The effect of social media influencer
relevance on online consumer
engagement with product endorsing social
media posts
Maartje van den Maagdenberg 11392476
17-08-2018 – Final version
Universiteit van Amsterdam – Faculty of Economics and Business
MSc. in Business Administration – Entrepreneurship and Management in Creative Industries
Supervisor: dhr. R.G.H.J. (Rens) Dimmendaal MSc Second Reader: dhr. dr. A. (Angelo) Tomaselli
Statement of originality
This document is written by Student Maartje van den Maagdenberg 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
TABLE OF CONTENTS Content Page Abstract 4 Acknowledgements 5 1. Introduction 6 2. Literature Review 9
Social Media Influencer 9
Brand and Product Endorsement 11
Consumer Engagement 15
Relevance of the social media influencer 17
Product category complexity 20
Popularity of the social media influencer 22
The full model 23
3. Data, Measures and Method 25
Social Media Influencers 25
Product Endorsements 25 Independent variables 26 Dependent variables 29 Methods 30 4. Results 32 Hypothesis 1 32 Hypothesis 2A and 2B 33 Hypothesis 3 34 5. Discussion 36
The effect of relevance on relative consumer engagement 36 The moderating effect of complexity of the product category 38
The negative moderating effect of SMI popularity 39
Other explanations for behavior on Instagram 40
Limitations 41 Future Research 44 Managerial implications 45 6. Conclusion 47 Reference List 48 Appendices 54
Appendix 1 – List of social media influencers 54
ABSTRACT
This thesis aims to contribute to the field of influencer marketing answering the question: How does the relevance of the social media influencer on a given topic affect the way
consumers interact with product endorsing content? For this the social media influencer (SMI) is defined as an individual who disproportionally impacts the spread of information by
shaping audience attitudes through the use of social media. For a lot of these SMIs a way to derive economic value from their influence is engaging in product endorsements- being consumer generated content containing publicly shared advertising messages whose subject is a collectively recognized brand. Marketing professionals use product endorsement by social media influencers as part of their campaign strategy to create consumer engagement or interaction with their products. To assess the effectiveness of a social media influencer as an endorser of a product, relevance is introduced as predictor variable, hypothesizing that it has a positive influence on relative consumer engagement (engagement per 10 000 followers). The complexity of the product endorsed and the overall popularity of the SMI need to be
recognized as moderating variables in this relation to come to a full model of the influence the relevance of the social media influencer has on relative consumer engagement.
ACKNOWLEDGEMENTS
For this thesis, data was made available by the Brand Communications team of L’Oreal Netherlands Consumer Products Division. I want to thank them for the opportunity to work with this very interesting data obtained from the Traackr database and sharing with me your state of the art influencer strategies. Special thanks to Brand Communication Manager Silvana Hoekstra for taking me under your wings and showing me the ropes on influencer marketing. I also want to express my gratitude to Nicolette Ognjanovsky, with your endless pink marker helping me revise. Lastly, I want to thank my thesis supervisor Rens
Dimmendaal for his patience and willingness to help me bring this research project to a (hopefully good) ending.
1. INTRODUCTION
In recent years marketing and branding professionals have discovered influencer marketing as a new way to reach audiences that are hard to reach with traditional advertising methods (de Veirman et al., 2017). Social networking site Instagram reported surpassing 700 million monthly active users and 1 million active users in 2017 (Our story, 2018). With the growth of the platform and the managerial interest in it, also came the growth of scholarly interest. Behavioral scientists were interested in reasons to use the platform (Sheldon and Bryant, 2016), and reasons it is to utilized to grow one’s celebrity (Zamudio, 2015; Marwick and boyd, 2011). Interesting parallels with celebrity culture (Banister and Cocker, 2013; Page, 2012) and widely researched celebrity endorsements (Kamins, 1990; McCracken, 1989; Rice et al., 2009; Jin and Phua, 2014; Biswas et al., 2006) sparked interest in influencer marketing from Academia. De Veirman et al. (2017) for example researched the impact of the number of followers of Instagram users on brand attitudes, finding a connection to literature about Word of Mouth; a product related statement made by a customer about a brand or a company (King et al., 2014). Influencer marketing essentially balances on the thin line between celebrity endorsement and consumer generated word of mouth reviews. Endorsers and specifically Social Media Influencers can have significant influence on brand attitudes and brand equity (Spry et al., 2011) but how this influence works and what traits of the social media influencer affect this influence is still under researched.
However, within the scope of celebrity culture and celebrity endorsement some interesting theories and contributions were made concerning endorser effectiveness in advertisement (Ohanian, 1991; Amos et al., 2008; Till and Busler, 2000) and their
effectiveness for shaping brand equity (Spry et al., 2011). With the emergence of social media platforms, a new type of celebrity arose, celebrity was no longer only for the famous people consumers knew from movies, TV, or the music they listened to, but could happen to people
appearing to be just like them; fellow consumers who gained a large following on a social networking site. Are those a priori mentioned theories in the scope of traditional celebrity culture applicable to the new “instafamous”, the group of celebrities of the internet? Or do we need to build influencer marketing theory based on consumer marketing theories such as electronic Word of Mouth (eWOM) with the knowledge that some consumers have bigger reach and influence than others?
The purpose of this thesis is to contribute to answering these broad questions by addressing the question: How does the relevance of the social media influencer on a given topic affect the way consumers interact with product endorsing content? This thesis addresses this question by using data from the influencer database Traackr. As the main effect, the influence of the relevance of the social media influencer on the relative consumer engagement (calculated by the sum of likes and comments). Because this effect can be influenced by different factors, also the moderating effects of product category and total number of followers of the social media influencer were incorporated in a final model. The research concludes that there is a statistically significant effect of relevance on relative consumer engagement, but the direction of this effect is contradicting throughout the research. Also, positive moderating effects of product category complexity, and negative moderating effects of social media influencer popularity were supported by the data. The results of this research are coherent with hypothesis based on theories derived from celebrity endorsement. This implicates that research in influencer marketing shows interesting parallels with research in celebrity culture and would fit as an addendum to this stream of research as well as to streams of social media marketing research or eWOM research.
This research is conducted in the scope of the Dutch Beauty Industry. Worldwide the beauty industry was estimated at 205 billion euros in 2017 and is estimated to grow at a rate of 4 percent (L’Oreal Group, 2018). The beauty industry is a very interesting empirical setting
to investigate influencer marketing. In 2018, consulting group Celebrity Intelligence found that 70% of the 385 marketing professionals in the beauty industry who were interviewed, thinks their budget for influencer marketing will increase, more than half say that currently between 10 and 30 percent of their budget is spent on influencer marketing.
The report mentioned Instagram as top social media platform for marketing efforts in the beauty industry, exceeding Facebook, YouTube and Twitter (Celebrity Intelligence, 2018). Most marketing professionals in the report (67%) say they take a data-led approach to selecting the most suited social media influencer for their campaign, giving this research real world application. Other selection criteria mentioned were social media influencers talent or skill (46%), and content themes and areas of expertise (43%). This thesis aims to tie those managerial practices together and provide scientific insights to corroborate these practices
The data for this research was provided by the L’Oreal Group who provided access to the Traackr database and disclosed their collaborations with social media influencers. The L’Oreal Group is the biggest corporate brand in the cosmetics market worldwide (L’Oreal Group, 2018) with an operating profit of 4.68 billion euros worldwide. They have been on the forefront of digital advertising, allocating 38% of their media budgets to investments in digital opportunities (L’Oreal, 2018). With expanding media budgets allocated to digital
advertisement opportunities, L’Oreal group has been a big commissioner of collaborations with social media influencers.
2. LITERATURE REVIEW
The scope of this research is product endorsement by social media influencers. Therefore, we first aim to define those concepts and try to place them in the current research paradigm. After that the hypotheses will be introduced.
Social Media Influencer
Influencer marketing is a relatively new term to marketing research (De Veirman et al., 2017) It combines insights from a wide range of research fields including advertisement (Amos et al., 2008; Araujo et al., 2016; de Veirman et al., 2017; McCracken, 1989), marketing (Banister and Cocker, 2013), electronic word of mouth (eWOM) (Hennig-Thurau et al., 2004; King et al., 2014) and behavioral research (Belk, 2014; Djafarova and Rushworth, 2017, Kamins, 1990). These various streams of literature all use their own terminology for the phenomenon that this thesis will refer to as the Social Media Influencer (SMI). The use of the term SMI has become commonplace in our modern environment and is often used interchangeably with simply ‘influencer’ (Bakshy et al., 2011; Sheldon and Bryant, 2016; de Veirman et al., 2017) in lay speak.
The basis of all of these definitions is derived from that of an “opinion leader” as defined by Katz and Lazarsfield (1955): “Someone who is able to influence his or her close personal ties by exerting social pressure and social support.” Even though this definition was formulated long before the emergence of the internet, the concept of an opinion leader is still embedded in more modern definitions. A widely used definition is: “Individuals who disproportionally impact the spread of information or some related behavior of interest” a notion adopted by Bakshy et al. (2011) in Influencer and applied to modern scenarios. These definitions include both strong ties- such as ordinary friends who communicate with each other on social
journalists,(semi-) public figures, and celebrities who share their opinion with an anonymous audience (Bakshy et al., 2011). The scope of this research for the purposed of this thesis will be limited to (semi-) public figures influencing an anonymous audience; therefore, the definitions given by Bakshy et al. (2011) and Katz and Lazarsfield (1955) are too broad and we must give a more focused definition.
Another distinctive characteristic for the social media influencer is their use of social networking sites (SNS) to rise their celebrity status. Freberg et al. (2011) define the social media influencer as “a new type of independent third-party endorser who shape audience attitudes through blogs, tweets and the use of other social media”. By this definition, followers are thought of as fans, rather than friends and family. This asymmetrical status is reason to believe that social media influencers have some commonalities with celebrities (Marwick and boyd, 2011). Freberg et al. (2011) also defines the social media influencer as an endorser. An endorser is a human brand (Thomson, 2006) that engages in a partnership with a corporate brand “to communicate the merits of the corporate brand or its products” (Zamudio, 2015) and establish positive consumer attitudes (Kamins, 1990). Thereby they limit the scope of the social media influencer to those who engage in (mostly paid) partnerships with brands and have monetized their social media profiles. But this is not necessarily always a defining character of the social media influencer, there can be examples found of individuals who have a disproportionate influence over anonymous others through their profiles on social networking sites but do not monetize this influence by endorsing brands or products.
The research by De Veirman et al. (2017) narrows down the scope of the definition by adding the concept of “content creation” to the definition of influencer. This addendum to the definition posits that one of the traits of an influencer is that they produce the content they post themselves (De Veirman et al., 2017). They do this by relating the emergence of influencer marketing to electronic word of mouth (eWOM): “Any positive or negative statement made by
potential, actual or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet” (Hennig-Thurau et al., 2004, p. 39; King et al., 2014). In this context, in influencer is a consumer who produces word of mouth that reaches a disproportionate amount of people (De Veirman et al. 2017; Katz and Lazarsfield, 1955). As a result, celebrities that star in other people’s content (for example, a made-for-social-media campaign, by a well-established company) but still post this on their own online profiles instead of creating their own content or voicing their own word of mouth, would not be social media influencers according to this definition. Therefore, in our context a social media influencer is not necessarily a content creator.
The scope of this research is product endorsement by social media influencers. So for the purpose of this research, we adopt the following definition:
Individuals who disproportionally impact the spread of information by shaping audience attitudes through the use of social media.
Brand and Product Endorsement
The concept of brand and product endorsement is very broad. It reaches from (electronic) word of mouth, used by consumers to voice their opinion about a brand or product (Hennig-Thurau et al., 2004; King et al., 2014) to advertisement practices by corporate brands (Zamudio, 2015). For this research it is important to separate two types of endorsement, the one initiated by the social media influencer and brand endorsement initiated by the brand. Product endorsement initiated by the social media influencer is a form of eWOM; communication from a (fellow) consumer about a product or service they have purchased (Hennig-Thureau et al., 2004). Hennig-Thurau et al. (2004) found that self-interested consumers make up the biggest group of eWOM senders. This would mean that social media influencers talk online because of their own interest and less because they feel the need or obligation to help others by providing
their opinions or expertise (Hennig-Thurau et al., 2004; Sheldon and Bryant, 2016). Communicating about a brand or product can stem from a true liking or positive attitude towards the product but also from a self-interested need to communicate one’s individual identity (Kapitan and Silvera, 2015). People use different self-representation strategies to express themselves, or to make their identities tangible. One of the ways by which to do this is associating themselves with material objects and places (Schau and Gilly, 2003). This can be embodied by possessions, or pictures with objects that reflect individual identity by demonstrating a person’s accomplishments, tastes or unique creative efforts, which for the purpose of this thesis includes content creation or curation (Schau and Gilly, 2003). Therefore, endorsing a product on a social media profile and communicating eWOM about it can be a self-representing strategy (Schau and Gilly, 2003).
eWOM is a valuable and reliable source of information for new customers in a (pre-) purchase stage (King et al., 2014), and eWOM referrals have strong impact on new customer acquisition (Trusov et al., 2009). This means that a product shown by a social media influencer is likely to enter a potential consumer’s consideration set (Keller, 2003) if they follow, like, or otherwise have confidence in their opinion. The content posted by fellow consumers is seen as more reliable and trustworthy than the same message appearing in market generated content which refers to content distributed by the company that also makes the product (Mangold and Faulds, 2009). Overall, personal communication is more powerful in affecting attitudes compared to mass media communication (Katz and Lazarsfeld, 1955). Adding to that, elasticity, the magnitude of consumer response to a change in the status quo (You et al., 2015), is found to be about 20 times higher for eWOM than other marketing activities (Trusov et al., 2009). Marketing professionals recognize the power of this consumer generated content, but the brand managers and marketers often struggle with controlling or directly shaping eWOM (King et al., 2014). This can be due to the issue of credibility. For example, reviews posted on independent
review websites have a higher volume and valence elasticity than reviews on a company website (You et al., 2015). This indicates that consumers trust fellow consumers more than they will trust the information provided by the company. With influencer marketing it is easier to shape the valence of the message (Uzunoglu and Kip, 2014; Katz and Lazarsfeld, 1955), as social media influencers who post about products on their social media profiles are, by consumers, often seen as fellow social media users/peers (Jin and Phua, 2014). Therefore, their eWOM about brands merits more credibility and trust compared to that of the company directly, whom many may view as just trying to sell goods (Jin and Phua, 2014). By providing influencers with test products, inviting them to exclusive events, or paying them, brands stimulate them to endorse their products in manner that is viewed as more trustworthly or honest, by consumers (Uzunoglu and Kip, 2014).
This brings us to the second type of product or brand endorsement; endorsement initiated by the corporate brand. One defining characteristic of a social media influencer as mentioned before is that a they attract large scale attention from the public (Rindova, Pollock and Hayward, 2006). This gives them a unique position of disproportionate influence over other consumers who talk online (Bakshy et al., 2011). Ergo, the greater the number of people who know and pay attention to a certain social media influencer, the greater the value of that social media influencer is to a brand (Rindova, Pollock and Hayward, 2006). Traditionally, these audience sizes were only seen at the level of celebrity. Using celebrities to endorse a product is and has been a relatively common practice in advertisement (eg. Centeno and Wang, 2016; Jin and Phua, 2014; McCracken,1989; Rice et al., 2009; Thomson, 2006). Celebrity being “a property of the actor’s relationship with an audience rather than a characteristic of the actor himself” (Rindova, Pollock and Hayward, 2006). A celebrity therefore is a well-known individual who receives significant media-attention who has this type of relationship with their audience (Rindova, Pollock and Hayward, 2006; McCracken, 1989). Most of the traditional
celebrities include public figures such as Royals, TV and film stars, athletes, and musicians. Recently there has been a rise of new digital celebrities such as bloggers, vloggers, and Instagram personalities (Djafarova and Rushworth, 2016) claiming their place in a continuum of celebrity. They do so by a practice of micro-celebrity; a process to create a celebrified online persona (Page, 2012). They commodify themselves (Thomson, 2006) and construct personal brands around their social media profiles (Olausson, 2017). On these profiles the self-presentation is carefully curated to be consumed by the audience- their followers. By making their everyday life choices and values, things that are intrinsically private, available to the public they endear themselves to their followers, ensure their image as “fellow social media user”. At the same time they create brand identities (Thomson, 2006), which are carefully constructed and maintained to maximize their economic and social value (Centeno and Wang, 2016).
Social media influencer endorsement has blurred the line between the consumer and the advertisers by creating consumer-generated advertising (Jin and Phua, 2014). These ads are consumer generated content containing publicly shared advertising messages whose subject is a “collectively recognized brand” (Campbell et al., 2011) and are usually seamlessly woven into the daily narratives and posts by the social media influencer (de Veirman, 2017). This is called native advertisement, it is perceived as more credible and trustworthy by the consumer (Jin and Phua, 2014) and classified as eWOM rather than direct advertisement (King et al., 2014) in the mind of the consumer. The balance between personal content and content containing consumer generated advertisement needs to be carefully curated (Rice et al., 2011). Endorsment of too many brands is found to be harmful to both the brands endorsed and the social media influencer (Rice et al., 2011) by endangering the trustworthiness of the social media influencer (Amos, et al., 2008).
For the purpose of this research, the definition by Campbell et al. (2011) of Consumer Generated Advertisement fits with the concept of brand or product endorsement to be discussed: product endorsement is consumer generated content containing publicly shared advertising messages whose subject is a “collectively recognized brand” (Campbell et al., 2011) and are usually seamlessly woven into the daily narratives and posts by the social media influencer (de Veirman, 2017).
In the case(s) presented henceforth the social media influencer willingly communicates about a certain brand or product, integrating this in the content published on their social media profiles. Some social media influencers in the sample got a monetary payment to endorse a specific product on their social media page, others were sent a free test product as compensation for endorsing the product on their social media profiles. As addressed in this thesis, all the endorsements were brand initiated, but this is not necessarily the case for all products or brand endorsements (Kapitan and Silvera, 2015).
Consumer Engagement
One of the a priori mentioned characteristics of the social media influencer is the disproportionate number of people that follow their social media accounts, when compared to the average social media user (Bakshy et al., 2011). This number of followers is what attracts the attention of marketing or brand managers (Uzunoglu and Kip, 2014). In case of a brand endorsement, all of these followers could potentially see the advertisement and be turned into potential customers. The interaction that the consumer has with the post endorsing the product is at the center of the concept of consumer engagement. Vivek et al. (2012) define consumer engagement as “The level of the customer’s (or potential customer’s) interactions and connections with the brand offerings or activities, often involving others in the social network created around the brand”. At the center of this definition is the consumer’s interaction with the
product or service offered (Dessart et al., 2014). They acknowledge in their definition that others in the social network around the brand can be involved in this engagement, meaning that interaction with a brand on the social media profile of a social media influencer is a form of consumer-brand engagement in much the same way as direct brand advertising.
There are many ways consumers can engage with a post they encounter on social media and these vary across and between individual social networking sites. For example, on social networking site Instagram only the number of likes and comments can be obtained by a third party and are therefore the forms of engagement that are focused on for marketing purposes. Kim and Yang (2017) found that different engaging behaviors (liking, commenting and sharing) have different underlying psychological drivers. Liking a post has a more emotional driver. They found that photos elicited more likes than posts containing only text on social media. This is corroborated by evidence from research in the field of journalism studies where photographs are able to elicit more emotional responses (Brantner et al. 2011). Based on this evidence, liking a photo is a sign of affection towards the content portrayed in the picture (Kim and Yang, 2017). Comments, on the other hand were found to be much more cognitively triggered (Kim and Yang, 2017). However, unlike commenting and liking, viewing a post, measured by impressions, is seeing a post, but choosing not to actively engage with it, is not accessible to a third party. As such it is hard to assess the number of actual followers that viewed the post, since this data is only disclosed to the owner of the account (Kim and Yang, 2017). As a result, for data acquisition, some researchers and practitioners use the ‘reach’ of a social media influencer. The reach is defined as the total number of people following that account (Grave, 2018) and is used as an indication of the potential number of impressions.
Marketing professionals use not only the total engagement; a sum of the likes and comments (Grave, 2018) attributed to a certain post or photograph, but also the engagement rate. To obtain the engagement rate the total engagement is divided by potential number of
impressions- the reach-, or total audience of the social media influencer. Therefore, the engagement rate represents the percentage of the audience of the social media influencer that has engaged with the content posted (Grave, 2018). Another metric used is the relative engagement. This is the engagement on a post corrected for the number of followers a social media influencer has (Grave, 2018) For SMIs, engagement per 10.000 followers is widely used as a benchmark of relative engagement (Traackr, 2018). Both engagement rate and relative engagement allows marketing managers to compare the engagement of different social media influencers.
The comparison of different social media influencers on their relative engagement raises questions about what could drive this difference in engagement between different social media influencers. This research aims at providing some explanation for the difference in variance of the consumer engagement on different posts, taking various predictors such as SMI relevance, product category and SMI popularity into account.
The relevance of the social media influencer
In the context of endorsement, the concepts of expertise and trustworthiness are often named as an explanatory factor of endorsement effectiveness (Ohanian, 1991; Goldsmith et al., 2000; Amos et al., 2008; Till and Busler, 2008). Expertise is defined as “the extent to which a communicator is perceived to be a source of valid assertions” (Hovland et al., 1953, p.21; Winterich et al., 2018) and trustworthiness is defined as: “the degree of confidence in the communicators intent to communicate the assertions he or she considers most valid” (Hovland et al., 1953, p.21; Winterich et al., 2018). Central to both of these definitions is the extent to which the consumer finds the communicator of the message, in our case the social media influencer, a valid source that is trusted to convey honest information (Winterich et al., 2018)
Trustworthiness and expertise are found to be the strongest predictors of attitudes of consumers towards product endorsement (Amos et al., 2008).
These assumptions are in line with source credibility theory, which refers to the whether the source appears to have some kind of relevant expertise on the subject matter and is trusted to communicate honestly about this. (Winterich et al., 2018; Ohanian 1990) It posits that that more credible sources are more effective product endorsers than less credible sources (Kelman, 1961).
However, there are additional factors beyond trustworthiness and expertise that affect engagement with product endorsement. The source attractiveness model (Joseph, 1982; Hovland et al., 1953) hypothesizes that not the expertise but the perceived physical attractiveness of the endorser determines the effectiveness of the endorsement (Joseph, 1982). Indeed, Biswas et al. (2006) found that endorsements where the endorser has high perceived expertise are not always more effective than a highly attractive endorser. Biswas argues that it is not expertise nor attractiveness alone that serves as the determining factor of the endorsement effectiveness but the perceived congruency between the endorser and the product (Biswas et al., 2006).
A second theory used among extant literature is the match up hypothesis. The match up hypothesis (Kamins, 1990; Till and Busler, 1998) suggests that the fit between the endorsement and the endorser can be measured by both attractiveness and expertise but that the brand-endorser match is the most important determinant of brand-endorser effectiveness (Till and Busler, 1998). One of the hypotheses produced within this theoretical frame is that “more attractive endorsers are more effective when endorsing products that are used to enhance one’s attractiveness” (Kamins, 1990). McCormick (2016) found that brand-endorser match contributed to attitude formation: When the consumer views an image of an unfamiliar celebrity and the image of the product as being congruent or having a good fit, the attitudes towards the
brand would still be positive. Therefore, we pose that the match, congruency or fit between the endorser and the product endorsed has a positive effect on consumer attitude measured by consumer engagement
Different sources in sponsorship literature find the importance of fit in that field as well. Sponsorship is a form of communication where a brand or firm ‘sponsors’ via cash or product compensation an individual as a means by way to maximizing the sponsee’s commercial potential through use of their reach via social media platforms. (Cornwell et al., 2005; Olson, 2010). The effects of sponsorship-fit found in this context could be applied to social media influencers and brand match. A higher fit is found to be related to higher effects, both in sports and non-sports contexts (eg. Cornwell et al., 2005; Olson, 2010).
Previously we addressed the similarity between the traditional celebrity and the social media influencer. Therefore we pose that using the source credibility model, as used within celebrity endorsement and advertising reserach also fits the scope of product endorsements by social media influencers. The match between the brand and the content created by the social media influencer is one of the essential criteria of selection by marketing managers (Uzonoglu and Kip, 2014).
Within the scope of social media influencers, credibility and match can be addressed by the concept of relevance. The relevance of a social media influencer is the contextual affinity of the social media influencer on a given topic (Mulenok, 2018). This contextual affinity is made up of the frequency a social media influencer posts about that given topic (Mulenok, 2018). The more frequent a social media influencer posts or talks about something, the more likely they are to enter the mind of the consumer associated to that given topic (Keller, 2003) increasing their perceived credibility (Biswas et al., 2006) and perceived fit (Till and Busler, 2000) This leads us to believe that an increase in the relevance score attributed to the social media influencer leads to an increase in the relative consumer engagement on that post.
Hypothesis 1 (H1)
The relevance score of a social media influencer positively influences the relative consumer engagement on a post containing a product endorsement.
Product category complexity
According to categorization theory, people understand the world around them by categorizing it (Sujan, 1985), and new stimuli are placed within existing categories. This works the same for product categories (Chaudhuri and Holbrook, 2001). The previous knowledge of a certain product category will determine the type of evaluation the brand stimulus will evoke (Chaudhuri and Holbrook, 2001). Therefore, previous knowledge about different product categories could influence the way consumers process product endorsing posts on social media profiles of social media influencers.
It might be possible that for some subjects or products the relevance of a social media influencer is more important than for other subjects. This could be the case for product categories where for example expertise on the subject is more important to be a valuable source of information (Biswas et al., 2006) This would be the case with for example skincare products. When they are recommended by a expert beautician, the information conveyed feels a lot more valuable to a consumer.
An important determinant of the expertise needed to endorse a product is the perceived risk a consumer feels when potentially purchasing a product (Biswas et al., 2006, You et al., 2015). This perceived risk can be divided in performance risk and financial risk. Performance risk is “the risk associated with uncertainties regarding the product not performing according to expected levels” (Biswas et al. 2006). Financial risk is “the risk associated with the costs and expenses involved with the product, and uncertainties whether the product is worth that amount of money” (Biswas et al. 2006). For the products involved in this research, performance risk is
the most relevant. Some product categories require higher levels of expertise to assure the performance of the product. In the study by Biswas et al. (2006) a division between high and low technology-oriented products is made. They find that for products with high performance risk, the expertise of the endorser is more important, for example, when buying a new computer or laptop, expertise of the endorser minimizes the risk that te computer will not perform according to your wishes as a consumer.
The expertise of the social media influencer, as well as their trustworthiness and their fit is addressed by the concept of relevance. In this project, where beauty product endorsments are assessed, the highest performance risk is found in high complexity products. High complexity products are products that require a certain skill or knowledge when purchasing or using them such as skincare or make-up products. Low complexity products on the other hand are usable by anyone without particular skill or knowledge and will perform satisfactory for them. Examples of those product categories are haircare-such as shampoo- or nailpolish.
Because of the associations of complexity related to those product categories, consumers are more likely to process encounters with those product in a complex category. For complex products in which more expertise is required to be a trustworthy source of information, users want to engage with influencers on both a affectionate (liking) or cognitive (commenting) level. Therefore we propose that for endorsement of high complexity products, the influence of relevance on relative consumer engagement is bigger than would be the case for low complexity products. Subsequently we pose that the opposite is the case for low complexity products. Meaning that relevance is less important to relative consumer engagement when a low complexity product is endorsed.
Hypothesis 2A (H2A)
The relationship between relevance and relative consumer engagement is positively moderated by high complexity of the product endorsed.
Hypothesis 2B (H2B)
The relationship between relevance and relative consumer engagement is negatively moderated by low complexity of the product endorsed.
Popularity of the social media influencer
A criterion that has been used a lot by practitioners to assess whether a social media influencer could be effective as a product endorser is total number of followers (Uzunoglu and Kip, 2014; Zhang and Dong, 2008). The number of followers reflects the total network size of the social media influencer and serves as an indication of their popularity (Djafarova and Rushworth, 2017; de Veirman et al., 2017, Jin and Phua, 2014, Utz, 2010)
Quantitative indicators of social influence such as the total number of followers on a social media account can also be used as cues of source credibility (Jin and Phua 2014). Consumers may perceive a SMI as more trustworthy and competent, suggesting that audience size, or number of followers serves as a first cue or proxy for source credibility, before expertise, trustworthiness, and fit. (Jin and Phua, 2014). It was indeed found that there exists a significant correlation between perceived source credibility and number of followers (Jin and Phua, 2014). Araujo et al. (2016) found that in the context of Twitter users with a high number of followers are more likely to be retweeted after tweeting brand messages, compared to users with strong ties, such as offline friendships, tweeting brand messages to each other. This indicates that users are more susceptible of brand messages tweeted by a social media influencer or celebrity with a large number of followers than when its tweeted by one of their friends (Araujo et al., 2016). This corroborates the hypothesis that the total number of followers or
popularity serves as a cue for consumers to assess the credibility of a social media influencer (Jin and Phua, 2014, Amos et al., 2008; McCormick, 2016; de Veirman et al., 2017). Take for example George Clooney, endorser for Nespresso coffee. He is not necessary an expert on coffee or a known connoisseur, but he is very well-known. Does this mean that at some height of fame, it doesn’t matter anymore whether the endorser has relevant expertise, their fame takes over as a cue to the consumer? It has also been shown that top expertise endorsers are more influential across different categories than smaller in-category endorsers (Zhao et al. 2015). This suggests that consumers are willing to engage affectively (liking) or cognitively (commenting) with a post even though the fit between the social media influencer and the product they are endorsing, but may not be an expert in, because of the popularity of the influencer themself. This indicates that more popular social media influencers can get away with more product endorsement before losing their perceived trustworthiness.
This hypothesis is directly contradicted by Rice et al. (2011), who has found that as the number of brands endorsed by a single celebrity increases, the attitude of consumers towards that celebrity decreases (Rice et al., 2011). This suggests that even though a popular endorser can endorse more brands, it is not always the best strategy to do so.
This leads us to pose that when the social media influencer has more followers and following Utz (2010) therefore is more popular, the effect of relevance on relative consumer engagement will be less.
Hypothesis 3 (H3)
The relationship between relevance and relative consumer engagement is negatively moderated by the popularity of the social media influencer.
The full model
Figure 1 - Full model of hypotheses
The aim of this research is using relevance to explain some variance in relative consumer engagement, acknowledging that this relationship might be moderated by product category or social media influencer popularity.
According to H1. The higher the relevance of the social media influencer on a given subject, the more consumers will engage with this post by liking or commenting on it.
The moderation of product category is hypothesized by the complexity of the product category in H2A and H2B. For high complexity product category, the relevance of the social media influencer has a greater explanatory value than for low complexity categories. Meaning that when endorsing more complex products, being relevant to the given topic is more
important.
Product category complexity (High | Low)
Popularity of the social media influncer (Number of followers)
Relevance of the social media influencer (Relevance score)
Relative consumer engagement (Sum of likes and
comments) +
-+
-3. DATA, MEASURES AND METHOD
For this study, data was obtained from the Traackr Database with help from the Brand Communications team in the Consumer Product Division (CPD)of L’Oréal Netherlands. This database contains data about 600 different social media influencers, selected by the L’Oréal Benelux Brand Communications teams. For this data set only the SMIs who were known by the Brand Communications Team of the aforementioned CPD was used. Within these SMIs, 262 traceable brand endorsements were picked for this research. Posted were deemed traceable if the SMI tagged them with hash-tags that were related to the product.
Social Media Influencers (SMIs)
The sample used for this research contains information 66 female SMIs who created 262 Instagram posts between 2017 (144) and 2018 (118) that contained a brand endorsement for one of the four CDP brands; L’Oréal Paris, Maybelline New York, Garnier or essie. A complete list of the social media influencers used in this study can be found in Appendix 1. Since January 2017, L’Oréal Netherlands disclosed 114 brand endorsements for which the SMI had been commissioned and received monetary compensation. Utilizing the search function of the Traackr database, 148 additional, nondisclosed brand endorsements were found. These brand endorsements were not paid for, but rather the SMI received the product mentioned as a gift- with the intent to use it as a test-subject. This means that all the product endorsements in the sample were brand-initiated.
Product Endorsements
The product-endorsements in the sample are categorized per brand and per campaign within that brand. For L’Oreal Netherlands it is a general marketing practice to include the sending of test-products to a selection of SMIs as part of a campaign or product launch. These PR-gifts are well documented; therefore, they were able to be tracked for this research. An
overview of the different brands, campaigns, and the product category the endorsed product belongs to can be found in table 1.
Table 1 - Overview of product endorsements
Brand Campaign Product Category
Frequency* Percent
Maybelline Maybelline Squad Make-up 107 40.8
Gigi X Maybelline Make-up 15 5.7
Master Holographic Make-up 5 1.9
Total Temptation Make-up 3 1.1
Total 126 48.1
L’Oreal Paris Stylista Haircare 27 10.3
Colorista Haircare 24 9.2
Hydra Genius Skincare 8 3.1
X Fiber Mascara Make-up 8 3.1
Color Riche Shine Lipstick Make-up 7 2.7
Balmain Lipstick Make-up 5 1.9
Elvive Dream Lengths Haircare 4 1.5
Sugar Scrub Skincare 3 1.1
Infallible Eyepaint Make-up 3 1.1
Miss Babyroll Mascara Make-up 2 0.8
Elvive Phytoclear Haircare 1 0.4
True Match Foundation Make-up 1 0.4
Total 98 37.4
Garnier Ambressadeur Skincare 13 5.0
Loving Blends Haircare 7 2.7
Micellair Gelwash Skincare 2 0.8
Power of Micellair Skincare 1 0.4
Total 22 8.4
Essie Essie Kingsday Nailpolish 6 2.3
Essie Treat Love Color Nailpolish 5 1.9
Essie Winter Nailpolish 5 1.9
Total 16 6.1
TOTAL 262 100
* The total number of product endorsing posts within the campaign.
Independent variables
Relevance
To assess the credibility and fit of the social media influencer, the relevance score is used. Relevance is the contextual affinity of the social media influencer on a given topic- which is defined by a range of keywords and can be found in Appendix 2. The Traackr algorithm calculates a relevance score based on these keywords between 0 and 100 across four product
categories- make-up, skincare, haircare and nail polish. The Traackr algorithm factors in keyword mentions, keyword diversity, content production rate, the freshness of content, and some contextual measures. Keywordmentions refers to the total number of times the keyword
is mentioned in total and within one unit of analysis (eg. One blogpost). Keyword diversity is a measure borrowed from Search Engine Optimization practices and weighs in the number of times related keywords or synonyms are used instead. Content production rate indicates how often the SMI posts new content, both related and unrelated to the given keywords. The freshness of the content refers to the time that has decayed since the production of the content. Most algorithms such as for example Facebook’s Edgerank consider older content less important than new content (Bucher, 2012). Contextual measures weigh for example where the keyword was placed in the text. Keywords mentioned in titles are considered more important than keywords mentioned in the last paragraph of the text (Mulenok, 2018). It assesses these measures across all known social media platforms (e.g. Facebook, Twitter and Instagram), and websites of the SMI. As a result of this factor analysis, a SMI who posts a lot about make-up products will receive a higher relevance score than a social media influencer who posts primarily about fashion. It is important to note that keywords contained no brand names, sub-brands, or product names to assure that SMIs whom had posted about L’Oreal brands previously did not receive higher scores. Additionally, both English and Dutch keywords were used, since Dutch social media influencers use both Dutch and English keywords in their content creation. To minimize the chance of relevant keywords being left out, different ways of spelling were used. To make sure keywords wouldn’t be too broad, for keywords consisting of multiple words, exact matches (using “…”) were used. Table 2 contains the descriptive statistics for the variable “relevance”.
Product Category Complexity
The four product categories mentioned before- make-up, skincare, haircare and nail polish- were divided into a dummy coded variable based on their complexity. Make-up and skincare are hypothesized to be the more complex product categories featured in this research as they require a better developed skillset to use them (make-up; also including the need of further application tools) or there is an inherent difficulty in choosing the right product (skincare). Therefore, make-up and skin care have a higher performance risk compared to haircare and nail polish, which were dummy coded as low-complexity products. A haircare product (e.g., shampoo) does not take extraordinary skill to purchase nor use, with the same being true for nail polish. For all four product categories some obvious exceptions can be made. For example, professional application of intricate nail art may require extra precision and practiced skill but did not influence the above parameters for classifying a product as high or low-complexity. The products endorsed in the posts sampled for this research do not cover this use of niche products and were categorized for complexity on general usage (i.e. simply covering a nail with a coat of polish). Descriptive statistics for product category complexity can be found in Table 2.
Popularity of the social media influencer.
To calculate the popularity of a social media influencer, their total number of followers was used. The total number of followers was downloaded all at the same time and was not corrected for temporal changes. Therefore, there is no correction for the number of followers a SMI had at the time of her endorsed posting compared to when the data for number of followers was extracted. The descriptive statistics for the total number of followers can be found in Table 2.
Dependent Variables
Relative Consumer Engagement
Total consumer engagement is calculated by the total of the likes and comments on the Instagram post containing the brand endorsement. Since a SMI with a large number of followers is more likely to have more likes and comments on a post, this variable needs to be adjusted to compare across SMIs with different audience sizes. A commonly used practice to atone for this is to use relative consumer engagement. This measures the consumer engagement on a post per ten thousand followers.
The relative consumer engagement used in this analysis is computed by summing the likes and comments on a post and correcting them for the total number of followers. To measure if both likes and comments are measuring the same concept, namely relative consumer engagement, a Cronbachs alpha test is used. Results show that Cronbachs alpha would be too low in this data set to consider likes and comments measures as being both a part of the same concept (α = 0,207), namely relative consumer engagement. Having discussed the different reasons for consumers to like (show affection) and comment (cognitive triggers) on a post, this could be an explanatory theory of the two variables not exactly measuring the same concept. However, because of the commonality of calculating engagement this way in managerial practice (Grave, 2018; Celebrity Intelligence, 2018), we decided to continue using this variable for the realworld application of this work. Using relative consumer engagement as a sum of likes and comments will resonate more with previous research into this topic.
Lastly, since data was downloaded all at once, there is no distinction made in how old the post is and the relation of time to the number of likes and comments. Since data was collected from a third party and not from the SMIs themselves it is almost impossible to obtain the number of followers they had at a certain time in the past. This method of data collection did not skew the information in any ways, as the algorithm of Instagram takes novelty of the
content into account. Therefore, it is highly unlikely that the post was shown to consumers later than a couple of days after the initial posting by the SMI. Being visible is in a sense being selected by the Instagram Algorithm, that closely resembles Facebook’s Edgerank. Since the algorithm values Time Decay as one of the factors to select newer content over older, it is unlikely that old content will be shown to followers (Bucher, 2012)
Table 2 - Descriptive Statistics
N Mean SD Min Max Skew. Kurt.
Relevance 262 35.81 28.84 0 96 0.468 -0.96 Product Category Complexity 262 0.69 0.46 0 1 -0.850 -1.287 Popularity 262 79,814.10 117,699.32 1295 715,972 3.54 14.351 Relative Consumer Engagement 262 360.33 265.47 31 1644 2.31 6,99 Methods
The dependent variable relative consumer engagement takes the form of a nonnegative integer and calls for the estimation procedures typical for count data. Such as a Poisson, Negative Binomial, or a zero-inflated model. Because there are no cases within the sample for which Relative Consumer Engagement is 0, we can rule out the use of the zero-inflated model. To check whether the data would follow a Poisson Distribution, a One Sample Kolmogorov-Smirnov test, testing for a Poisson distribution; Z=9.28 p=0.00. Therefore, we can say that our sample does not follow a Poisson distribution for the dependent variable, and t we therefore opt for a negative binomial regression. This was corroborated by a test of overdispersion.
To address our moderation hypotheses (H2A, H2B and H3), our independent variable relevance and the moderator variables, Product Category Complexity for H2A and H2B and Popularity for H3B, needed to be mean centered and were computed to test the moderation.
4. RESULTS
The results shown in table 3 test our theoretical reasoning with respect to the relative consumer engagement on Instagram posts containing product endorsement. The Models 1 to 5 report the estimates from the count based on negative binomial regression. Each Model introduces new variables that are relevant to our aforementioned hypothesis. Model 5 includes all the variables and estimates the coefficients of the full model.
Table 3 - Relative consumer engagement; GLM Estimates of Negative Binomial Models
Variables Model 1 (H1) Relevance Model 2 (H2A) Int 1 Model 3 (H2B) Int 2 Model 4 (H3) Int 3 Model 5 Full Model
Relevance (Mean Centered) 0.004** (0.0012) -0.003 (0.0024) 0.007** (0.0015) 0.002 (0.0013) -0.006* (0.0023) Relevance x High Complexity Product Category (Int 1) 0.010** (0.0028) 0.010** (0.0027) Relevance x Low Complexity Product Category (Int 2) -0.010** (0.0028) -0.010** (0.0027) Relevance x Popularity (Int 3) -8.368e-8** (1.7969e-8) -8.647e-8** (1.7598e-8) Constant 5.880** (0.0383) 5.851** (0.0379) 5.851** (0.0379) 5.825** (0.0377) 5.792** (0.0372) Observations 262 262 262 262 262 Log-Likelihood -1750.317 -1744.193 -1744.193 -1738.262 -1730.821
Likelihood Ratio Chi-Square 9.727** 21.976** 21.976** 33.837** 48.718**
Estimated dispersion parameter
(Negative Binomial)
0.382
(0.0817) 0.366 (0.0305) 0.366 (0.0305) 0.351 (0.0294) 0.333 (0.0280)
Note: Standard errors are in parentheses. *p<0.05; **p<0.01.
H1: The relevance score positively influences relative consumer engagement
Hypothesis 1 advance a positive effect of the relevance score of a social media influencer on the relative consumer engagement. Model 1 in Table 3 includes the variables that
are relevant for testing this hypothesis. The coefficient estimates reported in this model support this hypothesis. This model is statistically significant over a null model (G2: 9.727, p < 0.01).
As expected, the relevance -used to measure the credibility and fit of a SMI and the product category- exerts a positive and statistically significant effect on relative consumer engagement (B:0.004, p < 0.01). An increase of one point in the relevance of the social media influencer leads to an approximate 0.4% increase in the relative consumer engagement on an endorsed product post on Instagram.
H2A: The positively moderating effect of high complexity product categories.
Hypothesis 2A proposes that products classified as high complexity have a positive moderating effect on the positive relationship between the relevance of the social media influencer and the relative consumer engagement. The coefficients reported in Model 2 in Table 3 lend support to this hypothesis (B: 0.010; p < 0.01) and were found to be a statistically significant predictor over a null model (G2: 21.976, p < 0.01). For high complexity product
categories, the effect of relevance of the social media influencer on relative consumer engagement increases by 1%, supporting the idea that to endorse a more complex product, the relevance of a SMI is more important than when endorsing a product within a less complex product category. Interestingly, the estimates from this model suggest that the main effect of relevance on relative consumer engagement is negative (B: -0.003), which would mean that if the relevance of the social media influencer, the estimated relative consumer engagement on the post decreases, but this effect is not statistically significant (p = 0.161).
H2B: The negatively moderating effect of low complexity product categories.
Hypothesis 2B posits that low complexity of the product category has a negative moderating effect on the positive relationship between the relevance of the social media
influencer and the relative consumer engagement. The coefficients that support this hypothesis (B: -0.010, p < 0.01) are reported in Model 3 that can be found in Table 3. This model is statistically significant over a null model (G2: 21.976, p < 0.01). This result shows that the effect
of relevance of the social media influencer on relative consumer engagement decreases by 1% for low complexity product categories, suggesting that to endorse a low complexity product category, relevance is a less important predictor of relative consumer engagement. This means that you do not need as much relevance on the subject matter, to endorse a low complexity product successfully. When endorsing low complexity products, other predictors are likely to become more important, to be successful and obtain a high relative consumer engagement.
H3: The negatively moderating effect of SMI popularity
Lastly, we look at the relationship between relevance and relative consumer engagement. In Hypothesis 3 we state that relevance and relative consumer engagement together are negatively moderated by the popularity of the social media influencer. The coefficient estimate reported in Model 4 lends support to this hypothesis (B: -8.368e-8, p < 0,01). The total model was statistically significant over a null model (G2: 33.837, p < 0.01).
This outcome suggests that if the effect of the relevance of the social media influencer on relative consumer engagement decreases as the popularity of a SMI increases. The effect of this is appears to be very small. Unstandardized coefficients will show an increase in one unit, meaning an increase of only one follower as a unit of analysis for popularity. This does not show a visible interaction effect on relative consumer engagement. The introduction of the interaction effect of relevance and popularity in Model 4 led to loss of statistical significance for the main effect of relevance on relative consumer engagement (B: 0.002, p = 0.257).
Conceptual Model: Testing the full model
Just as in our theoretical reasoning our four hypothesis (H1, H2A, H2B, and H3) work towards one full model that tries to explain the effect of the relevance of the social media influencer on the relative consumer engagement. This full model was tested in Model 5 as shown in Table 3. The estimates reported in the full model (Model 5) mostly confirm the results discussed so far. The model proposed is statistically significant and is better predictor than a null model (G2: 48.718, p < 0.01). Interestingly, it shows that the main effect- the effect of
relevance of the SMI on relative consumer engagement now has a negative coefficient (B: -0.006, p < 0.05): An increase of one on relevance of the SMI decreases the relative consumer engagement by 0.6%. This contradicts Hypothesis 1, but confirms Hypotheses 2A (B: 0.010, p < 0.01), 2B (B: -0.010, p < 0.01), and 3 (B: -8.647e-8, p < 0.01). Figure 2 shows a visual representation of this model with the estimated coefficients from the negative binomial regression analysis.
Figure 2 - Full model with coefficients
*p < 0.05; ** p < 0.01
Product category complexity (High | Low)
Popularity of the social media influncer (Number of followers)
Relevance of the social media influencer (Relevance score)
Relative consumer engagement (Sum of likes and
comments) -0.010**
-0.006* 0.010**
5. DISCUSSION
The aim of this thesis is to provide insight into the consumer behavior surrounding product endorsements on social media via interaction with content posted by SMIs. To address this concept, we focused on the relative consumer engagement and how this is affected by the relevance of the social media influencer in a given context. Results from this study will be discussed hereafter in the text both in relation to past research and as a template for recommendations for future research.
The effect of relevance on relative consumer engagement
H1 states that the relevance of the social media influencer has a positive effect on the relative consumer engagement on a product endorsing post. When analyzing this hypothesis in a model only containing relevance as a predictive variable of relative consumer engagement this hypothesis was confirmed. However, when more variables were added to the models as well as in the result testing the full model, the data provided evidence to reject this hypothesis. Explanation for this contradiction will be addressed below.
The relevance of the SMI measures the credibility and fit of the SMI on a given subject matter, defined by a set of keywords. In our theoretical reasoning we showed that these concepts are named as explanatory factors of endorsement effectiveness following source credibility theory (Ohanian, 1991; Goldsmith et al., 2000; Amos et al., 2008; Till and Busler, 2008). However, the source attractiveness model provides alternative reasoning stating the not the expertise, but physical attractiveness of the endorser is the most important determinant of endorser effectiveness (Joseph, 1982). Amos et al. (2008) name source attractiveness as fourth most important source effect of endorser effectiveness, after negative information about the celebrity, source expertise, and source trustworthiness, but before source credibility. This could be a confounding variable within the scope of this thesis, as data presented here did not include
a variable that tested for the concept of endorser attractiveness. The attractiveness of an endorser is highly prejudiced by the consumers personal preferences and therefore hard to measure in a research design using third-party data. Because most SMIs are less known than traditional celebrities, it is hard to find an unbiased consensus of the endorsers attractiveness. Nonetheless, it is possible that the perceived attractiveness of the 66 different SMIs used in this research has influenced the outcome especially when more possible predictors were entered into the model.
Deviations from expected outcomes may also be consequent of the way consumers process content posted on a social media platform. The content created for Instagram is mostly targeted towards a broad audience, maximizing the potential number of followers that is attracted to the content (Celebrity Intelligence, 2018). For example, the visually pleasing aspect of the content is much more important for content on Instagram than it would be on, for example, a written review. This can trigger a superficial examination of the message, in this case a post on Instagram, in the mind of the consumer (Kelman, 1961). When the message is processed superficially, variables such as aforementioned attractiveness, similarity between the SMI and the consumer, likeability of the SMI, positive image associations, and positive emotional content come into play (Kapitan and Silvera, 2015). These are all associations that the consumer attributes to the source of the information, in this case the social media influencer. These are also associations that trigger affectionate responses, rather than cognitive responses, probably resulting in more likes (Kim and Yang, 2017). If these associations replace the need for relevance on a certain topic to be an effective endorser of a product, this might provide some explanation of the minimized role of SMI relevance as a predictor of consumer engagement.
The moderating effect of complexity of the product category
H2A and H2B hypothesize a moderation effect of the complexity of the product category, with a positive moderation for high complexity product categories and negative moderation for low complexity product categories. This means that for product categories with complex products, the effect size of relevance as a predictor variable of relative consumer engagement would be bigger than for product categories with less complex products. The results shown in Model 2 and 3, summarized in Table 3, support these hypotheses.
This can be attributed to the perceived risk that a consumer feels when purchasing make-up or skincare products which, for this research, were categorized as high complexity products. To get the best possible results out of products in these categories, you need a certain level of skill or knowledge. For example, the incorrect use of sun protection products, can result in severe sunburns and even health risks such as skin cancer (Photobiology - L’Oreal Group, 2018). When searching for product information about these products, consumers are more likely to trust information sources that have a high fit with the product (Till and Busler, 1998). For the high complexity category makeup another explanation can be found in the professional make-up artists or “make-up gurus” on Instagram. These social media influencers create content centered around make-up, probably receiving high relevance scores in the product category make-up. In their posts they not only endorse the make-up product, they also use it to create their signature content. This content, mostly in the form of pictures of their own face (selfies), is used as a record of results from usage of a product (Gannon and Prothero, 2016). Because their Instagram profiles are records of product experience, the make-up guru gains legitimacy as a source of expertise regarding make-up. Instagram is one of the main platforms where so-called “make-up junkies” gather looking for this expertise and simultaneously engaging with endorsements of make-up products.
For low complexity product categories, the opposite is found to be proven by the data (Table 3, Model 3). For low-complexity product categories (haircare and nail polish), our theoretical reasoning is supported by results showing that these product categories contain less performance risk for the consumer (Biswas et al., 2006).
The negative moderating effect of SMI popularity
The relationship between relevance and relative consumer engagement is negatively moderated by the popularity of the social media influencer (Hypothesis 3) was supported by the data in this research. In cases where the SMI had more followers, this negatively moderated the effect of predictor value relevance on relative consumer engagement.
In the chapter of this thesis visiting theoretical foundations for the research the number of followers was named as a cue for source credibility (Araujo et al., 2016; Jin and Phua, 2014) insofar that SMIs with a large number of followers are perceived as more credible by consumers than SMIs with less followers.
Another explanation is provided by a theory called the million-follower fallacy (Cataldi and Aufaure, 2014). It is stated that follower size is not directly related to influence on social media. A lot of practicing marketing managers hold the rule of thumb that the larger the audience of the SMIs the lower their engagement rate is. This corroborates the idea that the more well known a social media influencer becomes the closer they are to celebrity status. This turns followers into fans rather than friends enhancing an asymmetrical relationship (Marwick and boyd, 2011). In this case the celebrity social media influencer is perceived to be less accessible, believable, intimate, and easy to relate to (Schau and Gilly, 2003), which results in less consumer engagement. Figure 3 shows that this is also the case in this sample. The engagement-rate (computed by dividing the sum of likes and comments by the total number of followers of a SMI) declines as the number of followers grows, to start growing again after
about 270,000 followers, showing that the SMIs with the highest number of followers have a growing engagement rate.
Figure 3 - Number of Followers and Engagement Rate
Note: function for the fitted line: y=4.11-1.21e-5*x+2.24e-11*x*x
Other explanations for behavior on Instagram
At the root of this research lies the assumption that consumers recognize content that endorses a product as such- a product endorsement. It is possible that consumers who use Instagram engage with product endorsement not as type of advertisement but simply as content created by people they follow or admire (Banister and Cocker, 2013). Sheldon and Bryant use Uses and Gratifications Theory to suggest an audience’s motive for using Instagram may be driven purely by narcissism. Uses and Gratifications theory poses that psychological characteristics influence gratifications sought, in this context the audience’s motives for using Instagram (Sheldon and Bryant, 2016). They find Surveillance- to see what other people share, Documentation- to share your life with other people, Coolness- to become popular, and
