Facebook Fan Page Interaction

(1)

Facebook Fan Page Interaction

The Key to Satisfied Customers?

(2)

2

Title Facebook Fan Page Interaction: The Key to Satisfied Customers?

Author Robbert G. Warners

Studentnumber s1908049

Organization University of Groningen

Faculty Economics and Business

Study MSc BA Marketing Management and Marketing Research

Year August 2013

First Supervisor Dr. J. van Doorn Second Supervisor Dr. S.F.M. Beckers

Contact

information Schuitendiep 33a

9711RA Groningen The Netherlands (06)45286507 (050) 5014609

(3)

3

Management Summary

This research focuses on the influence of Facebook fan page interactions on overall customers satisfaction. Forward-looking companies are moving away from centrally controlled mass broadcast, towards the development of personal and localized relationships with knowledgeable, demanding customers. This results in a shift from traditional (one-way) forms of marketing to a two-way communication environment where customers can actively interact with a company. Facebook fan pages give companies the opportunity to enhance brand attractiveness, attract consumer attention and establish closer ties with their customers trough such an interactive online environment.

Despite the convergence of trends that strongly support Facebook’s adoption and use by consumers and companies, strong interest among the marketing community at large, and vivid anecdotes of marketing success using Facebook, empirical evidence regarding the effectiveness of Facebook fan pages is lacking (Hennig-Thurau et al. 2010). This research will therefore concentrate specifically on making insightful what the influence of Facebook fan pages interactions are on members’ overall satisfaction with companies that are actively involved in Facebook fan pages.

Empirical data is the backbone of this research. A diverse group of 79 companies from 18 different economic sectors has been examined and both the American Customer Satisfaction Index (ACSI) and Facebook fan pages interaction data of these companies is collected over a period of four years ranging from 2010 until 2013. Using panel data, a fixed effects panel data model is adopted to assess the effects of the metrics of Facebook fan page interactions on the ACSI . Because both the ACSI and Facebook fan page interactions are likely to be endogenous a lagged ACSI variable is included in the model.

The outcomes of the fixed effects panel data model show that an increase in likes on company-generated posts on Facebook fan pages, also increases the likelihood of getting a higher American Customer Satisfaction Index score. In other words, there is a significant positive relationship between the amount of likes on company-generated posts and the level of customer satisfaction.

(4)

4

Preface

This thesis is written to complete my Master Marketing Management and Marketing Research at the University of Groningen. After graduating the Bachelor Communication at the Hanzehogeschool Groningen, I decided to follow my heart and I started with the pre-master Marketing. I am glad I took this opportunity, as Marketing is what I like most in business. With this thesis, the great student time I had comes to an end, and I finally get the chance to practice my knowledge in real life in my new function as a product manager at Philips.

The road to finalizing my thesis has not been without setbacks and difficulties, therefore I would like to use this preface to thank some persons whom have made my graduation possible. Firstly, I would like to thank my supervisor dr. J. van Doorn for her advice and helpful feedback during the thesis writing. I would also like to thank my second supervisor dr. S.F.M. Beckers for his helpful feedback and suggestions on my thesis. In addition I would like to thank Quintly for their provision of the data for this research.

Finally, I would like to thank my parents, sister, girlfriend and friends for their everlasting support during my study period. Their love and encouragement helped me in the process of writing my thesis and to finish my study. I am grateful for that!

Robbert G. Warners

(5)

5

Introduction

(8)

8

1. INTRODUCTION

“90 percent of the companies still wonder what to do with online social media, the other 10 percent are heading to become market leaders..” – Tarmo Groot

1.1 Background

During the last couple of years, online social media has been embraced by consumers at the speed of light. Especially, online social network platforms like Facebook, MySpace and Twitter have seen exponential growth. Online social networks, which are types of online communities, offer users the possibility to interact with friends, family and acquaintances and to build an online network. While online social media used to be the territory of young users who were fast adopters of new technologies, it nowadays covers a much wider demographic domain where almost 75% of the Internet using adults, also use online social media at a regular basis (Stephen and Galak, 2010). It is therefore impossible to think of a world without online social platforms, online search engines and mobile applications where people can interact with each other.

(9)

9 By now it becomes clear that online communication tools can be useful in various organizational domains. However, implementing online communication tools will not automatically result in benefit for the company. Companies more increasingly experience challenges in setting up and maintaining flourishing online communities (Moran and Grossieaux, 2010). Therefore, is it critical to understand how online social media influences customers and to understand how it operates alongside traditional media (Hollenbeck and Kiakati, 2012).

This paper specifically focuses on Facebook as it is the most used online social network site in the world and has recently reached the status of one billion active users as of October 2012 (Facebook.com, 2012). Research of Nielsen (2011) indicated that Internet users are spending more time on Facebook (an immense 8 hours per month on average) than they spend on Google, Yahoo, YouTube, Microsoft, Wikipedia and Amazon combined. To put the success and popularity of Facebook in perspective: “if Facebook were to be a country, it will be the third largest after China and India” (Ang, 2011). Although the number of users and their time spending on Facebook are quite impressive, still the primary source of revenue comes from paid company advertising (Gangadharbatla, 2008). Therefore, Facebook offers several paid advertising possibilities, ranging from rather explicit forms like banner advertisement, to more implicit forms of brand communication and advertisement options, for example the option to create a fan page on Facebook (Lipsman et al., 2012). Especially, these fan pages are increasing in popularity. Recently, Barnes and Andonian (2011) found that 58% of the Fortune 500 companies created Facebook fan pages already, with services companies such as retailers, insurance providers and banks being among the most enthusiastic adopters. Additionally, a February 2011 survey revealed that social media adoption by U.S. small businesses increased from 24% to 31% during 2010, and 87% of these adopters had a fan page on Facebook (Small Business Success Index 2011). One only needs to log on to Facebook to see this popularity of fan pages themselves. Over 140 million brands including Skittles, Pringles, Starbucks, Red Bull and many others have Facebook fan pages, from which Coca Cola and Disney are leading with a massive 40 and 32 million fans respectively (Fan page list, 2012).

(10)

10 Moran and Grossieaux, 2010). In order to subscribe and get full access to the fan page of the firms, Facebook members make use of the “like button”. The goal of operating fan pages is not only to persuade as much fans as possible to click the like button or collecting as much “likes” as possible, but also persuade them to want to own or use targeted brand products and services (Lin and Lu, 2011). It is therefore important for companies, while creating fan pages, to keep in mind that most users tend to go to Facebook to socialize with friends and not to shop and make actual purchases (Harris and Dennis, 2011).

Algesheimer et al., (2010) show that there is an increasing interest in (online) social communities in academic research. Mostly, this academic research focusses on how to turn users into active users, and what factors influences commitment, brand involvement and loyalty. For example, research results indicate that customers that participate in an (online) social community in general have high levels of engagement with the products and brands of the firms (McAlexander et al., 2002), are motivated to help other customers (Bagozzi and Dholakia 2006), are very loyal, and actively recruit others to the community (Algesheimer et al., 2005). Other research of Nielsen and Facebook, also found an increase in ad recall, awareness, and purchase intentions when a users’ news feed indicate that friends have become fans of a particular brands’ profile page (Neff, 2010). Besides promoting the companies and the company products, fan pages provide customers an online platform through which inquiries and comments can be posted (Lin and Lu, 2011), in the same way as customers create content in their own social network. This two-sided way of communication is important for building and maintaining strong brands and plays an significant role in fostering brand relationships as fans/members are strongly committed to the brand and can become brand advocates (Matzler et al., 2011; Algesheimer et al., 2005; Bagozzi & Dholakia, 2006).

1.2 Goal and academic relevance

(11)

11 Their findings support the idea that Facebook fan pages are useful for deepening the relationship with customers. However, it is still not insightful what is happening inside the “black box” of fan pages (Jahn and Kunz, 2012). Less attention has been dedicated to the effects that the consumers participation level in online brand communities (fan pages) has on consumer behavior in terms of satisfaction (e.g., Ansari et al., 2011; de Valck et al., 2009). Compared to brand communities and online communities, fan pages on Facebook have their own unique characteristics and selected audience (as you cannot like fan pages on Facebook when you not have a Facebook account). As a consequence, the research on brand communities and online communities is most likely not entirely transferable. This research will therefore concentrate specifically on making insightful what the influence of Facebook fan pages interactions are on members’ overall satisfaction with companies that are actively involved in Facebook fan pages. Hence, the research question of this paper is:

“What is the effect of interactivity on a company’s Facebook fan page on the company’s overall customer satisfaction?”

To give a valid answer to the main question, the following sub-questions are added to this research:

 What is the effect of Facebook Fan Page growth rate on the level of customer satisfaction?  What is the effect of the amount of company-generated posts on the level of customer

satisfaction?

 What is the effect of the number of likes on company-generated posts on the level of customer satisfaction?

 What is the effect of the number of user comments on company-generated posts on the level of customer satisfaction?

1.3 Structural organization

(12)

(13)

13

Theoretical

Framework

(14)

14

2. THEORETICAL FRAMEWORK

The second part of the thesis consists of a theoretical framework. At first relevant literature about online social communities and brand communities will be described. Next, brand pages in online social networks are discussed. Brand pages differ from other brand communities because they comprise an aggregation of members and are exclusively accessible via online social networks (Kim et al., 2008). Facebook fan pages can be seen as a type of brand pages with their own unique characteristics. Therefore, the possibilities that companies have to interact with consumers through Facebook fan pages are discussed. In the second part of the framework the constructs satisfaction, loyalty and de American Customer Satisfaction Index are described. The last part consists of an explanation of the conceptual model and explanation of the independent variables.

2.1 Online Communities

(15)

15 connections with others but also make it possible for users to join groups, communities and fan pages that are in line with the users’ interest and give companies a platform to reach millions of potential fans (Ellison et al., 2007, Chi 2011).

Balasubramanian and Mahajan (2001) define online communities based on five basic characteristics. First, online communities are constituted of an aggregation of people (Armstrong and Hagel 1997; Preece 2001;Ren et al. 2007;Kim et al. 2008; Madupu and Cooley 2010). Second, community members expect to receive utilities or benefits from their participation in online communities (Dholakia et al., 2004, Ko, Cho and Roberts 2005; Nambisan and Baron 2009). Third, members mostly interact with each other via computer-mediated mechanisms and do not have physical contact (Szmigin and Reppel, 2004, Ren et al. 2007; Kim et al., 2008; Madupu and Cooley 2010). This means online communities are not constrained by space or time (Muniz and O’Guinn, 2001, Szmigin and Reppel, 2004, Kim et al., 2008). Interaction is one of the most important aspects of online communities (Keller, 2009; Algesheimer et al., 2010). There are four major interaction structures in online communities according to Valck et al., (2007), Those are: member-to-member interaction, organizer-to-member interaction, organizer-to-community interaction and the community site itself, which is the basic requirement for interaction to take place. However, not every member interacts with every other member in an online community (Balasubramanian and Mahajan, 2001). Fourth, community members are engaged in a mutual social-exchange process of production and consumption. While each of the community members’ is engaged in some level of consumption, not all of them are necessarily engaged in production. As a result a distinction is made between different groups of users based on their level of active participation (Preece 2001; Ren, Kraut and Kiesler 2007; Wiertz and de Ruiter 2007). Fifth, the social interaction between members revolves around a widespread focus of interest or goal, which can be shared objectives, shared identity or a shared interest (Hollenbeck and Zinkhan, 2006, Reysen et al., 2010). This shared focus of interest results in commitment among community members makes online communities unique (Williams and Cothrell 2000; Ren et al. 2007; Kim et al. 2008). All five online community characteristics that are mentioned above are applicable on brand pages that are exclusively accessible through online social networks as they can be seen as communities of shared interest in a particular brand. In the following paragraph brand communities will be explained in more depth.

2.2 Brand communities

(16)

(17)

17 characterized by relationships between a member and four aspects of the community: the product, the brand, the company, and other members (Muniz & O’Guinn, 2001, McAlexander et al. 2002). In general, customers with higher brand connection and/or commitment may be more likely to participate in brand communities to learn about other brand users, to enhance their knowledge about a brand and its uses, and to spread one’s own knowledge and experiences among others (Schau, Muniz, and Arnould 2009). When researching brand communities it is important to keep in mind that one can distinguish between brand communities that are corporately sponsored and those that are independent (Lee et al., 2011). Especially among brand pages in online social networks many examples exist of independent communities. Lee et al., (2011) indicate these independent brand pages are perceived more credible by consumers as companies have no influence on the content. Nevertheless, corporately sponsored brand pages cannot be completely controlled by companies as members are free to express themselves among those pages (Ren et al., 2007). For the scope of this research only brand pages that are corporately sponsored will be taken into account.

2.3 Brand pages

(18)

18 show a lot of resemblance common online communities (Kim et al., 2008). However there are also some differences compared to regular (brand) communities. First of all brand pages are only accessible for members of the online social network (Kim et al., 2008). Besides, joining the brand page is relatively easy because the only requirement is that one joins the online social network. Gangadharbatla (2008) found out joining brand pages is not among the major reasons for becoming a member of an online social network in the first place. People have different motivations like communicating with acquaintances or pleasing their needs for belonging and information. People that join brand pages often do so in order to receive discounts and show brand support to their friends (Harris and Dennis, 2011). Especially the brand support is distinctly different from ordinary brand communities. Rather than solely relying on marketing communication, companies should therefore try to participate more in their customers’ social media activities, in order to understand the impact of these on their brand image and also to facilitate interaction with potential customers (Heinonen, 2011). Compared to the success of the 30-second radio and or television spot, brand pages are starting to compete with traditional advertising methods by converting the main action of social media, sharing, into something potentially even better for branding than television ads: electronic Word-of-Mouth (Hof, 2011). Hof (2011) therefore states that sharing is the most valuable form of marketing. Nonetheless it is hard to build and even more difficult to control. Due to its importance to online brand pages, the concept of (electronic) Word-of-Mouth will be described in more depth in the following paragraph.

2.3.1 Word-of-Mouth (WOM) and Electronic Word-of-Mouth (eWOM)

(19)

(20)

20 outcomes (Dellarocas and Narayan 2006). The following sections therefore give a review of the literature about satisfaction and the related constructs.

2.4 Satisfaction

(21)

21 itself (such as corporate brand name or in this respect a Facebook brand page) may influence the affect evoked by the evaluative process inherent in satisfaction formation. One could say that customer satisfaction with a buying experience might be associated with happiness, pleasure or misery (Oliver, 1993). So when consumers buildup experiences with a certain company, this results in a reaction that predisposes an association of their consumption experiences with particular emotions such as happiness or affinity (Da Silva and Syed Alwi, 2008). It therefore seems that outcome satisfaction construct is the more suitable customer satisfaction approach for the study. According to Anderson and Srinivasan (2003) there are two issues concerning satisfaction measures. They are likely to be positively biased (Peterson & Wilson, 1992) and establishing the relationship between satisfaction and repurchase behavior has been hard to pin down for many firms (Mittal & Kamakura, 2001).

(22)

22 a consumer expresses his or her satisfaction with the performance of the product or service also directly through loyalty (Newman and Werbel, 1973). Helgesen (2006) argues for example that the main consequence of satisfaction is consumer loyalty. Additionally, they argue that at first the relationship has to pass a certain threshold before influencing loyalty at all. This influencing has to be positive and finally has a diminishing effect on increasing loyalty (Helgesen, 2006). Jones and Sasser (1995) found out that the strength of the relationship between satisfaction and loyalty depends upon the competitive structure of the industry. In a more recent paper, Oliver (1999) found that a satisfying experience leads to brand loyalty, but true loyalty can only be achieved when other factors such as an embedded social network are present. In this light other researchers (e.g. Jang et al., 2008) noted that brand loyalty can be increased specifically by online communities. Therefore, the following section will be dedicated to the construct loyalty.

2.5 Loyalty

(23)

23 literature can be classified into two other separate streams according to Delgado-Ballester and Munuera-Alemán (1999). This classification is based on psychological – and sociological orientation. The psychological orientation stream primarily concerns the cognitive processes supporting the development of brand attitude strength (Delgado-Ballester and Munuera-Alemán, 1999). The sociological orientation stream is directed towards the hedonic motives of brand loyalty and might be seen from a relational point of view where brand loyalty is examined, considering relational characterizing variables like commitment and trust. (Delgado-Ballester and Munuera-Alemán, 1999).

(24)

24 loyalty are explained based on relevant literature the American Customer Satisfaction Index that will be used in explaining the effect of Facebook fan pages will be described.

2.6 The American Customer Satisfaction Index (ACSI)

A regularly used method to measure customer satisfaction is the American Customer Satisfaction Index (ACSI) which is developed at the University of Michigan’s Ross School of Business. The ACSI is an independent national benchmark of customer satisfaction with the quality of products and services available to household consumers in the United States (Theacsi.org, 2013). It measures overall customer satisfaction that is uniform and comparable. The concept behind the market based ACSI cause-and-effect model requires a methodology with two fundamental properties and therefore specifies between the antecedents and consequences of customer satisfaction (Fornell et al. 1996).

In figure 1 the three antecedents of customer satisfaction: customer expectations, perceived quality, and perceived value are described. Customer complaints and customer loyalty are the outcomes of customer satisfaction in the model.

Figure 1 - The cause-and-effect model of customer satisfaction (Fornell et al. 1996)

(25)

25 other hand, the appraised reliability with which the company will be able to deliver this offering (e.g. does the product/service work as it is supposed to?). The antecedent ‘Perceived value’ in the ACSI model is the customer’s subjective evaluation of the quality of the product related to the price of the product. When looking at the outcomes of customer satisfaction ‘Customer complaints’ can be simply defined as the number of complaints a customer has filed with the company within a certain time period. ‘Customer Loyalty’ in the ACSI model, is the likelihood that a customer will repurchase from the company and the extent of a customer’s price tolerance (Fornell et al. 1996). As stated earlier, customer satisfaction is positively related to customer loyalty (retention) and that is leads to less customer complaints (e.g. Yang and Peterson, 2004; Helgesen, 2006; Jang et al., 2008 and Davis-Sramek, 2009). In addition, Fornell et al. (1996) assume that there is a relationship between customer complaints and customer loyalty. The relationship will be positive if the firm is able to successfully handle complaints and turn complaining customers into loyal customers. On the contrary, if a firm handles complaints unsuccessfully it will have a negative effect on customer loyalty.

The next section of the paper consists of the related constructs of Facebook fan pages as these have their own specific characters compared to other brand pages in online social networks.

2.7 The specific case of Facebook fan pages

(26)

26 As the aim of this research is to determine if there is a correlation between Facebook fan page activity and overall customer satisfaction, various Facebook fan pages are followed and different statistics are measured. On Facebook, companies can take several decisions on how to customise their fan pages to enable information sharing, which creates different statistical data and measures. Although several of these measurements can be used for measuring fan page activity (e.g. amount of user posts and response times on user posts), these are not all included in this research. Companies can - for example - allow members to post comments on the brand page’s wall or choose to only admit company-generated content. Measures therefore can differ between companies, which makes it difficult to compare brand fan pages with each other. However, there are also several statistics that are similar between all fan pages, which can make the comparison between the different fan pages possible. These measures are fan page growth, the amount of company-generated posts and the number of likes and comments on these company-generated posts. In the following paragraphs these measurements will be discussed in more depth.

(27)

27 Therefore, this research specifically focuses on the effect of Facebook brand fan page interactions on overall customer satisfaction. The next paragraphs are dedicated to the various ways customers can interact with companies on their Facebook fan pages.

2.7.1 Fan growth

There are some aspects which are unique about Facebook fan pages. First of all, only members of the online social network can access fan pages. Nonetheless, joining a corporate fan page is relatively effortless for Facebook users. Users just have to search for the brand name of the fan page they want to become a member of by using the search function and click one button to verify their choice. The action regarding becoming a fan page member is called liking. By liking a fan page and becoming a fan of that particular page, Facebook users tie their personal identity and online affinity to a brand.

Companies can use Facebook as a platform to inform their fans (e.g. likers) about upcoming events, new product launches, inside stories and general updates (Hof, 2011). An attractive identity on Facebook may lead visitors to ‘like’ the organization's page and join its community (Tuten, 2008). When a Facebook user joins a fan page, this is denoted on his or her user-profile and becomes visible on the news feed for all individuals (Zarella and Zarella, 2011). In other words, if an individual joins or participates on a fan page, all of his or her Facebook-friends are automatically informed about this. The news feed is the first page Facebook members see when they log onto the online social network, and it displays all activities of their friends over the past time period. As the reports of activities in the news feed concerns friends, people are more likely to pay attention to the messages that are shown (Gangadharbatla, 2008). On the other hand, one could imagine that Facebook users therefore also can feel social pressure when liking a page as Facebook combines the power of interpersonal persuasion with the reach of mass media (Fogg, 2008). As a result, when a brand focuses on acquiring and engaging fans, it can benefit from a significant secondary effect-exposure among friends of fans that often surpasses reach among fans (Lipsman et al., 2012). Consumers that see their Facebook friends get involved in a fan page could get curious about the page and after a quick scan become a fan of the fan page as well, increasing the growth rate of that page.

(28)

28 2011; van Doorn et al., 2010). As a result, engaged brand fans tend to visit the store more, generate more positive word-of-mouth, and are more emotionally attached to the brand than non-brand fans (Dholakia and Durham 2010). Besides, customers who take part in brand communities are believed to already have a baseline relationship with the brand, which is further influenced by community participation (Algesheimer et al., 2005). Therefore, the following hypothesis is predicted:

H1 – The Facebook fan page fan growth rate has a positive effect on the level of customer satisfaction.

2.7.2 Company-generated posts

On Facebook, companies have the possibility to actively participate with consumers through their fan page. Smith, (2012) found that companies generally post five types of entries to their fan page: direct marketing of products or services (e.g. launches of new products or announcements of sales); promotion of sponsored events; surveys (e.g. polls concerning future business locations); informational announcements (e.g. openings of new stores); and “fun” postings, usually in the form of questions related to recent or upcoming events. Fans can interact on these brand posts by liking or commenting on them (de Vries et al. 2012). This participation creates a feeling of community and increases loyalty to online communities (Kim et al. 2008). Loyalty is regarded as a fundamental reason for brand community participation, i.e. consumers join brand communities because they like the brand and feel loyal to it (McAlexander et al., 2002). Thus, by engaging in the community, loyalty can be further strengthened. Consumer satisfaction is likewise positively influence by customers’ affective responses such as their enjoyment, excitement and pleasure of using the service (Lynch et al., 2001; Wolfinbarger and Gilly, 2001), and these may be experienced due to higher levels of customer engagement. Besides - through interaction - social relationships are built and a sense of shared values originates among members (Porter and Donthu, 2008). In this way, members’ willingness to participate in brand communities increases (Hsu and Liao, 2007). According to the theory of reasoned action, attitudes are antecedents of behavior, and therefore antecedents of satisfaction as well (Ajzen and Fishbein, 1980). It is therefore expected that the number of company-generated posts leads to more positive attitudes towards a fan page and ultimately an elevated level of satisfaction with the brand. Therefore, the following hypothesis is expected.

(29)

29

2.7.3 Likes and Comments

Consumers engage in a number of behaviors that strengthen their relationship with the brand, which go beyond the traditional customer satisfaction and loyalty measures, such as frequency of visits and purchasing behavior (Gummurus et al., 2012). Brand fan pages in this respect, reflect part of the customers’ online relationship with the brand (McAlexander, Schouten, and Koenig 2002). It broadens the brand–customer relationship (Muñiz and O'Guinn 2001), and provides a source of information and social benefits to the members (Bagozzi and Dholakia 2002; Dholakia, Bagozzi, and Pearo 2004). In order to strengthen this relationship companies create posts on their fan pages to motivate consumers to bidirectional interact. These posts are often commented upon by members and receive “likes” from them. “Likes” is a specialty of Facebook and means that one gives a thumb up for a comment, picture, video, etc posted by the company (Gummerus et al., 2012). By liking or commenting on a brand post, fans get actively engaged with the company and other brand fans and state their opinion publicly. All these interactions are concerned back to the user’s Facebook profile and are displayed to the user’s network of friends in real time via their newsfeeds (Debatin et al., 2009). Liking in particular, gives brand fans the option to indicate they like the company-generated post. Due to its nature and the fact that a ‘dislike’ does not exists on Facebook, liking a post indicates a positive gesture of the fan towards the company-generated post and therefore the company. Liking a brand post is thus in some way similar as positive Word of Mouth communication (de Vries et al. 2012). Previously proven results, as higher brand satisfaction and loyalty, are signs that eWOM has impact on consumer behavior (Dellarocas and Narayan, 2006). Therefore, it is also expected that the number of likes will influence the level of customer satisfaction. Hence, the following hypothesis is expected:

H3 – The number of likes on company-generated posts has a positive effect on the level of customer satisfaction.

(30)

30 satisfaction differs with respect to types of online community behaviors (De Valck et al., 2009). This might also have to do with the fact that to create and maintain relationships between two parties both need to feel that they gain something (Gwinner et al., 1998). From a consumer point of view, engagement behaviors may thus be motivated by satisfying needs and gaining benefits from the behavior itself (Gummerus et al., 2012). One of the engagement behaviors of consumers in Facebook fan pages is the option to comment on company-generated posts. Individuals must first become a “fan” of the company’s page; being a “fan” indicates that the commenter has clicked a button indicating that he or she likes the organization’s products or services, or has a favorable opinion of the sponsoring business (Dekay, 2012). By commenting, fans have the freedom to easily and extensively express their personal feelings with the company. Fans may write comments that are critical of an organization’s products, services, or employees (e.g. Dekay, 2012), but these comments can also be of a more positive nature. The effects from such electronic Word-of-Mouth are proven to have a positive effect to fellow users and consumers when they are positive, but also when the message is less positive (Trusov et al., 2011). Adjei et al., 2010, stated that maintaining a brand community that allows customers to know the firm more intimately through consumer-to-consumer conversations will work in the firm’s favor, even if negative information is shared. Research in the service domain by Verhoef et al. (2002) shows that affective commitment has a positive relation to customer referrals. The same relation may be valid for a consumer and their brand, if the consumer has an affective commitment to the brand it is likely that he or she actively talks about it. Next to that, brand fans who disagree with these negative comments might rebut these by providing positive comments (e.g., Moe and Trusov 2011). People tend to differentiate their opinions and hence post multiple perspectives (e.g., Schlosser 2005). Moreover, the variance in posted comments seems to generate subsequent comments, which is an indication that negative comments are not necessarily bad (Moe and Trusov 2011). So, negative comments might not only lead to more negative comments (conformation), but also to more positive comments (differentiation). As a result, one could argue that the amount of both positive and negative comments has a positive effect on satisfaction. The following hypothesis is therefore anticipated:

(31)

31

2.8 The Conceptual Model

Based on the hypotheses and constructs described in the theoretical framework, a conceptual model can be established (see figure 2). The theory in the previous chapter displays the various levels in which Facebook users can interact with companies trough fan pages. The levels of interaction within a Facebook fan page vary in weight. A ‘like’ can be seen as the lightest level of interaction, as the users effort regarding the action is practically low. Commenting on a company post on a corporate fan page, on the other hand, demand significant time and action from the Facebook user and is therefore weighted the utmost form of fan page interaction.

Figure 2 - Conceptual model

ACSI level of Satisfaction Number of Likes on Company-generated Posts H1 (+) H2 (+) ((+) H3 (+) H4 (+)

Number of User Comments on Company-generated

Posts Number of

Company-generated Posts Facebook Fan Page Growth

(32)

32

Research design

(33)

33

3. RESEARCH DESIGN

This chapter contains the methodology part of this study. The aim of this research is to find out if there is a link between Facebook fan page interaction and overall company satisfaction. At first the sampling technique and data are described. Second, the research method will be explained. The chapter ends with a brief discussion about the indicated measures that will be used.

3.1 Sampling technique and Data

Data is the backbone of this research. A diverse group of companies has been examined and company statistics of both the American Customer Satisfaction Index (ACSI) and Facebook fan pages are collected. The goal of this study is to find a possible explanation of customer satisfaction with a company based on their online presentation on their Facebook fan pages. As described in the literature review, the ACSI is the dependent variable of this study. The independent variables are based on Facebook fan pages interactions and match the specific companies that are monitored by the ACSI as well.

3.1.1 The dependent variable

(34)

34 (Cronbach and Meehl 1955). To the extent that the model predictions are supported, the validity of the ACSI is supported.

From all the companies that are indexed by the ACSI, 79 are selected for this research. The most important criteria for selecting the 79 ACSI companies is based on the fact that their Facebook brand fan pages has to be publicly accessible so that the interactions with fans that take place at these fan pages can be tracked and benchmarked. Next to that the companies are selected based on several criteria (industry, the sort of products and services they provide) to arrange a diverse dataset. Besides, this research focuses only on commercial companies. Governmental organizations are excluded from this study. The selected companies range from airlines, internet service providers, limited service restaurants, specialty retail stores and automobile manufacturers among others. This research however does not include banks and other financial institutions. Due to the current credit crisis leading to trust issues of customers with respect to these companies one might expect problems with biased outcomes. In Appendix 1 a full list of companies is provided. The ACSI index records are collected over a four year time period, ranging from 2010 until 2013. The index records are provided by the ACSI on a yearly basis. Most of the companies followed for this research have a 2013 ACSI index. For some, however, the ACSI index has not yet been publicized as ACSI indexes are updated quarterly.

3.1.2 The independent variables

The independent variables of the model are Facebook Fan Page Growth Rate, Number of Company-generated Posts, Number of Likes on Company-Company-generated Posts, and Number of user Comments on Company-Generated Posts. The fan page data of the 79 companies that are followed in this research is collected by using the Facebook analytics tool of Quintly. The Quintly Facebook analytics tool is a professional social media benchmarking and analytics solution to track and compare social media marketing activities on company brand fan pages on Facebook (quintly.com, 2013). The data of all the independent variables is classified as count data and indicate the amount of interactions on brand fan pages followed over the specific time period selected.

(35)

35 took place from this period are available. Besides, not all Facebook fan pages that exist are monitored by Quintly. Only fan page profiles that are in the Quintly database already have historical data available. This database however is relatively extensive as Quintly is one of the leading social media analytics firms with 25.000 users (quintly.com, 2013).

3.1.3 Data transformation

After the ACSI companies were selected in the Quintly online dashboard, data about the fan page interactions ranging over the years 2010 until 2013 was exported from the dashboard to an Excel file. The 2010 fan page data started as of September that year and ended in July 2013. The data imported was separated by month and had to be combined to a yearly figure. However, after exploring the imported data it became clear that it was impossible to compare the different fan pages to each other as not all pages where followed by Quintly from the same point in time. For example: the fan page of eBay was followed right from the start of Quintly as of September 2010 while sports brand Adidas was only followed since February 2011. Combining the untreated data with absolute numbers would cause incorrect outcomes. To make sure companies could be compared with one and other the available data was added, divided by the amount of months of that specific year the data was available and then aggregated to a yearly number. Facebook fan page interactions of companies that were not yet followed in a specific year have been ruled out from the data for that specific year.

Next to that, a closer look was taken at the data of the American Customer Satisfaction Index. All the companies followed for this research had an index for the years 2010, 2011 and 2012. The year 2013 on the other hand not all companies followed had an index yet. The ACSI for companies in different economic sectors is updated quarterly resulting in missing indexes for companies in economic sectors that are not yet published. All Facebook fan page interactions of companies with missing indexes in 2013 where therefore excluded from this research.

3.2 Research method

(36)

36 analyze a number of important economic questions that cannot be addressed using cross-sectional or time-series data sets (Hsiao, 2003).

Panel data models take into account the cross-sectional and time variability of the information analysed for estimating the parameters. Two models are generally used are the fixed effects panel data model (equation 1) and the random effects panel data model (equation 2):

Fixed effects model: yit = ηi + βxit + υit (1) Random effects model: yit = α + βxit + (ηi + υit) (2)

Where:

i = 1…n, identifies the cross-sections υ = random disturbance .

t = 1…t, identifies the time sections η = cross-sectional heterogeneity

The difference between the two models lies in the relation between the cross-sectional heterogeneity (ηi) and the random disturbance (υit) (Baltagi, 2005). In a fixed effects panel model, heterogeneity is defined fixed or determining, while in a random effects panel heterogeneity is seen as the composition of a ordinary fixed part plus a specific random one for each individual. The decision to use one model or the other must be based on the character of the phenomenon studied. It cannot be based solely on the use of a statistical contrast (Baltagi, 2005). However, the most frequently used statistical contrast in this area is the Hausman Test, and its application is found to be relevant for making the decision when there is no certainty about the relationship between xit and ηi. The null hypothesis of the Hausman Test is the absence of correlation between xit and ηi. Under this assumption, the estimate of random effects (β RE) is consistent, as is the estimate of fixed effects (β FE) (although somewhat less efficient), such that the similarity is significant. There is no difference between the two estimators, and random effects are used (Hsiao, 2003).

(37)

37 American Customer Satisfaction Indexes of the companies followed for this research do not in all cases cover the whole four year time period, an unbalanced dataset is the result.

(38)

38

Analysis &

results

(39)

39

4. RESULTS

This chapter will discuss the results of the analyses. It will start with exploring the data by explaining the descriptive variables and by looking at correlations between the variables. In the next paragraph the results of the fixed effects panel data analysis will be discussed.

4.1 Descriptive statistics

Before analyzing the data in more depth, it is important to get a better understanding about the data. This is done by looking at the descriptive variables that are used in this research. Since this research is based on longitudinal data a closer look is given at the data collected. For this research data of both the American Customer Satisfaction Index (ACSI) and Facebook fan pages is collected over a four year time period ranging from 2010 to 2013. In this specific time frame 79, both domestic and foreign, firms with substantial U.S. market shares are followed. These companies represent 18 different industries (see appendix 1). When the indexes of all companies are pooled over time it becomes clear that the average satisfaction index slowly declines from 2010 onwards (see table 1).

N Mean 2010 41 78.85 2011 75 77.45 2012 76 77.80 2013 30 74.57 Total 222 77.44

Table 1 – American Customer Satisfaction Index

Table 1 also shows the total number of cases used for this research. One might expect this number to be higher since 79 companies are followed in this study. Especially in the years 2010 and 2013 a relatively small amount of cases is shown. This turns out to be the result of combining two separate data sets where the ACSI lacks complete 2013 indexes as these are updated quarterly and historic Facebook fan page data unfortunately goes not for all companies back until 2010.

(40)

40 data used for this research as these do not in all cases cover the researched years. Therefore, the data of the independent variables is calculated to an average number per year over the four year time period. One could argue that Facebook for business purposes has not yet reached its maturity stage as some fan page interactions have steadily increased over time. Table 2 shows the average fan page growth per month for each year. Despite fluctuating growth rates over the years Facebook fan pages still show high popularity amongst Facebook users. This is confirmed by the amount of likes on company-generated posts and comments on company-generated posts which, year after year, show a growing amount of interactions (see table 3 and 4).

N Mean 2010 41 191903.51 2011 75 98324.57 2012 76 198023.73 2013 30 102833.61 Total 222 150347.74

Table 2 – Average fan page growth per month

N Mean 2010 41 17686.12 2011 75 20880.56 2012 76 107446.93 2013 30 178104.32 Total 222 71172.38

Table 3 – Average number of likes on company-generated posts per month

N Mean 2010 41 22883.94 2011 75 26222.74 2012 76 113181.48 2013 30 187889.45 Total 222 77222.63

Table 4 – Average number of comments on company-generated posts per month

(41)

41 and therefore shift from posting an enormous amount of company-generated posts to a smaller but high quality themed content on their fan page (see table 5).

N Mean 2010 41 26.561 2011 75 33.705 2012 76 47.547 2013 30 39.643 Total 222 37.927

Table 5 – Average number of company-generated posts per month

Next to that, a correlation matrix is generated which contains all the variables that are used in this research. When computing a correlations matrix it is interesting to analyse the variables in more depth. Therefore, the mean and standard deviations are calculated for each of the variables (see table 6). Especially the Facebook fan page interactions (e.g. number of fans, number of likes and comments on company-generated posts) show to have quite a large variance in the data causing high standard deviations. The reason for this lies in the fact that the fan pages in this research differ from each other quite a lot with respect to size. This size difference also lies in the basis of the differences between the amount of interactions that take place on the fan pages.

Mean Std. Deviation N

American Customer Satisfaction Index 77.4400 6.3530 222

Average Nr of Fans per Month 150347.7419 318347.1231 222

Average Nr of Company-generated Posts per Month 37.9270 27.8775 222

Average Nr of Comments per Month 77222.6265 149863.0666 222

Average Nr of Likes per Month 71172.3854 142838.2202 222

Lagged American Customer Satisfaction Index 76.94 6.4730 145

Table 6 – Descriptive statistics

(42)

42 comments and the number of likes are highly positively and significantly correlated to each other as well (see table 7).

Correlations ACSI Average # of Fans per Month Average # of Company-generated Posts per Month Average # of Comments per Month Average # of Likes per Month Lagged ACSI ACSI Pearson Correlation 1 Sig. (2-tailed) N 222 Average # of Fans per Month Pearson Correlation .087 1 Sig. (2-tailed) .194 N 222 222 Average # of Company-generated Posts per Month

Pearson Correlation -.108 -.026 1 Sig. (2-tailed) .108 .701 N 222 222 222 Average # of Comments per Month Pearson Correlation -.011 .002 .260** 1 Sig. (2-tailed) .866 .974 .000 N 222 222 222 222 Average # of Likes per Month Pearson Correlation -.015 -.008 .251** .998** 1 Sig. (2-tailed) .829 .902 .000 .000 N 222 222 222 222 222 Lagged ACSI Pearson Correlation .917** .058 -.139 -.070 -.074 1 Sig. (2-tailed) .000 .491 .094 .405 .378 N 145 145 145 145 145 145

**. Correlation is significant at the 0.01 level (2-tailed).

Table 7– Correlation matrix

4.2 Panel data analysis

(43)

43

4.2.1 Fixed effects panel data model

Using the mixed model option in IBM SPSS Statistics 20, the fixed effects panel data model with a lagged dependent variable is generated. Because the dependent variable in this research is based on the American Customers Satisfaction Index, it is likely that there exists some dynamics in the data (Mittal et al., 1999). Therefore, a lagged ACSI variable is created to make it possible to observe changes in the satisfaction scores. The dependent variable in this model is the ACSI. The independent variables are; fan page growth rate, amount of company-generated posts, amount of likes on company-generated posts and finally the amount of comments on company-generated posts. Subsequently, the lagged ACSI is added and a time dummy variable has been included in the model, making it possible to differentiate between the years the data reaches.

As treated in the methodology part of this study the Facebook fan page interaction data is transformed to an average figure because of missing values due to the sampling technique. This data transformation makes it possible to pool the data of all companies together and thus results in better outcomes

Table 8 – Output Lagged Fixed Effects Panel Data Model

The inclusion of the lagged ACSI variable shows to be highly significant on a 99% confidence level (see table 8). Indicating that the lagged ACSI variable appears to be related to the effect of the current ACSI. Since the beta is positive the lagged effect appears to have a significant positive effect on the current customer satisfaction index.

Parameter Estimate Std. Error df t Sig. 95% Confidence Interval

Lower Bound Upper Bound

Intercept 8.47549 2.551907 128.276 3.321 .001*** 3.426206 13.52477 [Year=2011] -1.11994 .658427 76.091 -1.701 .093* -2.431284 .191407 [Year=2012] -1.09652 .569132 62.609 -1.927 .059* -2.233983 .040935 [Year=2013] 0a 0a . . . . . Fans 0.000000643 0.00000067554 103.436 .953 .343 -0.000000695747 0.000001983691 Own Post -.003495 .009895 129.412 -.353 .725 -.023072 .016083 Comments -0.00005567 0.00002662052 93.997 -2.092 .039** -.000109 -0.000002821756 Likes 0.000059002 0.00002774043 96.369 2.127 .036** 0.000003941156 .000114 ACSI-Lagged .907772 .033440 127.102 27.146 .000*** .841600 .973944

(44)

44 The model show significant values for both comments on company-generated posts and likes on company-generated posts (at a 95% confidence level). This outcome indicates, due to the lagged ACSI, that these specific fan page interactions appear to be significantly related to the change in the dependent customer satisfaction scores. When looking to the estimates of the variables one sees a negative coefficient for comments on company-generated posts. This negative coefficient can be interpreted as follows: when the value of the independent variable increases – indicating that there are more comments on a company-generated post – the likelihood of a larger ACSI score decreases. In other words, a larger amount of comments negatively influences the ACSI and consequently customer satisfaction. With the rise of Facebook fan pages it is not uncommon that consumers express their feelings with companies about products purchased or service experienced. These comments can have both a positive or negative character. Hence, the negative comments apparently prevail.

The other variable with a significant influence on customer satisfaction is the amount of likes on company-generated posts. However, this time the corresponding estimated coefficient is positive, so it should be interpreted the other way around. This means that if the value of the independent variable increases – indicating that the amount of likes on company-generated posts increases – the likelihood of getting a higher ACSI score also increases. In other words, a larger amount of likes on company-generated posts appears to have a positive influence on the ACSI and consequently customer satisfaction. It should however be noticed that the estimated coefficient is very small. This is understandable since the variance in the data is rather high. A single like on a company-generated post therefore has only a miniscule influence on the dependent variable.

The other two independent variables, fan growth and amount of company-generated posts appear not to be significant in this model with a lagged dependent variable. Because of the lagged ACSI variable the data of year 2010 is lost for this model. Therefore less cases are used in this model. When one takes a look at the time dummies in this model one can see that the years 2011 and 2012 do significantly differ from base case 2013 at a 90% confidence level.

4.2.2 Goodness of fit

(45)

45 effect as this is not included in the basic model. This basic model is however not further explained as the independent variables showed to have no significant influences on the ACSI. The rule of thumb with respect to the information criteria is that the lower the values, the better the relative quality of the model. One can see that the lagged model predicts better than the basic model.

Basic Model Lagged Model

Information Criteria Value Value

-2 Restricted Log Likelihood 1500.894 744.298

Akaike's Information Criterion (AIC) 1508.89 750.298

Hurvich and Tsai's Criterion (AICC) 1509.09 750.479

Bozdogan's Criterion (CAIC) 1526.36 762.058

Schwarz's Bayesian Criterion (BIC) 1522.36 759.058

Table 9 – Information criteria Next to the information criteria, the R-squared statistic is often used to determine the predicting power of a model. A mixed model procedure in IBM SPSS Statistics 20 does however not produce such a R-square statistic. Definitions for R-square become problematic in models with multiple error terms. As a result generally no R-square measures are produced in most software for these kinds of models. It is possible to compute a R-square value by measuring the squared correlation between observed and predicted values. It is however very well possible that different results are predicted with different definitions of R-square as computing the squared correlation between observed and predicted values may produce a different value than computing one minus the ratio of the sum of squared residuals to the sum of squared deviations of the dependent variable from its mean (IBM.com, 2013).

4.3 Answering the hypotheses

Based on the performance of the panel data model that is explained in the previous paragraph the hypotheses are either accepted or rejected. The hypotheses will be subjected systematically based on the outcomes of the model.

The first hypothesis H1, which states that the Facebook fan page fan growth rate has a positive

effect on the level of customer satisfaction, has to be rejected. There is no evidence that customer satisfaction (e.g. ACSI) is influenced by a higher amount of fans.

The second hypothesis H2, stating that the amount of company-generated posts on the Facebook

(46)

46 Hypothesis three H3, which assumes that the number of likes on company-generated posts has a

positive effect on the level of customer satisfaction, can be accepted. The fixed effects panel data model shows that likes on company-generated posts has a significant p-value and a positive coefficient. This means that the number of likes on company-generated posts can indeed be associated with a positive change in ACSI scores, where higher number of likes increase the likelihood of changes towards higher levels of customer satisfaction. The estimated positive coefficient is however relatively small due to the large variance in the data indicating that the effect of a single like on a positive change in the customers satisfaction score is rather small.

The fourth and last hypothesis (H4) presumes that the number of user comments on

company-generated posts has a positive effect on the level of customer satisfaction. Although the observed significance level of the number of comments on company-generated posts has a significant p-value (95% confidence level) its estimated coefficient is negative. This indicates that the opposite result is found, where an increase in amount of comments on company-generated posts turns out to decrease the likelihood of getting a higher level of customer satisfaction (i.e. higher ACSI score). From this it needs to be concluded that the hypothesis needs to be rejected.

(47)

47

Conclusions &

recommendations

(48)

48

5. CONCLUSIONS AND RECOMMENDATIONS

The last chapter of this research contains the overall conclusions and recommendations. The conclusions will be described first. In this part the theoretical literature will be linked to the results of this study. The second part of this chapter will describe the recommendations.

5.1 Conclusion and discussion

The goal of this research is to find a possible explanation of customer satisfaction with a company, based on their online presentation on their Facebook fan pages. It is critical to understand how online social media influences customers and to understand how it operates alongside traditional media (Hollenbeck and Kiakati, 2012). However, empirical evidence regarding the effectiveness of Facebook fan pages is lacking (Hennig-Thurau et al. 2010). Regardless of the significance of users’ active participation in Facebook, little research had focused on this area. Much research in the topic of online social communities focuses only how active participation among community members can be maintained (i.e. how to turn users into active users) (e.g., Koh et al. 2007; Wiertz and De Ruyter 2007). However, less attention has been dedicated to the effects that the consumers participation level in online brand communities (fan pages) has on consumer behavior in terms of satisfaction (e.g., Ansari, Koenigsberg, and Stahl 2011; De Valck, van Bruggen, and Wierenga 2009). Compared to brand communities and online communities, fan pages on Facebook have their own unique characteristics and selected audience (as you cannot like fan pages on Facebook when you do not have a Facebook account). As a consequence, the research on brand communities and online communities is most likely not entirely transferable. Therefore, this research concentrated specifically on making insightful what the influence of Facebook fan pages interactions on customer satisfaction. The following variables are used for this study: The American Customer Satisfaction Index, Facebook fan page growth rate, number of company-generated posts, number of likes on company-generated posts and number of user comments on company-generated posts. The research question of this paper is:

“What is the effect of interactivity on a company’s Facebook fan page on the company’s overall customer satisfaction?”

(49)

49

Hypothesis Result

H1: The Facebook fan page fan growth rate has a positive effect on the level of customer satisfaction. Not supported

H2:

The amount of company-generated posts on the Facebook fan page have a positive effect on the level

of customer satisfaction. Not supported

H3:

The number of likes on company-generated posts has a positive effect on the level of customer satisfaction.

Supported

H4:

The amount of user comments on company-generated posts has a positive effect on the level of

customer satisfaction. Not supported

Table 10 - Overview of the tested hypotheses.

The outcomes of the fixed effects panel data model indicate that hypothesis H1 is not supported.

The outcomes did not indicate a significant influence of the Facebook fan page growth rate on the level of customer satisfaction. A possible explanation might be the fact that most users tend to go to Facebook to socialize with friends (Harris and Dennis, 2011). The goal of operating fan pages is not only to persuade Facebook users to become a fan or collecting as much “likes” as possible, but also persuade them to want to own or use targeted brand products and services (Lin and Lu, 2011). In fact, the objective of a Facebook fan page is to get fans actively engaged with the company trough online discussions, instead of using the fan page as an extension of the traditional marketing attributes to reach as much (potential) consumers as possible. Gummerus et al., 2012 in this respect point out that fans experience relationship benefits from engaging with fan pages such as entertainment and as a consequence also become more satisfied and loyal.

In addition, hypothesis H2 can also not be supported. The results did not indicate a significant

(50)

50 From the observed significance levels of these first two variables, fan page growth and company-generated posts, it can be concluded that they do not appear to be significantly related to the customer satisfaction (see table 8). In other words, there is no significant influence on change of the ACSI score. This is an interesting finding which supports the idea that true two-way interaction on Facebook fan pages seem to have a greater influence on customer satisfaction than one-way interactions which show similarity with traditional forms of advertising. Companies should therefore need to post content on their fan pages that challenge fans to respond on the post as the more engaged customers are, the higher their satisfaction is expected to be (Brodie et al., 2011).

As can be seen in the table above, hypothesis H3 is supported. This means that the outcomes indeed

showed that an increase in likes on company-generated posts on Facebook fan pages, also increases the likelihood of a positive change of the American Customer Satisfaction Index score. In other words, there is a significant positive relationship between the amount of likes on company-generated posts and the level of customer satisfaction according to ACSI scores. A ‘like’ can only be interpreted as a positive expression from a brand fan. Besides, it shows customer engagement with the company as likes are reactions on company-generated posts. It is therefore coherent that the number of likes on a company-generated post has a positive significant influence on customer satisfaction.

Another interesting finding is related to hypothesis H4 which state that the amount of comments on

(51)

51 customers (de Vries et al., 2012). Since the effect of comments on company-generated posts shows to be negatively related to the ACSI, it could be very well possible that a greater part of the actual comments on company-generated posts is not followed up by the companies. This goes against the nature of online communities, in which interaction between members and the company is a crucial aspect of the success of an online community (Keller 2009; Algesheimer et al. 2009).

5.2 Overall recommendations

(52)

52

Facebook Fan Page Interaction