Digital influencers on Instagram and their impact on
consumers’ brand engagement.
Author: Georgia Ioannidou-Kati
Student number: 11386126
Thesis Supervisor: dr. J.Y. Guyt
Thesis final draft: 23-06-2017
Study field: M.Sc. in Business Administration-Digital Business Track
1
ACKNOWLEDGMENTS
This Master Thesis is the final individual assignment for the master Business Administration- Digital Business at the University of Amsterdam. It investigates the relationships between user-generated content, its perceived credibility and the power of the influencer that is sharing the content towards brand engagement.
I would like to use this opportunity to first thank my Thesis Supervisor Dr. Jonne Guyt, for his continuous support and willingness to help throughout the process. Furthermore, I would like to thank my family, my boyfriend and friends for their unconditional support and understanding throughout this intense year. Finally, I would like to especially thank my father for always being there.
Statement of originality
This document is written by Georgia Ioannidou-Kati who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2
ABSTRACT
The aim of this Thesis is to measure the impact of the user-generated content (UGC) produced by digital influencers and reposted on a cosmetic brand’s account, towards brand engagement. In addition, it is examined how the perceived credibility of the UGC can be considered as the mechanism or process that underlies in the above relationship and how the number of followers of each influencer can influence the strength of the relation. This has been tested with two complementary, quantitative studies. The first study was tested by conducting an analysis with data stemming from four cosmetic brands’ accounts on Instagram. The results indicated that UGC has indeed higher engagement than BGC (brand-generated content) measured by the number of likes. However, the same conclusion cannot be drawn when the number of comments is used as the engagement metric. In addition the number of followers was not proven to have an impact on the relationship between the independent and dependent variable. In the second study, data was gathered through an experiment that aimed at testing the perceived credibility of the UGC on the engagement and its mediation effect on the core relationship. The results stemming from the data analysis indicate that neither the mediation effect is statistically significant nor the UGC produces higher credibility than the BGC. On the contrary, the results proved that if there is a high perceived credibility towards a post it can actually generate higher levels of brand engagement. These results have meaningful managerial implications and can provide useful insights for marketers that are aiming at leveraging digital influencers to promote their products and boost sales.
Keywords: influencer marketing, digital influencers, user-generated content, perceived credibility, Instagram marketing
3 Table of Contents Acknowledgements.………...……..1 Statement of Originality………....…...1 Abstract……….………...…....2 1. Introduction………...……7 1.1 Background….……….……...7
1.2 Scientific and managerial relevance…...………...10
2. Literature Review………...11
2.1 Firm initiated word of mouth-Influencer Marketing...………...…....11
2.2 User-generated content (UGC)………..………...13
2.3 Consumer brand engagement……….……….…...15
2.4 Influential power………..……….….17
2.5 Source perceived credibility....………..………...18
3. Conceptual Framework and Hypotheses……...………20
3.1 Explanation of the variables in the model………..20
3.2 Hypotheses……….22
4. Methodology………..…………...….23
4.1 Study 1- Experiment methodology………...……….23
4.2 Study 2- Instagram methodology………...………24
5. Data………26
5.1 Data Collection……….26
5.1.1 Data Collection: Experiment………..………….…26
5.1.2 Data Collection: Instagram………...………...28
5.2 Instagram data preparation………...30
6. Results………...31
4
6.1.1 Testing part of the framework on Instagram data……….….31
Hypothesis 1….………...31
Hypothesis 2………33
6.1.2 Testing part of the framework on experiment data………....37
Hypothesis 3………....38
Hypothesis 4………39
Hypothesis 5………39
7. Discussion and managerial implications………41
8. Limitations and future research………...45
9. References..……….……..47
10. Appendices...52
10.1 Experiment pre-test results………....…...52
10.2 Experiment………..………….54
10.2.1 Condition 1: User-generated content……….…54
10.2.2 Condition 2: Brand-generated content……….….54
10.2.3 Questionnaire………..………..55
10.3 Correlation analysis-Instagram data……….57
10.4 Descriptive data of experiment sample………....58
5
List of Tables
Table 1: T-test results comparing UGC and BGC on the level of engagement…………..…32
Table 2: Hierarchical Regression model of brand engagement: number of likes………...….35
Table 3: Model Summary with dependent variable the number of comments…….……..…36
Table 4: Basic mediation output in Process………....38
6
List of Figures
Figure 1: Conceptual framework……….…..20
Figure 2: Conceptual framework- study 1……….…31
Figure 3: Conceptual framework-study 2……….….37
Figure 4: Final results study 1 and study 2………....39
Figure 5: Overview of the hypothesis acceptation……….…40
7
1. INTRODUCTION
1.1 Background
During the past decade, social media platforms have transformed marketing practices, offering brands new ways to reach, communicate, interact and engage with customers. This revolution caused by technological innovations, has created new possibilities for firms to expand their customer base and, thus raise their profits (Lamberton & Stephen, 2016). In fact, consumers are no longer passive recipients of marketing campaigns. Instead, they interact with each other by forming online communities, creating user-generated content (UGC) and producing online word of mouth (WOM). In addition, they cooperate with firms in order to influence potential customers’ purchasing behaviors and shift marketing strategies (Fader & Winer, 2012).
The emergence of firm-initiated word of mouth through endorsing consumer-to-consumer communications is an increasing trend on the marketing field, as it is generating brand engagement. This concept can be defined as ‘amplified’ WOM and it differs from ‘organic’ WOM, which occurs without the firm’s contribution (Chae, Stephen, Bart & Yao, 2016). It is also referred to as influencer marketing, viral marketing, and buzz. On one hand, marketers are leveraging social media networks to reach a wide range of potential customers, as well as to raise brand awareness and engagement. By urging users to interact with each other and more precisely to recommend a product or service, practitioners are cultivating an online community and building a mutual relationship based on trust. Consumers on the other hand, are evolving as co-creators of value and undertake an active role in the formulation of marketing campaigns by influencing potential customers to purchase new products (Kozinets, De Valck, Wojnicki, & Wilner, 2010).
8
Brands are gradually initiating to implement this new marketing strategy, especially in the social media platforms, where influence power is more evident and effective. An industry that largely depends on the above phenomenon is the cosmetics business that utilizes influencers on Instagram to drive high levels of engagement. Cosmetic brands rely on images and on the attractive visual tools of Instagram to advertise their products, through beauty endorsers. One relevant example is MAC Cosmetic’s campaign on Instagram with the hashtag #MACArtistChallenge that enabled MAC makeup artists to upload pictures with looks created exclusively with MAC products and to post them using this hashtag. Afterwards, MAC’s Instagram experts reposted them in the brand’s account and thus, they created an artistic album of photographs with striking visuals. Posts included a list of products required to recreate the look in order to inspire and influence the audience to proceed in a purchase.
The majority of players in the cosmetic industry are targeting young women, who use Instagram on a daily basis and are more prone to be influenced by third parties. Hence, these users may purchase products that are endorsed by their favourite Instagram celebrity (Djafarova & Rushworth, 2017). Digital marketers are leveraging this new opportunity by discovering people with persuasive power on social networks, to promote their products and thus raise conversion. Social media influencers are a new form of third party endorsers that can shape audience’s purchase behavior and public opinion, in general (Freberg, Graham, McGaughey, & Freberg, 2011). Marketers are selecting social media ‘leaders’ by judging their relevance with the brand, their reach and their resonance. The above three elements are referred to as ‘the Pillars of Influence’ and each one of them is required to define a social media influencer figure (Solis, 2017). Consequently, an influencer selected by a brand is required to generate content that is brand related and of good quality, to be able to reach a large audience, as well as to maintain its engagement.
9
Influencer marketing is undoubtedly a new trend in the digital marketing, offering firms new possibilities to interact and engage with customers. Therefore, brands take advantage of the user-generated content produced by social media influencers in order to establish credibility in the online environment and impact online or offline sales by strengthening brand engagement levels. This marketing strategy stems its value from three different sources: its wide reach to the audience, its original content-quality and its trust, which is cultivated between users in the online context.
For these reasons, influencer marketing is already being used in the cosmetic business industry and its potential contribution is evident. However, there is still little academic research on this field. Previous studies tried to measure the impact of the user-generated content leveraged by marketers, on financial indicators, such as stock market performance and sales. Up until now, no research has focused on estimating the effect of UGC on non-monetary indicators, such as consumer brand engagement. In addition, scholars have not paid any attention to the brands’ marketing strategy that leverages user-generated content produced by influencers on Instagram. Brands are constantly using influencers to promote their products or services, with a low cost. According to Bughin (2015), the reach of digital influence is growing and in some product categories: ‘5 percent of the recommenders accounted for 45 percent of the social influence generated’.
The proposed Thesis aims to investigate the impact of UGC generated by Instagram influencers and posted on brands’ account, on the consumers’ engagement and test if the perceived credibility of the post plays an important role in this relationship. In addition, the research will try to discover if the power of each influencer has an impact on the strength of the engagement that each post gains.
10
To fill in the gap from the previous literature the below question was formulated:
What is the impact of the user generated content produced by Instagram influencers and posted on cosmetic brands’ accounts, on the consumers’ engagement and how is it affected by the followers' base and the source's perceived credibility?
1.2 Scientific and managerial relevance
The aforementioned Thesis research question has a high academic interest, as it will shed light on a field that has received limited attention up until now. This field is the firm-initiated word of mouth strategy implemented by social media influencers. Moreover, the research will offer insights to marketing practitioners that want to raise their social media returns without disproportionately increasing their investment. The importance of this topic is highlighted by Bughin, Doogan and Vetvik (2010), which states that the firm-generated WOM has more than double the impact on sales, compared to paid advertising. Seeing that customers’ purchase behavior is being influenced by product endorsements, marketers will have to shift their attention on building sustainable relationships with the social media influencers.
11
2. LITERATURE REVIEW
During the last years, following the popular advent of social media many researches have been conducted concerning the consumers’ relationship with brands. Furthermore, the surge in the usage of social networks has changed the way consumers and brands communicate and has shifted the brands’ marketing strategy. According to Nielsen (2012), 46% of online users depend on social media recommendations when making a purchase. Nowadays, marketers face the growing pressure of having not only to acquire clients in order to raise profits, but also to retain them by strengthening their brand engagement. To capture the value emerging from the engagement efforts it is essential to monitor and influence the customers’ purchase behavior and their attitudes. Achieving this, could be more effective by prompting consumers to encourage other people to buy from the firm or to promote the brand online and, hence to raise the engagement levels (Kumar, 2015). The current chapter will introduce the previous literature upon the topic of this Thesis by examining past researches.
2.1 Firm initiated word of mouth-Influencer Marketing
Starting from the work of Godes and Mayzlin (2009), various papers have studied the marketing strategy of firm-initiated word of mouth that aims to drive sales. This firm initiative plans to initiate customer-to-customer communication regarding a product or a service. The above-mentioned study focused on identifying, which types of customers have the most influence upon others and therefore can be used to generate electronic word of mouth. The findings indicated that consumers, who are loyal towards a brand, are not necessarily the most influential people, as their peer network has probably already been informed about the latest news of the brand. Moreover, insights from this research led to
12
accepting the hypothesis that the less loyal customers can be more effective in producing chatter around a brand’s offering.
One aspect of the firm-initiated word of mouth are the seeded marketing campaigns (SMC), which include the firm's initiative to send free products to specific consumers and prompt them to spread their experience as positive word of mouth (Chae et al., 2016). Seeded marketing campaigns can be considered as paid influencer marketing strategy and the reward offered to potential seed consumers ranges from earning free products to earning a sponsorships. Those consumers are called social media influencers and they have the ability to leverage public opinion and affect purchase behavior by sharing online content. Some customers’ embracement or opinions have an excessive power and impact on others’ product adoptions (Van den Bulte & Joshi, 2007). Social media users trust the chatter generated from other consumers especially when they have a significant social network influencing power and a large number of followers. This dependence is evident and can be confirmed through a Nielsen report (2009), which states that 70% of consumers trust the referrals from other anonymous consumers, rather than traditional advertising on TV. Factors that can potentially foster this trust are the influencer’s credibility and the quality and uniqueness of their shared posts.
Another source of generating WOM is the earned influencer marketing, which involves preexisting relationships with social media opinion leaders that are not sponsored by the brand. Influencers generate brand related content on the internet in order to increase their social growth. Although this form of product placement is not being financed by the brand, it can be also driven by marketers. Practitioners can support and encourage those initiatives by establishing hashtag campaigns or contests that stimulate the influencers’ participation. Moreover, earned marketing is not only favourable due to its low cost, but also due to its
13
large impact on the customers’ engagement. According to Lovett and Staelin (2016), earned media plays a primary role in improving the enjoyment effect affiliated with the product or the service that is promoted. Social media users prefer and have more trust on the campaigns that are not sponsored and which are not directly guided by brands.
2.2 User-generated Content (UGC)
Due to the massive usage of social media, consumers are shifting their attitudes from passively searching information to actively communicating their experiences to others on online platforms, such as micro blogging websites and forums. The content that consumers produce online is commonly known as user-generated content and it can be considered as a broader phenomenon of the electronic word of mouth (Tirunillai & Tellis, 2012). UGC’s growth can be easily observed in China where the available online content is mostly user generated, rather than firm-generated (Smith, Fischer & Yongjian, 2012). This content might be produced individually or collaboratively and can be an essential tool for marketers to drive brand engagement, as it is an effective marketing strategy due to its availability, low cost, wide access and reach (Tirunillai & Tellis, 2012). Additionally, marketers are leveraging social networks and are producing content (BGC) in order to promote the brand to consumers. Furthermore, because of the simultaneous engagement of consumers and marketers on social media, consumers’ purchase decisions are often influenced by both UCG and BGC (Goh, Heng & Lin, 2013).
Prior literature has paid attention to three different streams of brand related user-generated content (Smith et al., 2012). The first one studies consumer-generated advertisements and brands, what motivates consumers to produce content, how the process of co-creation can be monitored by marketers and also discusses future implications for practitioners. The second
14
stream investigates to what extent consumers trust user-generated posts, as well as their efforts to engage with them. The third and last approach of brand UGC estimates its effect on financial indicators, such as sales.
Studies focused on the financial impact of UGC upon brands confirmed the hypothesis that UGC can contribute to changes in the firm’s monetary indicators. Tirunillai and Tellis (2012) indicated that negative UGC has a substantial negative impact on firm’s stock performance while this is not the case for positive UGC. Therefore, this is crucial for the marketers, as they should regularly monitor online negative word of mouth and take initiatives and actions in order to deter potential damage to shareholder value. Furthermore, Goh, Heng and Lin (2013) tried to measure the effect of community content produced by consumers (UGC) or by marketing managers (BGC) to purchase expenses. The results of this study indicate that an increase in community engagement raises sales numbers and UGC has an impact on purchase’s volume through informative and persuasive interactions. On the other hand, BGC affects purchase quantity only through persuasive interactions. Ideally, the right marketing strategy combines both UGC and BGC to drive sales.
According to Smith et al. (2012), who investigated the differences of brand related user-generated content across three social media networks (YouTube, Facebook and Twitter), YouTube is more powerful in fostering UGC due to its philosophy towards self-promotion. However, Twitter can be more appropriate for sharing news and cultivating discussions as the brand centrality on UGC is higher, than that of the other social networks. Some interesting findings for practitioners could be that compared to YouTube, Facebook and Twitter are more suitable to marketers for communicating with consumers and for raising the brand awareness.
15 2.3 Consumer brand engagement
Modern technology plays an essential role in maintaining loyal relationships among firms and customers; especially from the perspective of value co-creation. The interactive environment of social media favours the establishment of co-production and cooperation towards boosting brand loyalty and engagement ( Apenes Solem, 2016). Prior literature has focused on how companies can foster consumer engagement with the brand and which marketing strategies are the most effective towards this direction. The term consumer engagement in general can be defined as the direct interaction of the consumer with the brand and it can be built through a collection of positive experiences (Calder, Malthouse & Schaedel, 2009).
This Thesis examines the customer brand engagement as the interaction of customers and brands in the online environment that can be cultivated through the display of the user-generated content on brands’ social media accounts. Consumer brand engagement according to Hollebeek, Glynn and Brodie (2014), is a multidimensional construct that consists of a cognitive, an emotional and a behavioral nature. The first dimension consists of the thoughts each consumer has related to the brand. On the other hand, the emotional part of the engagement is the level of positive affection between the brand and the consumer. Finally, the behavioral dimension consists of the activities, the effort, as well as the time each consumer spends during the interaction with the brand. The latter two aspects of the engagement constitute the major determinants of the self-brand connection.
Another important aspect of the online consumer engagement is that it can be either passive, when the content produced by marketers is not being utilized by users, or active, when consumers are participating in the creation of brand-related content (Schamari & Schaefers, 2015). In the first occasion, users are referred to as followers, as they might follow
16
a brand on social media, but they do not interact with it through actions, such as sending messages or posting comments. On the other hand, the second category consists of people, who do co-create value and they are referred to as advocates, since their aim is to promote the brand across the social networks and increase brand loyalty (Barger & Labrecque, 2013).
Customers can contribute to the brand and raise its profits directly by buying products, as well as indirectly by influencing potential customers to engage with the firm (Kumar & Reinartz, 2016). The latter can only be achieved, if the customer is connected and familiarized with the brand. However, it is also essential that the customer is eager to promote the brand to others. If people have a sense of loyalty for a particular brand, they become keen on generating content related to a specific product or service of the brand or even the brand itself. In return, social media users will trust the endorsements stemming from another user, especially if the latter is showcasing a sense of credibility and is providing content of high quality. Consequently, the above activities are creating a vicious cycle that drives engagement levels and can augment the company’s profits.
In the academic literature, online engagement is related to specific metrics that are used to evaluate its influence on customers’ social media behaviour. Online engagement according to Peters, Chen, Kaplan, Ognibeni, and Pauwels (2013), differs from offline as it is more difficult to be sustained in the long run and it requires consistent monitoring. Engagement as a Key Performance Indicator (KPI) in the online environment can be measured more effectively by calculating online activity, such as likes, comments or shares. More specifically, comments and shares provide more value in fostering customer engagement, as
they can be more influential to potential customers. Additionally, online engagement can be more impactful compared to its offline counterpart, as highly engaged followers can play a
17 2.4 Influential power
Influencer marketing has innovated and empowered the marketing area, leading to new opportunities for marketing practitioners to reach a wider range of potential customers by leveraging the network effects and spreading the advertising message more effectively. The above marketing strategy leads to an increase in sales and to a reduction of the advertising costs (Li, Lai & Chen, 2011). The main issue that experts have to cope with is how to identify the most effective influencers that can significantly drive the purchase decisions of potential or existing clients. According to the ‘3 Pillars of Influence’ mentioned in the introduction, an appropriate influencer has to, not only relate to the brand and have an adequate audience, but also mainly to maintain a large percentage of the audience engaged. For those reasons, although an influencer might have millions of followers, if he or she is not interacting with the audience then the influencer will not be as effective and valuable to the brand as a micro influencer with a highly engaged audience.
Towards this direction, Li et al. (2011) conducted a research in order to discover which, factors affect the influential power of digital ‘leaders’ in the blogosphere and how these factors can be ranked according to their marketing effectiveness. For this reason, a Marketing Influence Value (MIV) Model was developed and the following three categories of factors were identified: the network-based, the content-based and the activeness-based. The first category involves the social-network related information and it can be divided into two parts, the ‘social connection’ and the ‘social interaction’. High ‘social interaction’ is evident when influencers with a high amount of comments and citations are creating a ‘buzz’ around their name and extending their power. The second category, namely content-based factors contains all the information related to the post: the length, the average living time and the level of subjectivity. Finally, the third category embodies all the related activities that generate and
18
maintain social media conversations. For example, when bloggers tend to reply frequently to their comments in order to raise digital reactions around them and thus gather more attention. The results of this study are providing significant information for enterprises that want to reduce their marketing resources and also to become more effective in targeting potential customers. Consequently, the most effective influencers should acquire simultaneously the following advantages: the support of society by having high social connection and interaction, the engagement from their followers, while also be able to produce high activity to sustain the ‘buzz’ around their names. In this paper, influential power will be measured by the amount of Instagram followers, as it is the most objective metric and according to Jin and Phua (2014) higher followers’ base means higher social influence.
2.5 Source perceived credibility
Influencer marketing can be an effective tool for marketers to foster interest around a product or service and to increase sales, compared to the traditional paid advertising. As mentioned earlier, people trust the user-generated content in social media platforms and consider it as useful information. A major mechanism that contributes to the process of generating trust between two different parties is credibility. Source credibility can be defined as “the perceived ability and motivation of the message source to produce accurate and truthful information” (Li & Zhan, 2011, p. 240). Existing literature related to this topic demonstrates that consumers are more prone to be influenced by an advertising message and adjust their purchase behaviour, when the source per se is considered as credible (Hautz, Füller, Hutter, & Thürridl, 2014). Two dimensions are the main constructs of source credibility and are referred to as: expertise and trustworthiness, according to Hautz et al. (2014). Expertise is considered as the prior knowledge related to the product/service acquired
19
by the user that disseminates the content. On the other hand, trustworthiness is the level of inspired confidence that the receiver of the message senses from the communicator in the process of providing objective information.
Source credibility determines the extent to which the consumer believes in the honesty and fairness of the provided information. Furthermore, when the UGC is considered as credible, people see it as unbiased and objective. This leads to consumers believing that there are no commercial interests and that the digital influencer posting this relevant content is promoting the product/service due to its high quality and without earning a direct profit (Owusu, Mutshinda, Antai, Dadzie & Winston, 2016).
20
3. CONCEPTUAL FRAMEWORK AND HYPOTHESES
The research question is established as follows and the below conceptual framework was developed to visualize the relationship between the independent and dependent variable and to present the mediation and moderation effects upon this.
R.Q: What is the impact of the user generated content produced by Instagram influencers and posted on cosmetic brands’ accounts, on the consumers’ engagement and how is it affected by the followers' base and the source's perceived credibility?
Figure 1: Conceptual Framework
3.1. Explanation of the variables in the model
The brands’ posts on Instagram will be the independent variable as it is considered the input in the examined relationship and has a binary value. It can consist of either a user-generated content (UGC) that is being produced by social media influencers and is being reposted on brands’ accounts or it can include photos or videos posted by social media marketers (BGC) on brands’ accounts. The consumers’ brand engagement is the dependent
21
variable as it is the output or the outcome stemming from the above relationship and it can be measured by the number of likes and comments on each Instagram post, as well as from the results of the experiment, mentioned later on. Perceived credibility plays the role of the mechanism in this model, since it is the factor that can contribute to the direct relationship and can be characterized as the mechanism. Furthermore, the influencer’s power can play the role of the moderator, as it can affect the strength or the relationship between the independent and dependent variable. The aim is to measure whether high influential power affects positively the engagement of the repost or if it does not add any further value.
In addition, there are two control variables that have an impact on brand engagement, and hence they should be held constant throughout the data analysis. According to Bono and McNamara (2011), the selection of the appropriate control variables renders the conclusions more valid and allows some generalizations. Control variables in a study should meet three core criteria in order to be considered as effective. First, there should be evidence that there is a relationship between the covariate, which is the control variable, and the dependent variable. In addition, there should also be some expectation that the control element is correlated to the independent variable. Finally, the control variable should not have a mediation or moderation effect or even be one of the hypothesized variables. Considering the above theory the following control variables were developed to investigate the validity of the results more effectively.
An essential factor that can influence the engagement level on Instagram is the perceived post quality. According to Figueiredo, Pinto, Belém, Almeida, GonçAlves, Fernandes and Moura (2013), a post has content related, as well as textual related features. The first category is more relevant to this research, as Instagram is a platform that depends mostly on images
22
and videos and not on textual features. In this Thesis, the perceived post quality will be evaluated by measuring the perceived creativity of the photo and the photo’s resolution.
Another metric that should be considered for the purpose of answering this research question is the familiarity with the brand. People, who already know the brand or have developed to some extent a relationship with it, may be more prone to get engaged afterwards. Thus, during the data analysis this variable should be controlled.
3.2 Hypotheses
The following hypotheses were formed according to the existing literature and they will be tested by the two research methods that will be analysed in the following section:
H1: The user-generated content posted on brands’ accounts, has a greater effect on the
consumers’ brand engagement than the brand-generated content.
H2: The power of the influencer moderates the relationship between user-generated content
and consumers’ brand engagement.
H3: The user-generated content posted on brands’ accounts, has a positive effect on the
post’s perceived credibility.
H4: The posts’ perceived credibility of the user-generated content has a positive effect on the
consumers’ brand engagement.
H5: The effect of the user-generated content on the brand engagement is completely or partly mediated by perceived credibility.
23
4. METHODOLOGY
For the purpose of this Thesis, a deductive quantitative research is conducted and the required data are gathered through two distinct primary research methods: an online experiment and a social media research. The aim of using two different research methods is to investigate the relationships between the variables through two different points of view. However, the two methods are complementary to each other, and each one of them is filling in the gaps-limitations of the other. The experiment is used to identify the effect of the mechanism, which is a more abstract concept (perceived credibility) and can be measured effectively by collecting primary data. On the other hand, social media research constitutes the data field part of the research and uses real time data, offering more valid results with fewer limitations.
4.1 Study 1-Experiment methodology
As mentioned previously, the first stage in the research is a web-based experiment that aims to test the hypothesis of the impact that the mechanism has on the relationship between the independent and dependent variable. Specifically, this research method is aiming to measure the perceived credibility of posts and its influence on the engagement of Instagram users. The online experiment method was chosen as an appropriate research method for this specific Master’s Thesis, as it provides numerous benefits to the researcher such as low cost, external validity, wider reach of participants and convenience through a high degree of automation (Reips, 2002). Prior to the actual experiment, a pre-test was conducted to evaluate the usefulness of each question in the experiment and test whether the participants are understanding and are answering the questions in a logical way. According to Reips (2002), pretesting is an effective tool for researchers, as it provides a feedback prior to the final experiment, preventing potential misunderstandings.
24
Following the analysis of the pre-test results (Appendix 1) the actual experiment was developed. In this experiment, two different types of posts were presented; each one of them was tested according to the degree of its perceived credibility. For the purpose of this experiment, a randomization scheme was chosen with a probability of 50% for each condition. Each participant was assigned to one condition, thus he or she was accessing the questions’ part referring to one out of two possible types of posts. The aim of the randomization feature is to minimize bias and create homogeneous treatment groups (‘Experimental Design’, 2017). Next, the study is investigating the impact that credibility has on the interaction of the user with the brand in the online and in the offline context. One of the posts included user-generated content that was posted on the brand’s account, while the other included brand-generated content, posted on the same brand’s account. Finally, contestants in the experiment would rate several perceptions that they had, when seeing the post, such as the level of credibility that they sensed from each post and their level of brand engagement. The results of the experiment led to useful conclusions that were used to test the mediation effect presented in the conceptual framework.
4.2 Study 2-Instagram methodology
The second research method was suitable for shedding light on the aforementioned research question as it examined real time data from the firm’s Instagram account. In June 2016 Instagram released its API update, making it hard for third parties to acquire data through web scraping. With this initiative, Instagram was aiming to raise revenues by leveraging its own data, and increasing traffic on the application (‘Instagram Platform Update’, 2017). Therefore, the data for the specific research was obtained manually and not through web scraping techniques which are utilized through programming.
25
The required amount of posts was extracted from the brand’s accounts and the content of each post was evaluated with regards to, whether it was user-generated or not. This evaluation was completed by considering whether the Instagram posts were being reposted on the brand’s account from another user’s account. In addition, as mentioned previously in the conceptual framework section, the number of followers that each influencer possesses is considered as the moderation effect in the relationship between the independent and dependent variable. Brands are using different kinds of influencers, according to their reach on Instagram to stimulate the consumers’ engagement. By analysing this dataset, the main aim was to identify the impact of the user-generated content on consumers’ brand engagement. This was measured by the number of likes, and comments each post has attracted. In addition, the data analysis led to the evaluation of whether the influential power of the digital influencer has an impact on the strength of the relationship between the independent and dependent variable.
The brands that were monitored are active on Instagram and are using the repost technique by sharing user-generated content related to their products. The required sample consisted of four popular cosmetic brand accounts, with two of them following a mixed strategy by sharing UGC and BGC simultaneously and the other following a UGC and a BGC strategy, respectively. The aim of testing brands with a different strategy was that the results emerging from the data analysis would be representative and useful conclusions would be drawn. Afterwards, a quantitative analysis was conducted to identify the relationships between the variables used in the conceptual framework and accept or reject the above hypotheses. Finally, conclusions and some limitations were drawn according to the results emerging from the analysis.
26 5. DATA
5.1 Data Collection
The sections below detail the procedures followed in the experimental design of the study and the procedure followed in the empirical analysis using real Instagram data.
5.1.1 Data Collection: Experiment
The experiment is designed to investigate the hypothesis concerning the perceived credibility of the user-generated content, and whether it affects brand engagement. For this purpose, a questionnaire was constructed with two alternative conditions. Each respondent was randomly assigned to one of the two experiment conditions. The experiment was conducted during the period ranging from the 10th to 18th of May.
The first condition contained a photo of a fictitious digital influencer that was reposted on the brand’s account, while the second one displayed the exact same photo, which however was a brand-generated content and it did not contain a specific reference about the displayed person. The post was created exclusively for the experiment and was hypothetically posted by the brand ‘MAC Cosmetics’. This firm was selected as it is the most followed cosmetic brand account on Instagram. Hence, most of the participants could be familiar with it and the results would be more representative. The perceived credibility and the brand engagement were measured by a set of questions, and each set was presented in a separate page with the same post remaining at the top of each page.
The questionnaire was created using the Qualtrics.com survey software. The construction of the questionnaire was based on the existing literature and the majority of the questions were selected from previous researches. The first two questions aimed at investigating the familiarity of the respondents with the ‘MAC Cosmetics’ brand and their Instagram usage
27
behaviour. The goal of those two introductory questions was to monitor if the prior knowledge, concerning Instagram and the displayed cosmetic brand, may influence the results.
Furthermore, the questionnaire aimed at investigating the level of engagement each type of posts generates. The brand engagement in this specific questionnaire consists of three distinct dimensions. The first one is related to the online context and measures the online behavior of the participants by testing how probable it is that they would press the ‘like’ button or post a comment, following a seven-point Likert scale. The second dimension is referred to as affective brand engagement and according to Hollebeek et al. (2014) it can be measured by specific questions, such as ‘I feel positive when I use the brand’. Additionally, the third dimension regarding the activation aspect of the engagement is also established by Hollebeek et al. (2014). It can be measured by the intention of users to recommend the brand to others, as well as the intention to purchase products of the brand.
Afterwards, the next scale aimed at evaluating the perceived credibility of the post. According to Hautz et al. (2014), two main constructs formulate the source credibility. These are the source’s trustworthiness and expertise, on which the questions of this research were exclusively focused on. A scale measuring the two constructs of credibility was extracted from two relevant articles, one is from Ohanian (1990) and the second one is from Hautz et al. (2014). Participants had to indicate in a five-point Likert scale to what extent they agree with the displayed credibility dimensions. Finally, the last questions in the experiment were related to the demographical data of the participants.
28
5.1.2 Data collection: Instagram
The data was extracted from the Instagram platform during the time period between the 11th of April and the 25th of April. Instagram is an internet-based photo and video sharing application founded in 2010 by Kevin Systrom and Mike Krieger in San Francisco ("About Us • Instagram", 2017). In April 2012 Instagram was acquired by Facebook for approximately 1 billion US dollars in cash and stock. So far, the platform has over than 600 million active accounts and its popularity grows in exponential terms. Moreover, through the years many additional features and updates have been integrated in the application. Instagram is considered a new effective marketing and advertising tool for businesses that want to grow their reach and increase their sales numbers (Laurinavicius, 2016).
The four cosmetic brands that were monitored were chosen based on their popularity on Instagram and whether they meet all the criteria that were previously mentioned in the Methodology section. More precisely, the brands should be active on Instagram platform and two of these brands should be using a mixed social media strategy by sharing both generated and brand-generated content, while one of the other two will only be sharing user-generated content and the final one only brand-user-generated content.
The three out of the four cosmetic brands are ranked among the top ten leading beauty brands on Instagram according to the number of followers they have accumulated. These brands are the following: ‘Anastasia Beverly Hills’ (2nd place with 12.900.000 followers), ‘Too Faced Cosmetics’ (5th placewith 8.200.000 followers) and ‘Urban Decay Cosmetics’ (7th place with 7.700.000 followers) (“Instagram: most-followed beauty brands 2017”, 2017). ‘Anastasia Beverly Hills’ is the brand that reposts exclusively content shared by digital influencers. These posts display products of the brand or the influencer himself or herself, using or ‘wearing’ a make-up product. The influencers are usually sharing this type of photos
29
on Instagram and are tagging the brands that they promote. Subsequently, the social media experts of the firm can monitor these posts and then determine if they will repost them on the firm’s account. The other two cosmetic firms are: the ‘Too Faced Cosmetics’ and the ‘Urban Decay Cosmetics’, following a mixed Instagram approach by sharing both content, produced by marketers, as well as content created by influencers. The fourth cosmetic firm in this research is ranked 13th in the list of the most followed cosmetic brands on Instagram, with 3.200.000 followers and it is known as the ‘Make up Forever’ cosmetic firm. This brand was selected, as out of all brands in the list, it was the one with the highest ranking that only shared brand-generated content.
The dataset consisted of four hundred Instagram posts. A hundred posts of each of the four companies was used. The extracted data included basic information, such as the name of the brand, the post description, the post URL, the post type (photo or video), the date, the amount of likes and the amount of gathered comments. In addition, it was determined whether the post was user-generated content by observing either if there was a hashtag, such as #regram or #repost in the description of the post or if there was another account mentioned or tagged, which was different from the account of the brand. Tagged accounts can be easily recognized by the ‘@’ symbol. If it was a repost, then additional data was extracted. This data included the name of the digital influencer, whose post was shared, the influencer’s number of followers and whether the influencer was displayed in the photo or if the photo was simply displaying a product. Finally, each post was assessed according to its perceived quality, which was estimated by monitoring two distinct features: the creativity and the resolution of the photo, as it was mentioned in the literature section.
30 5.2 Instagram data preparation
Firstly, the data was prepared to be imported into the SPSS software, as the data collection was completed in an Excel file. Two new columns were created in order to measure the likes and comments of each post, divided by the followers of each brand. The aim of this calculation was to produce objective results, as each brand had a different number of followers and the posts cannot be compared otherwise. In the next stage, each variable was characterized as string or numeric according to its content and as nominal or scale according to its type of measure. In the preliminary steps of the data analysis a frequency check was completed to identify the percentage of each type (UGC or BGC) of post, as well as to identify any missing value. The results indicated that there were no missing values and each variable had 400 recordings, equal to the number of posts that were extracted. In addition, the frequency check showed that there were 201 brand-generated, and 199 user-generated posts in the dataset.
31
6. RESULTS
6.1 Hypothesis Testing
6.1.1 Testing part of the framework on Instagram data
Figure 2: Conceptual framework - study 1
Hypothesis 1: The user-generated content posted on brands’ accounts, has a greater effect on the consumers’ brand engagement than the brand-generated content, according to the number of likes and comments.
The first hypothesis can be tested in two different ways: by testing the impact of UGC compared to BGC, either on the number of likes or on the number of comments. As mentioned previously, two new indicators were added to the dataset in order to estimate the engagement level of each post, based on its likes and comments. The first engagement indicator shows what percentage of the brand’s followers like the post and the second one what percentage of the brand’s followers comment on the post. As covered before, the percentages are useful in this data analysis, as the dataset consists of four different brands, each one of them with a different number of followers. Thus, it is essential to estimate these percentages, in order to produce representative results.
The hypothesis tested focused on investigating the impact of UGC compared to BGC on the number of likes, which is a type of consumer brand engagement. The user-generated
32
content group (N=199) is associated with a percentage of engagement based on the likes of each post with M=.0091 (SD=.0037). On the other hand, the brand-generated content group (N=201) is associated with an engagement percentage of M=.0064 (SD=.0040). To test the hypothesis that the UGC and BGC were associated with a statistically significant different mean of engagement; an independent t-test was performed in SPSS. The assumption of homogeneity of the variances (null hypothesis) was tested and satisfied via Levene’s F-Test, F(398)=.006, p=936. The independent samples t-test was associated with a statistical significant effect, t(398)=-7.043, p=.00. Consequently, the user-generated content is associated with a significantly larger mean of likes than the brand-generated content.
Furthermore, to test the hypothesis that the UGC has a greater effect on engagement than the BGC, regarding the percentage of comments on each post, a t-test was performed. The same process was conducted starting by the Levene’s F-Test, F(398)=.349, p=.555. The independent samples t-test was not associated with a statistical significant effect t(398)=-.541, p=.589. Thus, the UGC has a statistically similar mean of comments per post, when compared to that of the BGC, meaning that the UGC does not have a greater impact on the volume of comments on each post. The results can be verified by the following table.
Table 1: T-test results comparing UGC and BGC on the level of engagement
Outcome Group 95% CI for
Mean Difference User-generated content Brand-generated content
M SD M SD t Sigg(2-tailed) Likes/followers .0091452 981 .0037115877 .00640297 82 .004065548 1 .003507765, -.001976374 -7.043* 000 Comments/followers .0000368 568 .0000530659 .00003329 77 .000076325 0 .000016493, -.000009374 -.541 0.582 Note: Statistical significance: *p<0.05; **p<0.01
33
Thus, H1 can be partially supported through this dataset. The UGC has a greater effect on the volume of likes on each post; however, there is not a statistically significant difference between the two different groups on the volume of comments.
Hypothesis 2: The power of the influencer moderates the relationship between the user-generated content and brand engagement.
The second hypothesis can be tested through a regression analysis. A hierarchical linear regression was used to investigate the ability of the number of followers of each influencer to impact the relationship between the user-generated content and the brand engagement construct, according to the number of likes and after controlling for the quality of each post. The regression equation for this research is as follows:
Y= α + β1δ + β2X1 + β3δΧ1+ β4X2 + ε
The brand engagement is the outcome variable (Y), α is the constant and ε is the residual or error, which is the part of the results that cannot be explained by the regression model. The ‘δ’ is a dummy variable, with the value of 1 for UGC and 0 for BGC. In this Thesis, the aim is to investigate the influence of the two independent variables, their interaction and the impact of the control variable towards dependent variable. Firstly, the direct impact of the type of post (β1) οn the brand engagement is measured. Secondly, the effect between the
power of the influencer (β2) and the brand engagement is considered. Thirdly, the interaction
effect (β3) of the type of post and the power of influencer is examined. Finally, the effect of
the control variable (β4) is estimated.
Prior to the regression, a correlation analysis was implemented in order to see if there is a possibility of multicollinearity. As we can see from the correlation table (appendix 3), the
34
independent variable referring to the type of post and the moderator do not highly correlate (r=.275) on a significant level (p<.01). Therefore, it is not necessary to aggregate and centre the variables before multiplying them. For that reason, the interaction effect is calculated as the product of the independent variable with the moderator.
In the first stage of the hierarchical linear regression analysis, a predictor was entered, which is referred to as the perceived quality of the post and can be defined as the creative concept and resolution of the photo, included in the post. This model was statistically significant F(1,398)=5.75; p<.05 and explained 1.4% of total variance in brand engagement. The second step of the process involved the introduction of the two independent variables, which are the type of post, either user-generated or brand-generated and the number of followers of the digital influencer that has posted the content. In addition, in the same block of variables, the interaction effect was added. The second step led to a total variance explained by the model as whole, equal to 12% with F(3,395)=17.78; p=.00 < 0.01. This means that there is a large percentage of variance that cannot be explained by the two models. In the final model only the independent variable was statistically significant, with the type of post recording the higher Beta value (β=.35, p=.00) compared to the other variables. In other words, UGC generates 0.35 higher engagement than the BGC, all things being equal. The same result was produced by the t-test analysis, previously. On the other hand, the power of the influencer and the interaction produce insignificant results. This outcome means that the number of followers does not have an impact on the engagement neither does it influence the strength of the relationship between the independent and dependent variable. The results of the regression analysis are illustrated in the following table (Table 2).
35
Table 2: Hierarchical Regression model of brand engagement, based on the number of likes
R R2 R2 Change B SE β t Step 1 0.14 0.12 .014 Quality .001 .001 .119* 2.396 Step 2 .363 .132 .117 Quality .001 .001 .082 1.730 Post type (0=BGC, 1=UGC) .003 .000 .355** 7.227 Power of influencer 2.963E-8 .000 4.634 1.408 Interaction -2.023E-8 .000 -4.727 -1.436
Note: Statistical significance: *p<0.05; **p<0.01
The same analysis was conducted in order to investigate the effect of the moderator on the relationship between user-generated content and the brand engagement, according to the number of comments. In the first stage of the multiple regression analysis, the same predictor was entered, which is referred to as the perceived quality of the post. In the second stage the independent variables were entered, as well as the interaction effect. However, the results stemming from the regression analysis (table 3) indicated that the two models are not statistically significant and thus, there is no meaning in further analysing the results.
36
Table 3: Model Summary with dependent variable the number of comments
Model R R Square R Square
Change
significance
1 0.90 .008 .008 .073
2 .148 .022 .014 .132
Consequently, the hypothesis 2 is not supported by the gathered data, as there is an insignificant relationship between the interaction effect and the dependent variable.
The results of the first research method that was aiming at testing the first two hypotheses are not entirely satisfying. The first hypothesis is partially supported; meaning that the user-generated content is positive correlated with brand engagement and has a larger impact than the brand-generated content, depending on the number of likes. However, the same cannot be verified with regards to the number of comments. In addition, concerning the second hypothesis the results are not statistically significant for accepting the existence of the moderation effect. The variance explained by the model in the regression analysis is also not sufficient for providing general outcomes and conclusions.
37
6.1.2 Testing part of the framework on experiment data
Figure 3: Conceptual framework study 2
For testing the next three hypotheses a Process analysis was conducted in SPSS. The template 4 was used to test the basic mediation effect. The analysed data were collected from the experiment research method. During the analysis, several control variables were tested in order to investigate if they affect the relationships between the other variables. For example, the gender of the respondent and the familiarity with Instagram were added as covariates in the Process; however, the outcome of the analysis indicated that they do not produce significant results. Hence, only the familiarity with ‘MAC Cosmetics’ brand was added as a control variable. The aim of this test was to investigate if the familiarity with the brand that is reposting the user-generated content can influence the engagement level, as well as the perceived credibility. The results indicated that the familiarity with the brand does not have a statistically significant effect on the perceived credibility. On the other hand, the impact of the control variable on the dependent one, which is the brand engagement, is highly significant to the point of p=.0001. An explanation for this outcome is that people, who are to some extent familiar with the brand, are more prone to become engaged and develop some kind of relationship with the brand. To establish brand engagement, it is essential to firstly generate brand awareness across consumers. The following table presents the results from the Process analysis.
38
Table 4: Basic mediation output in Process
Consequent
Credibility (M) Engagement (Y)
Antecedent Coefficient SE Coefficient SE
UGC (X) a1 .1144 .1588 c1’ -.0959 .2280 Credibility(M) - - b1 .7278** .1360 Mac familiarity .2019 .2045 1.2010** .2943 constant i1 3.0646** .1994 i2 .3946 .5053 R2=.3118 R2=.1326 F(2,111)=.7829 p>.005 F(3,110)=16.6125** p=.00
Note: Statistical significance: *p<0.05; **p<0.01
Table 5: Direct, total & indirect effect
Effect SE P Direct effect c1’ -.0959 .2280 p>0.05 Total effect c1 -.0126 .2542 p>0.05
BootSE BootLLCI BootULCI
Indirect effect
a1b1 .0833 .0882 -.1539 .3226
Hypothesis 3: The user-generated content posted on brand’s accounts, has a positive effect on the post’s perceived credibility.
The Process results regarding the path a in the basic mediation, which are displayed in the above tables, are not statistically significant. This means that conclusions cannot depend on this data. Thus, the hypothesis cannot be supported by these results.
39
Hypothesis 4: The posts’ perceived credibility of the user-generated content has a positive effect on the consumers’ brand engagement.
The effect of perceived credibility on the consumer’s brand engagement is the path b1 of the basic mediation and it is statistically significant, with p=.00. The b1=.7278 indicates that
if two user-generated posts differ by one unit in their level of perceived credibility, they are estimated to differ also by .7278 units in their engagement metrics. Therefore, the hypothesis is being supported by statistically significant findings.
Hypothesis 5: The effect of the user-generated content on the brand engagement is completely or partly mediated by perceived credibility.
The second table of the basic mediation effect in Process indicates that the direct, indirect, as well as the total effect are not statistically significant. Thus, the hypothesis that the perceived credibility of the post mediates the relationship between the user-generated content and the brand engagement cannot be supported.
40
Hypotheses Results
H1: The UGC posted on brands’ accounts, has a greater effect on the consumers’ brand
engagement than the BGC.
Partially supported
H2: The power of the influencer moderated the relationship between UGC and
consumers’ brand engagement.
Not supported
H3: The UGC posted on brand’s accounts, has a positive effect on the posts’ perceived
credibility.
Not supported
H4: The posts’ perceived credibility of the UGC has a positive effect on the consumers’
brand engagement.
Supported
H5: The effect of the UGC on the brand engagement is completely or partially
mediated by perceived credibility.
Not supported
41
7. DISCUSSION AND MANAGERIAL IMPLICATIONS
The aim of the two studies was to investigate the impact of the user-generated content that is reposted by brands, on the consumers’ engagement and which factors and mechanisms interact in this relationship. The idea behind this thesis topic is based on the new trend of the digital influencers that are being hired by cosmetic brands, in order to promote products across social media. This firm-initiated WOM initiative is expected to generate more revenues than the traditional advertising, as the content shared by digital influencers is hypothesized to provoke higher credibility.
The results from the analysis of the data extracted from Instagram indicated that the user-generated content is indeed causing higher engagement levels than the brand-user-generated content, according to the number of likes. This outcome confirms the hypothesis from the existing literature and offers meaningful managerial implications. Marketers should leverage this insight, as it offers many opportunities for growth in this highly evolving business world. The firm-initiated WOM through influencer marketing is a lower cost option and simultaneously it has a larger reach than the traditional paid advertising. The target group of this promotion action is accessing social media more than other traditional channels; and hence the potential impact is larger. Therefore, it is essential for brands to make partnerships with digital influencers that share relevant content with the brand’s concept and use creative ideas to promote the products. For example, digital marketing experts should use influencers on Instagram or on YouTube to produce video tutorials of how to use brand’s products. In addition, through partnerships with influencers that acquire high numbers of followers, it would be effective to create contests in order to generate buzz across consumers. Through these contests consumers may obtain free products or have access to relevant events and therefore subsequently increase their positive feelings towards the brand, as well as the
42
probability of generating positive word of mouth and recommending the products to potential customers. These initiatives could lead to higher brand resonance and loyalty, thus contributing to brand’s profits.
With regards to the number of comments, the results were not statistically significant. This outcome can be explained by the different determinants, which are affecting the number of likes and comments on a brand’s account, according to De Vrie, Gensler and Leeflang, (2012). Comments are considered a stronger type of engagement and do not have the same frequency as likes. According to the authors of the article, marketing managers, who want to boost the number of comments in a specific brand post, should include highly interactive features, such as a question or the announcement of a contest. Other interactive characteristics, such as a website link do not have the same effect, as the user navigates away from the brand’s account and does not return to post a comment. However, the major factor
that contributes to the higher number of comments is maintaining the post in the top of the account for as long as possible. This approach is not feasible on Instagram, as the posts displayed in the account are displayed in chronological order. However, if a post has high interactivity on Instagram, it might be displayed in the ‘search’ feature and thus, has higher probability of being seen, even if it is not the most recent post of the account.
The second essential outcome of the study is that the power of the influencer does not significantly influence the strength or the relationship of the shared UGC with the brand engagement. This result can be verified through existing literature, as the ‘3 Pillars of Influence’ concept consists of three different constructs: the relevance of the influencer with the brand’s products, the influencers reach and his or her engagement. In this study only the
reach was considered, as the number of followers that each influencer has acquired was tested as the moderation effect. Marketers should identify people that acquire all three pillars in
