebook (Echtzzp, 2014). Active users, who had at least 200 friends on Facebook and posted at least 20 messages within the last three months, were selected (Marketingfacts, 2014). Out of lists of employees, five Facebook users were randomly selected for each brand that fit these criteria. This was harder than expected since people protect their content on Facebook for those who are not connected to them. Therefore employees were sent a friend request by the researcher to get access, containing a message asking to participate in this study. A total of 26 employees were approached, of which six did not answer the friend request. Therefore 20 participants remained in the sample, equally distributed over the four brands.

Coding training

A coding training was conducted to improve the reliability of this analysis by means of theoretical instruction and practical advice. Ten additional brand messages from a Red Bull employee were

retrieved to conduct this training, these messages were not part of the final sample. The codebook was explained to a second coder by the researcher and both coded these ten messages. Reliability was high since Kappa was between 0,8 and 1 for the outcome variables, except for personalization (K = 0,4) and emotion (K = 0,41). These variables were improved by adding examples, information and omitting response options. Five more messages were then coded and Kappa increased respectively to 0,8 and 0,9. During the training process it was noticed that a distinction could be made between employees’ activities on Facebook, in terms of liking, sharing, commenting or creating a message. This distinction enabled us to shed light on self-created messages of employees and their shared messages of brands. Therefore, the activity that employees perform was added in the research design of this content analysis. Codebook

In order to categorize Facebook posts of the remaining 20 employees, manual coding was performed. The researcher coded all messages in this analysis based on the modified codebook. A codebook was created to code the messages, based on previous studies (Daft & Lengel, 1984; Chew & Eysenbach, 2010; Verhoeven, 2012;). Only few codes were added following a coding development strategy (Glaser & Strauss, 1967). The codebook was divided into (1) the initial source of the message, (2) the type of topic that was shared and (3) the communication style features that were used. Activity of the ambassador was coded into ‘like’, ‘share’, ‘commented’ or ‘self-created’ (Facebook, 2014). The source of the message was coded as employee (0) or brand (1). Type of content was measured based on nine categories made by Verhoeven (2012) for a similar Twitter analysis; daily work, professional content, news about the brand, internal contacts, product/service, persuasive message, human resources, costumer communication and external professional contact. An open answer option was also added. The codebook, with an operationalization of all items, can be found in Appendix C.

Communication style was coded according to four features. Firstly the richness of messages was measured. Facebook offers the possibility to share a status, a photo, a video and a link. The more media were used, the richer the message was rated, ranging from 0 (not rich) – 6 (rich; Daft & Lengel, 1984). Personalization of messages was coded by answering the question: Are first-person pronouns used in the message? (Pollach, 2006; Davis & Brock, 1975). Both items were coded into personal (0)

and not personal (1). Thirdly, formality was coded into formal (0) or informal (1) based on definitions by Maeda et al. (2012). Next, emotions within Facebook messages were coded. Sixteen emotions were distinguished on an ordinal level (Chew & Eysenbach, 2010). The open answer option was added to the scale. Lastly, messages were coded on their sentiment into positive, neutral or negative, to identify the meaning of the used emotion (Koppel & Schler, 2006).

Results

A total of 330 brand-related messages were coded. The sample was skewed, 46,1% of all messages were derived from Red Bull employees. This means that employees of this organization post more content than ambassadors of KLM (24,7%), CoolBlue (26,5%) and ING (9%). Moreover, content of Red Bull employees was more often related to their company (48,2%) than KLM (35,6%) Cool Blue (26,5%) and ING (9%; Table 2). A great amount of messages were posted in Dutch (68%), the remaining part was written in English. Results show that employees performed various activities; 38% of them liked, 13% shared, 3% commented and 39% posted self-created content. The content was most often shared by friends (60,4%), the remaining part (39,6%) was shared from a brands’ Facebook or website. These activities made it possible to compare brand posts and posts initiated by employees.

----insert Table 2 about here----

This aim of this study is to have a better understanding of employees’ brand communication on Facebook in terms of content. Of all messages, the most frequent topics were about the daily work experiences (25,8%), followed by personal activities related to the brand (15,2%) and news about the brand (11,3%), persuasive posts (12,5%) professional content (10,9%; Table 3). Self-created posts by employees were mainly about daily work occasions (28,3%), followed by professional content (10,9%). Posts shared from the brand were mainly about daily work (39,3%), professional content (25%) and less about personal activities 14,3%.

The second research question asked what features of communication style employees use on Facebook. Messages contained on average all four types of media (M = 3.85, SD = 1.38), mostly text (82%) and photos (68%). Almost 80% of all messages were written in an informal way, only 4,3% was written in a formal style. The informal communication style was mostly used within both brand posts

(69,3%) and the self-created posts (84,7%). Furthermore, only 30,1% of all the messages was written in first person, for self-created messages (28,7%), but less for shared posts (7,1%). No messages expressed negative sentiment or emotion since 68,8% of the messages contained a positive sentiment and the remaining messages were neutral. Excitement was mostly expressed within all Facebook messages (23%), followed by joy (18%) and humor (14,5%). However, self-created posts did more often express emotion (68,1%) than shared posts (42,9%).

Conclusion and discussion

A content analysis was conducted to demonstrate the important role of employees’ communication on Facebook. When content and features of communication styles that were mostly used by employees are identified, a follow up study can be conducted to measure the source of effectiveness on Facebook. The analysis already shed light on shared messages of brands, which made it possible to look at brand-initiated posts as well. The analysis shows that approximately 40% of all retrieved Facebook messages were brand-related. This is more than on Twitter, Verhoeven (2012) showed that 33,3% of employees’ tweets was brand-related. This was not expected, as Twitter focuses more on professional content and Facebook is used to share personal experiences (Kwak, Changhyun & Moon, 2010). The sample frame was skewed, this may due to the motivation of employees themselves or the sector in which the employees work (Verhoeven, 2012). Nevertheless, the analysis yielded important results.

Results show that employees’ brand-related content on Facebook is slightly similar to topics Twitter. On Facebook, employees mostly post about daily work experiences, which correspondents with Twitter (Verhoeven, 2012). However, a new category, ‘personal activities’ (activities outside the workplace, but related to the organization) was added through open coding. This is the only different topic compared to Verhoeven (2012). However, range order of topic use differed. In particular, persuasive messages, product/service topics and human resources were more frequently used on Facebook compared to Twitter. On the other hand, news about the brand and professional content were more common on Twitter. All in all, employees posts (self-created and shared) were mostly about daily work experiences both on Facebook and Twitter.

Self-created messages from employees also seem to have the same style as shared posts of brands. Both use an informal communication style, as expected by Diffen (2014). The emotions of joy and excitement occurred the most. Employees’ posts were highly personalized, as Facebook was intended to be (Kelleher 2008). However, self-created messages contained much more personal pronouns (e.g., we, me, they) than shared messages. Employees use an informal, personal style on Facebook, adding an excitement to the message. The results of this content analysis are used to create the stimuli materials for Study 2, wherein the effectiveness of these messages is measured.

Study 2:

Effectiveness of Brand Ambassadors versus Brand Communication

Researchers do suggest high effectiveness of employees’ communication on SNS (Harris & De Chernatorny, 2001; Schade, 2007; Verhoeven, 2012), but actual effects were never shown. This study therefore tests this effectiveness, by comparing self-created posts by employees, brand posts shared by employees and brand posts on Facebook. Serra & Soto-Sanfiel (2014) suggest that brand content retrieved from friends creates more word-of-mouth than content retrieved from brands, since friends could be more relevant to receivers and may be less persuasive (Friestad & Wright, 1994). Moreover, less personal notifications (shared content) may generate more interest than advertisements of the brand itself, but less interest than self-created updates by friends (Thomson & Sinha, 2008; Reynolds-McInay & Taran, 2010). Based on these suggestions, this study will focus on the effectiveness of self-created and shared content of employees and brand posts on Facebook. Tie strength and persuasive knowledge will be discussed to explain this.

Strength of tie within SNS

Effectiveness of Facebook messages could be explained by its source, since perceptions about this source influences its effectiveness (Eagly & Chaicken, 1993; Petty & Cacioppo, 1986). The sources or senders of messages that are studied in this research have are connected to their receivers within social networks, described as social ties; something or someone with whom you have a connection or relationship (Boyd, 2004; Dwyer, 2007; Ellison, Steinfield & Lampe, 2007). Social connections may be created offline or online and may differ in their nature, as they represent weak and strong ties (Carrasco

& Miller, 2006; Ellison et al. 2007). Weak ties are loose connections between individuals who are not close and do not provide emotional support for one another. Strong ties are connections between people who discuss important matters or regularly keep in touch, like family or friends (Granovetter, 1973).

Previous studies show that strength of tie has influence on the effectiveness of communication messages. Strong ties are claimed to be more accessible and willing to be helpful (Krackhardt, 1992), which leads to greater knowledge exchange (Ghoshal, Korine & Szulanski, 1994; Hansen, 1999). People tend to give more attention to messages from people they know (Phelps, Lewis, Mobilio, Perry & Raman, 2004). Tie strength was also found to be a predictor of behavioral responses in form of receiver’s purchase decisions (Bansal & Voyer, 2000) and forwarding intentions of both commercial and non-commercial emails (Huang, Lin & Lin, 2009; Norman & Russell, 2006). However, Chiu, Hsieh, Kao and Lee (2007) showed that people are more inclined to forward marketing messages received from a strong tie, than from a weak tie or a commercial source. Phelps et al. (2004) confirm this and add that strong ties are perceived as more credible than weak ties or commercial sources. Hence, strong ties may be more effective since people are more willing to listen and absorb knowledge of trusted sources (Hovland & Weiss, 1951; Levin & Cross 2004).

Employees can be described as strong ties as they created their own network of connections on Facebook (Phelps et al. 2004). Their connections may perceive them as friends and not as ambassadors. Charron, Favier & Li (2006) state that messages are more effective when they are received from strong connections. Brand-related content from friends may generate interest and the desire to follow up with this. However, this may be different for weaker connections (Reynolds-Mcinay & Taran, 2010). Brands or commercial sources can be described as weak ties since they do not support people in an emotional way (Granovetter, 1973). Thus, employees may have more influence on people within their Facebook network than brands, while they are more trusted by their public (Brown & Reingen, 1987).

The effectiveness of employees’ communication may also differ between their messages since they can conduct different activities on Facebook. Employees can create posts themselves, but they can also share messages from their brand on their personal timeline through the ‘share button’ on Facebook (Facebook, 2014). The original source will always be visible within a shared message (Sanghvi, 2006). This makes identification of the actual source of the message ambiguous for its receivers. For instance,

a message that was originally posted by the brand, with whom receivers have a weak connection, is now shared by a friend, with whom receivers are strongly connected. Perceived credibility of a friend’s message is high, but trust may decrease when receivers notice that a message was originally written by a brand. Effectiveness of shared posts may therefore fall between brand initiated posts and self-created posts on SNS (Serra & Soto-Sanfiel, 2014). The following hypotheses are formulated:

H1: Exposure to brand-related messages from employees will lead to more positive affective and behavioral responses than exposure to messages from brands.

H2: Exposure to brand-related messages from employees will lead to more positive affective and behavioral responses than messages from brands.

Persuasive knowledge

The relationship between the source of Facebook messages and receiver’s reactions can be further explained by the underlying mechanism of Persuasive Knowledge (Friestad & Wright, 1994). People are less likely to perceive messages as persuasive when they receive them from people they know, since strong ties are more credible than weak ties (Rogers, 1995; Pornpitakpan, 2004; Brown, Dacin, Pratt & Whetten, 2006). People develop knowledge about how, why and when a message is intended to influence them, to help them cope with persuasive content (Friestad & Wright, 1994). This perceived persuasiveness will influence the effectiveness of messages. Van Noort, Antheunis and van Reijmersdal (2012) explain that viral SNS campaigns will be perceived as more credible and less likely as persuasive when they are retrieved from a person with whom the receiver has a strong connection. Consequently, people who do not perceive a campaign as persuasive hold more positive attitudes towards the brand and the campaign and have a higher forwarding intention.

In addition, the very presence of a sales agent or an advertising text causes the activation of persuasive knowledge (Cialdini, 1984; Campbell, 1995; Ahluwalia, Burnkrant & Unnava, 2014). We expected that the presence of a brand on Facebook in form of a fan page would cause activation as well, as people have previous knowledge about its commercial purposes. It is less likely that people activate their “market-place-related social intelligence” when friends within their social network posts the same message as receivers may not expect that friends will try to influence them (Rogers, 1995). Employees

may therefore be able to conceal the persuasive intent of their messages. However, the original source of the message is visible when employees share messages from brands. Consequently, persuasion knowledge may be activated. It is more likely that people activate their persuasion knowledge when the source of the messages is a commercial brand or is strongly related to this brand.

It is therefore expected that self-created posts of employees on Facebook will lead to more positive affective and behavioral responses among receivers within their network than posts shared from their company or created by their brand. As persuasive knowledge is activated for commercial sources, resulting in more negative affective and behavioral responses. The following hypotheses were formulated:

H3: Employees’ messages are less likely to be perceived as persuasive by its receivers than brand messages do, resulting in more positive affective and behavioral responses. H4: Employees’ self-created brand messages are more likely to be perceived as persuasive

by its receivers than when messages are shared from their company, resulting in more positive affective and behavioral responses towards the brand.

Presence on SNS

Although sources of brand-related messages are expected to have different effects on Facebook, they may be more effective than no presence on this platform at all. Literature suggests that being presence on SNS is effective for brands, while presence on social media creates consumer trust (Bleu, 1964; Luhmann, 1979). Social media interactions lead to higher brand involvement (Social Embassy, 2012). However, so far no studies demonstrated that presence is more effective than no presence at all. Kelleher and Miller (2006) conducted an experiment to test the potential advantages of brand blogs over traditional web sites. Blogs were perceived as more conversational than brands web sites, this correlated with more positive relational outcomes such as trust, satisfaction and commitment. Facebook posts could have this effect as well since they have the same characteristics as blogs (Herring et al. 2004).

Brand presence on SNS enables consumers to interact with a company (McAlexander, Schouten & Koenig, 2002). Consumers who are exposed to brands on SNS tend to be loyal and committed to them, regardless of its form. They are more open to receiving information about these

brands than individuals who are not exposed to them (Muniz & O’ Guinn, 2001). Brands can build some sort of relationship with others on Facebook, which in turn may create brand loyalty and sales (Dholakia & Bagozzi, 2006). Dholakia and Durham (2010) measured to what extent online brand content influences its consumers. Respondents who were exposed to this content, tend to visit the store more, generated more positive word-of-mouth and were more emotionally attached to the brand. However, it is not clear whether this accounts for brand communication on Facebook as well. This study will therefore test if Facebook messages are more effective than no presence at all. As stated in previous hypothesis it is expected that posts have more positive effects when employees create these themselves than when they share posts from brands, than posts created by brands. However, all sources are expected to gain more positive effects than no brand presence. This leads to the last hypothesis:

H5: Brand-related Facebook messages gain more positive affective and behavioral brand reactions when messages are self-created by employees, than shared, than created by a brand. But all sources gain more positive effects than no messages at all.

Method

An experimental study was conducted to test the proposed hypotheses. A between-subjects design was employed, wherein participants were randomly assigned to one out of four conditions (self-created post vs. shared posts by employees vs. brand initiated posts vs. control group), with attitude towards the message, brand attitude, brand image, brand impression, purchase intent and forwarding intention as dependent measures and perceived persuasiveness of the message as mediator.

----insert Table 4 about here---- Pre-test

A pre-test was conducted to make sure that all manipulations were interpreted in the right way, or that the source of the message was noticed by respondents. A total of 20 respondents were approached within the social network of the researcher to participate in this pre-test. The sample was formed by 10 male (50%) and 10 female (50%) respondents. Participants had a mean age of 22 years (M = 22.08, SD = 3.31). The majority of respondents have a Bachelor education or higher (65%). Respondents correctly

identified the manipulated conditions, respondents assigned to the self-created post correctly recognized this (100%). Shared messages (100%) and brand posts (100%) were also correctly identified.

Participants

Through convenience sampling, approximately 220 Facebook friends were approached within the network of the researcher to take part in this experiment. A private Facebook message was sent to ask if participants wanted to take part in this study. A total of 15 respondents were excluded from the data because their response contained missing values. The final sample contained 205 respondents and was formed by 84 male (41%) and 121 female (59%) respondents, with a mean age of 23 years (M = 22.89, SD = 4.03). The majority of these respondents completed a Bachelor education or higher (84.9%). A total of 81.5% of all respondents use Facebook on a daily basis (M = 6.6, SD = 1.1) and 79.5% daily check Facebook’s news feed (M = 6.59, SD = 1.07) More than 70% of respondents can be described as spectator, a person who only reads Facebook messages (71,2%). But 17,6% is active and post messages on Facebook themselves. Furthermore, 25.4% of all respondents did “like” the Red Bull Facebook page and 47.8% of all respondents consumed Red Bull less than once a month.

Stimuli materials

The independent variable used in this study is the source of a message and was manipulated through one Facebook message derived from three different sources. The manipulated conditions, a self-created post, shared post and brand initiated post, displayed a photo of a range of Red Bull mini coopers in a street, ready to drive away. The top of the post showed the message “Just another day at the office, Red Bull gives you wings!” This message was formulated based on results of Study 1, showing that employees mainly post about daily work experiences in an informal style, containing a photo and using an enthusiastic message. These posts differed in their headline, which showed the source of the post.

The self-created post contained the name and photo of a Red Bull employee as source of this message. “Tessa van Dijk” was used as the name for this employee. The post shared by the employee looked similar, but a text was added which indicated that this employee shared a message from Red Bull on Facebook. The headline contained “Tessa van Dijk shared Red Bull’s photo”. Furthermore, the brand initiated post contained the name and photo of the Red Bull brand in the headline; “Red Bull”.

Lastly, a control condition was added which did not show a manipulated message. Each manipulation served as a between-subjects variable. Appendix D shows the Facebook posts that purportedly affect the affective and behavioral reactions towards the brand.

Measures

This study used the affective measures attitude towards the message, brand attitude, brand image, brand image and brand impression since this was in accordance with previous literature on effects within SNS (Feick & Price 1987; Katz & Lazarsfeld 1995; Lee, Rodgers & Kim, 2009). The behavioural measures purchase intent and forwarding intention were taken into account for the same reason, as they are commonly used for studies in similar fields (Chevalier & Mayzlin, 2006; De Bruyn & Lilien, 2008; Chu & Kim, 2011).

Attitude towards the message. The attitude towards the message was measured by asking respondents to what extent they thought the message was (e.g.,) uninteresting/interesting, irrelevant/relevant, ineffective/effective and not credible/credible (Ajzen, 1991). This variable was measured using seven-point scales. Reliability was very high ( = 0.98). The attitude towards the message variable was computed by taking the mean of these seven items (M = 3.60, SD = 2.27).

Brand attitude. Attitude towards the brand was measured by asking respondents’ overall feeling towards the brand on five seven-point scales, unappealing/appealing, bad/good, unpleasant/pleasant, unfavorable/favorable, unlikable/likable (Mitchel & Olsen, 1981). This scale was very reliable ( = 0.92). Therefore the brand attitude variable was constructed (M = 5.03, SD = 1.27).

Brand image. General brand image was measured using six statements on a seven-point Likert scale, ranging from totally disagree to totally agree, e.g., The brand provides good value for money, There is a reason to buy the brand instead of other brands, I have a clear image of Red Bull consumers and Red Bull is different than its competitors proposed by Aaker (1996). The scale was reliable since Cronbach’s Alpha was 0.88. (M = 5.25, SD = 0.89).

Brand impression. Brand impression was measured to assess whether the respondent perceived the company’s identity as the company intended. Respondents were asked on a seven-point scale to what extent they think of the brand as e.g., not energetic/energetic, not sportive/sportive, not

committed/committed and closed/open (Red Bull, 2014). The brand impression variable was computed with these items (M = 5,25, SD = 0.89), Cronbach’s Alpha was 0.84.

Purchase intent. Purchase intention was measured through the following statements Would you try Red Bull if it was on sale?, If Red Bull would be given to me for free I would accept this? and Next time I a buy a drink at a party I will buy Red Bull, using a scale of 0 (would not try) to 7 (would try) (Ajzen, 1991). This scale was reliable; Cronbach’s Alpha was 0.87 (M = 3.72, SD = 1.55).

Forwarding intention. Forwarding intention was measured on a seven-point scale (totally disagree/totally agree) by statements about the probability to like, share or comment on the Facebook (Hoffman & Fodor, 2010; Cvijkj & Michahelles, 2013). The scale was reliable; Cronbach’s Alpha was 0.86 (M = 2.18, SD = 1.67).

Perceived persuasive intent. Respondents’ perceived persuasive intent of the Facebook post was measured by asking them to what extent the post was intended to persuade or influence (Rozendaal, Buijzen & Valkenburg, 2010; Van Reijmersdal, Smit & Neijens, 2010). This was asked on a seven-point scale (Totally agree/totally disagree), items correlated significantly (r = 0.83, p <.01) and formed a reliable scale ( = 0.90) (M = 3.16, SD = 2.11).

Control variables

To be able to control for confounds, several control variables were measured. First, it is likely that respondents already consume Red Bull and therefore have an attitude towards this brand, which might influence results. Therefore we asked respondents how often they usually drink Red Bull on a scale from never (1) to everyday (7) (M = 3.05, SD = 1.68; Mahon, Cowan & McCarthy, 2006). Second, the intensity of Facebook use was measured using the same scale. This variable may influence the way people perceive Facebook content (M = 6.59, SD = 1.07). Third, the occupied role on Facebook (producer, critic, collector, member, spectator or inactive) may influence their forwarding intention while producers are more likely to share content than inactive Facebook users (Forrester, 2010; Verhoeven, 2012).

Lastly, tie strength of sender and receiver may influence persuasive knowledge and was measured in terms of closeness and time spent with the sender (Marsden & Campbell, 1984). Closeness

was measured by asking respondents about their relationship with the sender: Just an acquaintance, not that close but important and close (Perlman & Fehr, 1987). The first item was reversed coded to fit the answer scale and the second item was deleted to increase reliability (= 0.85). Remaining items formed a new variable (R=0.81, p <.01), (M = 3.70, SD = 1,7). Time spent was measured by asking respondents to what extent they agreed with two statements; I spent a lot of time with the sender and I have frequent contact with the sender (M = 2.87, SD = 2.01). Reliability was high ( = 0.85).

----insert Table 5 about here---- Manipulation check

A manipulation check was added to the design of this experiment. One item measured whether respondents in this study recognized the source of the Facebook message. Respondents could rate the message as created by Red Bull, created by Red Bull but shared by a Red Bull employee, self-created by an employee working for Red Bull or as none of these options.

Procedure

Participants were recruited through the social networking site Facebook. A link to the experiment was spread by the researcher through private messages on Facebook and was available during the period from the 15th April until the 25th April 2014. Individuals assessed the experiment by clicking on this link and were randomly assigned to one out of four experimental conditions (Appendix E). The opening page informed respondents that results were anonymous and were used for scientific purposes only. The dependent variables, attitude towards the message, forwarding intention, perceived persuasive intent and a manipulation check were added within the three manipulated conditions. Next, the dependent variables, brand attitude, brand image, brand impression and purchase intent were measured within all conditions. Subsequently, the control variables; Red Bull consumption, Facebook usage, conducted role and social connection with the sender were measured. Thereafter, demographic questions (i.e., gender, age, education) were asked. Finally, respondents were thanked for their participation.

Results

Manipulation check

A manipulation check measured if respondents noticed the source of the Facebook messages within this study. A total of 89,5% of all respondents who saw a message created by an employee also recognized this correctly. For the Red Bull post this was 97,9% and the shared post was correctly identified by 93,9 % of all respondents in this condition. A Chi square test is conducted to test this manipulation check. Results of this test show that the manipulations were significantly recognized by participants, X2 (12) = 548,8, p < .001. Thus, the manipulation of the source within a message worked as intended.

Control variables

Before the hypotheses could be tested, correlation analyses were conducted to see whether the control variables correlated significantly within the studied relationships (Table 3). Closeness and time spent with social connections seemed to have significant effect on all dependent variables as Person correlations were fairly weak to moderate (r = 0.28 to 0.48) for all dependent variables (p < .01). Furthermore, Facebook usage correlated significant with brand attitude (r = 0.17, p < .05) and brand image (0.18, p < .01). The role that respondents adapt on Facebook did not significantly correlate with any of the dependent variables (r = 0.12, p > .05). Red Bull consumption significantly correlated to all dependent variables (r = 0.19-0.57, p < .01) except purchase intent and perceived persuasion. All significantly correlated control variables, with correlations above 0.1, (i.e., closeness, time spent, Facebook usage and Red Bull consumption) were included into the main analysis as covariates.

Testing hypothesis

A PROCESS simple mediation analysis of Hayes (2013) was used to test the first four hypotheses in this study (Appendix F). This model proposes a causal antecedent X variable as influencing an outcome Y through and without a single intervening variable M, along with bootstrap confidence intervals for effect size inference (Preacher & Hayes, 2004). The analysis was conducted separately for each dependent variable in this study. The fifth hypothesis was measured using an UNIANOVA analysis. Social tie and social frequency were control variables within all analyses. Facebook usage was added to

all analyses with brand attitude and brand image as dependent variables. Lastly, Red Bull consumption was controlled for all dependent variables within the analysis except for purchase intent.

Hypothesis 1 and 3

To test the direct and indirect effect of Facebook messages created by the brand or the employee a PROCESS analysis is conducted, with a dichotomous variable for source (employee vs. brand) as an independent variable, perceived persuasive intent as mediator (Preacher & Hayes, 2004) and all the dependent affective and behavioral variables, separately, as outcome variables in this model. The number of bootstrap samples in this analysis was 1000, with a level of confidence interval of 95%. Hypothesis 1 stated that employees’ Facebook posts lead to more positive affective and behavioral responses than brand posts. Results show a significant direct effect of the message source on all outcome variables. Participants who were exposed to a post created by an employee are estimated to score, on average, between 0.83 and 1.26 units higher in their affective and behavioral responses towards the brand than those who see a brand post. Moreover, the direct effect was significantly different from zero while confidence interval was estimated between 0.04 and 1.78 (Preacher & Hayes, 2004). In addition, source of the message has a direct effect on all outcome variables, independent of perceived persuasion. The first hypothesis is therefore fully supported. The control variables had no significant effect on this relationship.

----insert Table 6 about here----

Hypothesis 3 stated that this relationship, between source message and outcome variables, was explained by receiver’s perceived persuasive intent of the message. We therefore look at the indirect effect described in the PROCESS analysis that was conducted. First, the effect of source message to the perceived persuasion is estimated (a). Results show that those assigned to the employee condition are, on average 1.07 lower in their perceived persuasion than those assigned to the brand condition. However, the amount of perceived persuasion does not lead to more positive outcome variables (b), as this path was not significant (p > 0.05). Those who perceive a lower amount of persuasion do not have more positive on outcome variables; no total indirect effect was shown (Table 7). The third hypothesis is therefore rejected. However, few control variables seem to have great influence on the indirect effect.

Red Bull drink was highly significant for brand attitude (p < .001). Social tie was significant for forwarding intention (p < .001) and social frequency did determine a part of the indirect effect on brand impression and message attitude (p < .01).

----insert Table 7 about here---- Hypothesis 2 and 4

The second and fourth hypotheses were again answered through a PROCESS simple mediation analysis, with the created measure for condition (shared vs. self-created) as an independent variable, perceived persuasive intent as mediator and all outcome variables separately added into the design.

Hypothesis 2 predicted that exposure to employees’ self-created messages will lead to more positive outcome variables than exposure to employees’ messages shared from brands. Results show a significant direct effect of the message source on outcome variables. More specifically, participants assigned to the shared condition are estimated to be on average between 0.48 to 1.22 units lower in their affective and behavioral responses towards the brand than those in the self-created condition (Table 8). However, the direct effect was not significant for brand attitude and purchase intent. The null hypothesis is partly accepted with 95% confidence. The control variables had no significant effect on this relationship.

----insert Table 8 about here----

Hypothesis 4 stated that the relation between the source of a message and the outcome variables was mediated by perceived persuasive intent. Results show that an indirect effect is possible for all outcome variables, while all estimates are between 0.01 and 0.28. We therefore look at the first part of the mediation path (a). Results show that those assigned to the shared condition are, on average 0.73 units higher in their perceived persuasion than those assigned to the self-created condition. Furthermore, the second path (b), show significant effects for the attitude towards the message, forwarding intention, brand image and brand impression as effects range from 0.11 to 0.16. Thus, perceived persuasion does lead to more positive reactions on those four outcome variables (Table 9). Hypothesis 4 is marginally accepted. However, some control variables seemed to have influence on the indirect effect. Red Bull

drink was highly significant on all four variables that showed a significant direct effect (p < .001). Social tie was also significant for attitude towards the message and forwarding intention (p < .001).

----insert Table 9 about here----

Hypothesis 5 stated that Facebook self-created messages derived from an employee have more positive affective and behavioral responses than shared messages, brand initiated messages or no brand presence at all. An UNIANOVA analysis was conducted, with message source as the independent factor and brand attitude, brand image, corporate appearance, forwarding intention and purchase intent separately as dependent measures. Pairwise comparisons were conducted within the analysis, to test if the conditions significantly differed from one another.

Source of the message differed significantly (p < .05) on forwarding intention, brand image, purchase intention and brand impression. These variables differ from high to low in the order of self-created post, shared post and brand initiated post (Table 10). All variables showed a significant effect in the order of employee, shared, brand and control condition, from high to lower effectiveness on affective and behavioral responses. Only brand attitude did not show this result since it was rated higher in the control group than in the other conditions. This hypothesis is marginally accepted for all affective and behavioral outcomes, except for brand attitude.

----insert Table 10 about here---- Conclusion

Results of this experiment showed that employees, as a source of brand messages have indeed a more positive effect on affective and behavioral responses to this message and the brand, than posts by a brand have. This finding corresponded with suggestions that employees are more credible than brands within SNS since they are more socially connected to its receivers (Charron et al. 2006).

No mediation effect of perceived persuasive knowledge was found through a PROCESS analysis, as was expected by Friestad and Wright (1994). There was an effect from source message to perceived persuasive intent, but this did not positively affect outcome variables. This may be due to the way people deal with persuasive knowledge. Such knowledge does not always lead to more negative consumer evaluations, positive reactions were also found (Wright, 2002; Wei, Fischer & Main, 2008).

Furthermore, this study showed that self-created messages had more positive results for attitude towards the message, forwarding intention, brand attitude and brand image, than shared messages. The mediation effect of perceived persuasion did occur for these outcome variables. The study showed an order in source effectiveness through an UNIANOVA analysis. All outcome variables scored more positively in order of self-created, shared, brand initiated and no content at all.

General conclusion and discussion

This study is an important step towards a better understanding of employees’ brand communication on Facebook and the effectiveness thereof in terms of affective and behavioral responses. Perceived persuasive knowledge was taken into account as mediator within this relationship (Friestad & Wright, 1994). Two studies were conducted, Study 1 was the first to shed light on the communication content and style features that employees use on Facebook through a content analysis of their Facebook posts. This analysis was based on research of Verhoeven (2012), who studied which brand-related topics employees use on Twitter. The topics he identified were used in this study to measure if similar topics were used on Facebook. Communication style features were studied in the current research as extension of this study, as literature showed that this also influences the effectiveness of communication (Homans, 1958; Peters & Kashima, 2007). Study 1 demonstrates that employees mainly share experiences about their daily work life. They use emotions and photos within their messages and have an informal and personalized communication style.

Based on these findings of Study 1, in Study 2 an experiment was conducted to shed light on the effectiveness of employees’ communication. More specifically, this study was the first to make a distinction between the source of posts and compared employees’ self-created posts to shared posts, to brand posts and to no brand presence at all. Brand reactions were respectively positive to more negative. Study 2 demonstrated that employees’ posts were more effective than brand posts for both affective and behavioral outcomes. Furthermore, employees’ self-created posts ensured a more positive attitude towards the message, forwarding intent, brand attitude and purchase intent than shared brand posts. However, this was mediated by persuasive knowledge.

The overall study contributes to online marketing literature, especially in the poorly understood role of employees as brand ambassadors in SNS. No studies did yet look into employees’ communication on Facebook, in terms of content and communication style. In addition, the effectiveness of employees’ communication within SNS was never studied before, only suggestions about their effectiveness were made (Harris & De Chernatony, 2001; Verhoeven, 2012) or effects in the field of reputation management were studied (Deuze, 2007). This study is the first to compare different sources of messages on Facebook (self-created posts, shared posts, brand posts) and compare these to a condition were brands were not presented on Facebook at all.

Three remarks of these findings have to be made. First, the PROCESS analyses that were conducted to measure direct and indirect effects of source message were analyzed separately for each outcome variable since PROCESS can handle only one dichotomous variable as an independent factor. This may yield a distorted picture and the effects since not all variables were taken into account in these analyses. It is advisable to retake these analyses when such a PROCESS model is available. Second, the experiment was based on people’s interpretation, they were asked to imagine that the manipulated message within the survey appeared in their actual Facebook News Feed. Future research should retake this study as longitudinal experiment using real posts within the network of employees to avoid bias. Lastly, people may already have positive reactions towards the used brand in this study, Red Bull, despite the manipulated condition in which they were randomly selected. Future research should conduct this research with an unfamiliar or new brand to see if the effects are still similar.

Theoretical implications and future research

This study has several implications for future research and theory. First, employees’ natural communication on SNS was studied and showed that personalized and informal styles were mostly used. This is important for future research since communication effectiveness could be tested based on these findings. Although employees were already suggested to be important as marketing tools (Verhoeven, 2012), this study is the first to demonstrate their content and communication style. Moreover, a difference was found in the communication style between employees and brands. Brand communication was less personalized and contained less emotion than employees’ communication.

This study did not look into these differences and used employee’s content and style features for all manipulated items in Study 2 to shed light only the effectiveness of source messages. Future research should look into the differences within communication style features between employees and brands since that could also lead to differences in effectiveness (Kelleher & Miller, 2006).

Second, this research studied a new form of advertising within SNS. The objective of this study was to demonstrate the effectiveness of employees’ communication. This study showed that employees, posting brand-related content, do indeed attract more positive affective and behavioral responses towards this organization while effectiveness of employees on Facebook. This was never demonstrated as important predictor in reactions towards brands before. This implies that future research should pay more attention to employees as brand ambassadors within marketing strategies.

Lastly, the perceived persuasive intent proved to explain the relation between source message and responses towards this message. Theories on persuasion knowledge could therefore play an important role in studying sources of marketing activities within SNS. Future research could focus on how persuasion knowledge of message sources increases over time and how this influences brand reactions. However, not all outcome variables were explained by persuasive knowledge. Consumer evaluations may rather depend on factors such as familiarity to the brand or the attitude towards marketing (Campbell & Keller, 2003). Social tie did seem a significant predictor of the effect and social connections should therefore be integrated in theories on marketing activities on SNS.

Practical implications

Several important practical implications can be derived from the current findings. First, this study showed that it pays off for marketers to use employees as brand ambassadors on SNS. Strong ties are more effective than weak ties or brands on these platforms (Serra & Soto-Sanfiel; Van Noort, et al. 2012). Marketers need to implement employees in their SNS strategies since their communication gets around the activation of persuasive knowledge (Friestad & Wright, 1994). Thus, they must be stimulated to communicate about their brand on SNS. Employers need to involve employees within the brand they work for, employees need to have a positive image of their company and strongly