Effectiveness of Branded Digital Content

(1)

(2)

Effectiveness of Branded Digital Content ½ Master Thesis by Maike Holland 1

University of Groningen

Faculty of Economics and Business

MSc Marketing Management

MASTER THESIS

Effectiveness of Branded Digital Content

Brand Posts on Social Media: Determining the Effects of Social

Media Marketing on Customer Engagement

January 15

th

, 2018

Author Maike Holland (S3294951) M.Holland.1@student.rug.nl Elsternweg 10, 40668 Meerbusch, Germany

1st Supervisor 2nd Supervisor 2nd Supervisor

(3)

(4)

(5)

Effectiveness of Branded Digital Content ½ Master Thesis by Maike Holland

4

1. Introduction

Today social media is the most dominant digital communication tool (Hudson et al. 2016; Chappuis, Gaffey & Parvizi 2011). The availability and the use of social media is continually increasing, and customers are online anytime, anywhere and interacting with all kinds of devices (Hudson et al. 2016; Maecker, Barrot & Becker 2016). Social media is not only used to communicate or share information, but customers are also able to learn by using the platform to interact with brands they might consider for purchase or evaluation (Edelman 2010). According to a recent survey (CMO Survey 2017), 83% of companies include social media in their communication strategies – with an uprising trend. Additionally, Maecker, Barrot and Becker (2016) discovered that customers who interact with the brand on social media are more profitable than those customers who are not interacting with the brand. By giving customers the chance to respond using online platforms, companies adopt the advantages of interactive social media (Sashi 2012). Furthermore, web 2.0 is a trend that marks a new world wide web technology of generational change in web-based communication (Wirtz, Schilke & Ullrich 2010; Cheng, Dale & Liu 2008). Internet evolved from pure information retrieval towards a platform of increasing options of interactivity and communication (Campbell et al. 2011). Consequently, the web of today gives consumers new opportunities to connect and communicate with each other and participating companies on specific platforms. One of many opportunities to use social media is networking which increased to be the most popular tool for online communication (Ross et al. 2009). Companies make use of various networking pages such as Facebook, YouTube, LinkedIn, Xing, Flickr, Couchsurfing and so forth to establish fan pages and foster relationships with customers (Kaplan & Haenlein 2010; Boyd & Ellison, 2007). Creating a brand fan page offers companies and customers a shared platform for interaction. This is displayed through the use of likes and comments from both parties (de Vries, Gensler & Leeflang 2012; SAS HBR 2010; Kaplan & Haenlein 2010; Trusov, Bucklin & Pauwels 2009; McAlexander, Shouten & Koenig 2002; Muñiz & O’Guinn 2001). However, before companies can start engaging with social media, they need to understand customer interactions with brands on these online channels (Schivinski, Christoduolides & Dabrowski 2016). Even though the use of social media is steadily increasing in companies, there is still a lack of understanding to relate online metrics to the company (Hoffman & Novak 2012). This need of comprehension underlines the importance for companies to increase the understanding of customers’ interaction towards digital branded content.

Despite recent papers and studies, the reasons why specific brand’s social media posts have more engagement than others are still not purified. Advertising can be a very successful tool in marketing but the factors of success are hard to determine (Harnett et al. 2016). Previous literature focused on so-called cookbooks to develop a magical recipe of advertising success (Harnett et al. 2016; Stewart & Furse 1984). Results show that there is no strict guideline leading towards the effectiveness of advertising.

(6)

5

relationships, interactions of the type of content and creative aspects leading towards customer engagement will be studied. Academic research on the effectiveness of branded digital content mostly focuses on brand posts (de Vries, Gensler & Leeflang 2012), creative strategies (Ashley & Tuten 2014) and user-generated video content (Berger & Milkman 2012). Although a firm’s use of social media ultimately seeks to successfully create sales (Hudson et al. 2016, Maecker, Barrot & Becker 2016), the relation between the content itself and the customer engagement on video advertisement is still inconclusive and a matter of research. Therefore, this study will give new insights into the success factors of branded digital content and the relations of content towards customer engagement.

De Vries, Genlser and Leefland (2012) investigated the popularity of branded fan-page posts and took during their study the following six types of content into consideration for advertising’s increasing factors: vividness, interactivity, informational content, entertaining content, position and valence of comments. These characteristics are assessed by the number of likes and comments. This study is built upon the study by de Vries, Gensler and Leeflang (2012) and hence taken into account while formulating the study’s concept. Informational and entertaining content in advertising is mentioned in prior literature (de Vries, Gensler and Leeflang 2012, Dolan et al. 2016). Informational content is described by Luo (2002) as the most vigorous motivator to interact while entertaining content is described as more engaging than other types of content (Cvijikj & Michahelles 2013). Since this study builds upon the study by de Vries, Gensler and Leeflang (2012), the prior described variables of informational and entertaining content are included into this study with the goal to further understand which is the most engaging content. Moreover, some literature mentioned celebrity endorsement, describing its effectiveness towards a brand (Spry, Pappu, Cornwell 2011; Dwivedi, Johnson & McDonald 2015). Consequently, this variable is included in the analysis of relations towards customer engagement. Furthermore, the main difference of the study by de Vries, Gensler and Leeflang (2012) is the platform that was used: This study is going to use YouTube as a representative for social media platforms, where the focus is purely on video-content, while the previous study used Facebook. Thus, the variables of music and voice-over apply to video content. Previous literature describes the appearance of music as an influencer of engagement (Galan 2009) while the use of voice-over by a character appeared as a negative influencer for sales (Harnett et al. 2016). The appearance of music in videos and the use of voice in the form of character- or narrator speaking (Chatman 1999) are a common tool in advertorial videos and consequently need further investigation. Correspondingly, the variables music and voice-over are added, which result in the conceptual framework (figure 1).

The main contribution of this study is consequently to determine the effects of social media marketing on customer engagement. Therefore, the following research question is applied:

(7)

Effectiveness of Branded Digital Content ½ Master Thesis by Maike Holland 6 As previously mentioned, YouTube will provide the social media platform for the data collection. Established in 2005, YouTube is after Google the second most visited homepage worldwide (Alexa 2017) and offers the possibility to build a company-owned channel to communicate with customers, fans and users. Additionally, this platform is by definition of Kaplan and Haenlein (2010) a social networking platform since the platform offers customer engagement in any form such as sharing, creating new online content and commenting. Consequently, results of this study are transferable to similar platforms offering to post video advertising.

To answer the research question, the paper is structured as follows: first, the conceptual model with its variables is introduced. After developing the framework, hypotheses are formulated and linked to previous literature. Chapter 3 starts with an indication of the research design, followed by a short explanation of data collection and ends with a plan of analysis for collected data. Chapter 4 presents the results of the data analyses. Finally, chapter 5 concludes this study with a discussion, managerial implications and limitations while providing possible suggestions for future research.

2. Theoretical Framework

2.1 Conceptual Model

(8)

7

Therefore, celebrity endorsement, music and the use of voice are expected to interact with the type of content. Additionally, the use of voice is predicted to interact with music concerning customer engagement.

2.2 Theoretical Background and Hypotheses

2.2.1 Customer Engagement

Customer engagement on social media strongly emerged in the past few years (Hussein & Hassan 2017) while companies continuously try to engage with customers in many different ways (Pansari & Kumar 2017). According to marketing literature, engagement is described as a customer’s activity towards a company and is consequently called customer engagement (Pansari & Kumar 2017; Vivek, Beatty & Morgan 2012; Brodie et al. 2011; Kumar et al. 2010). Hence, customer engagement is a metric scale for a firm's marketing activity outcome (Brodie et al. 2011; Kumar et al. 2010; Vivek, Beatty & Morgan 2012).

For example, Kumar (2013) describes the activities of a customer as the demonstration of engagement levels: activities are for instance a purchase, customer recommendations and conversations about the brand or a product. Hudson et al. (2016) investigate the relationship of social media towards consumer engagement. They state that social media is positively related to higher consumer-brand relationships and word-of-mouth communication (ibid). Van Doorn et al. (2010) define customer engagement as a common marketing objective and confirm that customer engagement focuses on brand or firms. Furthermore, previous literature describes the positive influence of social media activities that result in more engaged consumers and higher willingness to participate and share (Ashley & Tuten 2014). Through the use of higher customer engagement, the traffic towards the brand or firm might increase. Moreover, customer engagement on social media can appear as the number of views, likes, dislikes and comments (Reese et al. 2016; de Vries, Gensler & Leeflang 2012). Although the number of dislikes is a negative metric, it demonstrates increasing traffic-numbers.

2.2.2 Type of Content

(9)

8

2012). Further literature distinguishes two primary creative strategies in advertising: firstly, emotional and transformational and secondly, functional and informational (Aaker & Norris 1982). Of note is that functional messages lead to high consumer-involvement towards a brand, whereby transformational messages result in low-involvement (Areni 2003), which might extend in the form of informational content leading to high customer engagement. Moreover, information-based strategies build a healthy relationship with high-involvement customers (McMillan, Hwang & Lee 2003). This relationship might influence the effectiveness of advertising on customer engagement. Therefore, the differentiation of content while using informational as one type of content is decided on. The descriptive study by Golan and Zaidner (2008) similarly describes the basic approach of indicating advertisings creative strategies as informational or transformational. Their statistics show that both forms of advertising strategies are used with the same frequency. What distinguishes this study from previous research is that it will analyse the usage of informational messages without pairing it with transformational messages.

Moreover, social network advertising that focuses on informational content tends to achieve more positive attitudes towards the advertising itself, than less-informational content (Taylor, Lewin & Strutton 2011). Hence, informational content in video advertising might confirm the positive attitudinal change in the form of high customer engagement. The success of informational content is also used to explain reasons why people consume brand-related content (Muntinga, Moorman & Smit 2011). Accordingly, customers engage towards informational content positively (ibid). Furthermore, the pursuit of information can explain the reason for customers consuming brand-related content. Consequently, informational content might lead to higher customer engagement numbers than non-informational content. In line with this reasoning, the following hypothesis is introduced:

H1: Informational content in a video advertising influences customer engagement positively.

Entertaining Content

Aforementioned, entertaining content can be seen as another primary strategy to increase customer engagement (Aaker & Norris 1982). Entertainment appears in numerous industries, for example in publishing, music, sports, broadcasting, gaming, event and the tourism industry (Foutz 2017). Its impact in the marketing world is increasingly growing and influences consumers more and more (ibid). Videos including entertaining content do not show product features in details, but might still be related to the product itself (Rahman et al. 2016). Entertainment in previous literature is mentioned as content that does not relate to the product, service or brand (Rahman et al. 2016; de Vries, Gensler & Leeflang 2012). It is a form of media gratification that is offered to escape problems or the daily routine and is supposed to work as an emotional relief and can serve as a relaxation tool (Muntinga, Moorman & Smit 2011).

(10)

9

content while focusing on entertainment, which is not based on information about the company, brand, product and so forth. Entertainment in the form of social media content is then supposed to increase customer engagement whilst delivering consumers enjoyment (Muntinga, Moorman & Smit 2011). Previous literature states that the entertainment-approach leads users to consume, create, and contribute to branded content on online platforms. Besides, advertisements using entertaining content are perceived as fun, exciting and positively influence the attitude toward advertisements (Taylor, Lewin & Strutton 2011). Next to that, entertaining content increases the desire to return to a website (Raney et al. 2003). Hence, this might be transferrable to customer engagement, for instance in the form of multiple views. Literature describes the significance of entertaining content on increasing numbers of likes, shares and clicks on Facebook posts (Cvijikj & Michahelles 2013). The study finishes with the result that entertaining content is the most influential form of branded content (ibid). Furthermore, entertaining content is described as more engaging than other types of content (ibid). Compared to previosly described informational content, entertaining content may influence consumer engagement more positively than informational content. Despite the result of de Vries, Gensler and Leeflang (2012), stating that entertaining content has a negative effect on the number of likes, literature confirms the positive effect of entertainment on customers. Being the most influential form of branded content (Cvijikj & Michahelles 2013), the outcome of this form of content might be more positive than the other types of content, which is in this study design informational content. Consequently, the second hypothesis is following: H2: Entertaining content in a video advertising influences customer engagement more positively than informational content. 2.2.3 Celebrity Endorsement

Celebrity Endorsement in the marketing context is not a new phenomenon, being used since the late 19th century (Kaikati 1987). ‘Celebrity’ typically refers to a famous or infamous person: this person is known to an audience, whereas the audience is unknown to the celebrity (Gamson 2001). Furthermore, celebrities are likely to represent a symbolic or aspirational reference for customers (Escalas 2004). Linkage of celebrities to brands intends to endorse brands with the desirable associations of a celebrity (Keller 2013; Till 1998). Originally not connected, the celebrity and brand are bonded through an endorsement process (Spry, Pappu & Cornwell 2011; Till 1998). The use of celebrities increases, for instance, brand recognition, brand recall and might even influence a customer’s purchase behaviour, such as increasing purchase intentions (Spry, Pappu & Cornwell 2011). Celebrity endorsement is supposed to influence the effectiveness of advertisement positively (ibid), which underlines the importance of celebrity endorsement in this study’s design.

(11)

10

Cornwell 2011). Further research demonstrates that celebrity endorsers are classified as an influencing instrument for consumers’ brand connection (Dwivedi, Johnson & McDonald 2015). Some articles mention the endorser's credibility as a source for trustworthiness, leading to a customer's opinion on advertising, based on the celebrity itself (Spry, Pappu & Cornwell 2011; Goldsmith Lafferty, & Newell 2000; Ohanian 1990). Besides, not only high credibility endorsers influence a brand positively - even celebrities with somewhat low influence prove the ability to build a brand (Spry, Pappu & Cornwell 2011). Consequently, since the appearance of a celebrity has a positive effect, no further specification of the celebrity him-, herself or multiple celebrities is needed.

Additionally, literature mentioned celebrity versus non-celebrity endorsement as a tool to increase purchase intentions (Mehta 1994). According to previous studies in this field, differences in advertising attractiveness between endorser and celebrity endorser are visible (Nataraajan & Chawla 1997). Hence, not only endorsement but celebrity endorsement itself might influence the numbers of customer engagement. Nevertheless, empirical evidence positions that celebrity endorsement generates a higher recall for endorsed brands and the related advert (Atkin & Block 1983) and leads to a more favourable attitude towards the brand (Till, Stanley & Priluck 2008). Resulting in a high recall, increasing engagement and the positive positioning of celebrity endorsement in previously listed literature, a positive effect of celebrity endorsement on customer engagement is expected. To achieve results concerning the appearance of celebrity endorsement in advertising and possibly confirm outcomes of previous studies, celebrity endorsement’s effectiveness is analysed in the following third hypothesis:

H3: Celebrity endorsement in a video advertising influences customer engagement positively.

2.2.4 Music

(12)

Effectiveness of Branded Digital Content ½ Master Thesis by Maike Holland 11 digital content. The frequent mention of music in comments during videos leads to a further investigation of these effects in the world of advertising (Campbell et al. 2011). The next step is to differentiate the type of music to investigate these effects. Past research showed that music formats that involve users are more effective than less involving music formats (Sullivan 1990), such as the forms of mood music compared to background music. Additionally, limitations of the described study state to gain more information about different forms of music or radio programmes (ibid). Hence, the division involving or less-involving music seems to be applicable to this study. Other literature tries to define a music-fit, based on style, tempo, rhythm and so forth to match the product message (Zander 2006). This leads to the status quo which describes the ongoing search for the perfect fit of music in advertising. The likeability and style of music in advertising as used in previous literature of music-fit (Zander 2006; Galan 2009) is highly opinionated and therefore not included in this study. Other specifications such as style, rhythm and so forth need the aid of music-professionals. While music influences advertising for customer’s first impressions (Zander 2006), different forms of customer engagement towards specific music might result. Zander (2006) uses in a study two different types of music, dividing it up into one mainstream-amusing song and one piano-ballade, a fairly classical pop music song. While the results only describe the significant transfer of music towards a product in any form (amusing and ballade) (ibid), an example of significant results to describe different types of music is discovered. Harnett et al. (2016) differentiate music into mood music compared to less quiet music such as elevator music and no music. Especially mood music, which invites customers and users to create a particular mood, is perceived as less efficient on sales (ibid). Furthermore, background music is described as important and effective to enhance advertisement’s message recall (Park, Park & Jeon 2014). This can be applied to higher customer engagement numbers as a result of background music since the presence of background music alone enhances advertising recalls (Hoyer, Srivastava & Jacoby 1984). After considering a variety of papers dividing music in different types, the differentiation of mood and background music is made. With the following hypothesis, the effects of music will be analysed, based on the positive effect of background music (Park, Park & Jeon 2014; Hoyer, Srivastava & Jacoby 1984) and the negative effect of mood music on sales (Harnett et al. 2016). Taking previous statements into account, hypothesis 4 tests whether background music increases customer engagement more successfully than no music and mood music. H4: Background Music in a video advertising positively influences customer engagement compared to mood music and absence of music. 2.2.5 Use of Voice

(13)

12

A previous article describes the appearance of web-text functions in advertising, divided into ideational, interpersonal and textual (Tan 2010). Ideational describes the way experience is presented and interpersonal describes the person presenting, for instance in the form of a speaker, writer and his or her relationship with the audience. The mentioning of speaker(s) in advertising is viable in this context and is not a new phenomenon. Besides, Chatman (1999) introduces the difference of narrators in movies into voice-narrated and character-voice in the figure of the first person without linking it to marketing approaches. Another reason to do so is the similarity between different forms of moving pictures, such as movies and videos. Hence, both narrator- and character-voice are needed to formulate the fifth hypothesis.

Furthermore, Harnett et al. (2016) investigate the effectiveness of creative aspects, using different types of voices, namely voice-over with characters and no voice-over. They conclude their study with the result that the voice of a character speaking, negatively influences sales. Hence, the variable voice-over will be included in this research and following hypothesis according to Harnett et al. (2016) and the results of negative influence of sales by character-voices will be added. Furthermore, due to the negative influence of character-voice on sales, a similar outcome on customer engagement is expected. Consequently, this study will investigate whether the effect on sales corroborates with customer engagement. H5a: Character Speaking in a video advertising negatively influences customer engagement. On the one hand, the use of voice is described as a negative tool for sales (Harnett et al. 2016). On the other hand, music is described as one of the most critical elements in advertising for moving media (Galan 2009). Since both variables are described as very close to each other (Harnett et al. 2016), the relation and difference towards customer engagement is highly important for this study’s results (see also chapter 2.2.6 for indirect effects). Hence, the following hypothesis is added to analyse whether music has a stronger effect on customer engagement than the appearance of voice-over. Based on previous hypotheses and literature, the effect of music is expected to be stronger than the effect of voice-over, since music is described as most critical for advertising (Galan 2009). To answer the question of the more engaging creative aspect, following hypothesis 5b is introduced. H5b: The effect of music on customer engagement is more positive than the effect of use of voice in video advertising. 2.2.6 Interaction Roles of Celebrity Endorsement, Music and Use of Voice

Advertising’s creativity contributes to its effectiveness and attracts interest in

(14)

13

music and the use of voice, are expected to enhance or mitigate the type of content on customer engagement.

The first moderation role to introduce into the study is the interaction of celebrity endorsement and the type of content. Celebrity endorsement on its own is described as a positive influencer of customer engagement (see 2.2.3). Conferring to the previously listed direct effects of celebrity endorsement, the appearance of a celebrity possibly results in increasing customer engagement (Till, Stanley & Priluck 2008; Atkin & Block 1983). The next step is to combine this effect with the type of content and test whether the creative aspect, here in the form of celebrity endorsement, influences customer engagement. Since videos are posted on a brand’s YouTube channel, celebrity endorsement as one creative approach might influence engagement indirectly since customers might not immediately know that the video contains celebrity endorsers. Furthermore, people’s attention is described as automatically concentrated (Lang 2000). This means that people tend to copy a preferable treatment to a stimulus, such as celebrity endorsement, that is related to their own goals or preferences (ibid). Moreover, previous literature studied the customer and the moderating roles of attention and interests in relation to celebrity endorsement - without leading to significant results about cognitive moderators (Knoll & Matthes 2016). Still, the approach appears transferrable for celebrity endorsement and the preference of type of content to gain further insights about the appearance of celebrities in advertisings. Hence, the indirect effect of customer engagement on the type of content might give an extended understanding of digital branded content’s success. Since a positive and direct effect is expected, the indirect effect is also expected to be positive. Therefore, the relation is tested with the following hypothesis: H6: Celebrity Endorsement in a video advertising enhances the effect of type of content on customer engagement.

(15)

14

Furthermore, literature indicates with the results that music might have an impact on customer’s moods, which then leads to purchase intentions (Alpert & Alpert 1990). Other literature describes the change of purchase behaviour due to music compared to no music (Guéguen & Jacob 2010). The authors did not find differences within the type of music (ibid). Consequently, this moderating role will occur in its appearance of music with no differentiation of the type of music. Since previous literature does not consider music as a moderating role to increase customer engagement but states different effects of it, the investigation of music as a creative appeal occurs as mandatory to gain results about branded content’s effectiveness. The interaction of type of content and music is expected to increase customer engagement and tested with the following hypothesis: H7: Music in a video advertising enhances the effect of the type of content on customer engagement.

Besides, the third creative aspect use-of-voice is supplementarily estimated to influence the relation of the type of content on customer engagement. As Harnett et al. (2016) state, a negative effect of voice-overs on sales is expected. Since customer engagement and the voice of a character speaking are relatively close, this variable will be used in its general version, as use of voice. As previously described, this effect might occur similarly on customer engagement and result in less engaging numbers than videos without voice-overs. To gain as much information about the appearance of voice-over, the interaction with the type of music is particularly interesting to investigate whether the creative aspects influence the relation of the type of content on customer engagement. However, Tan (2010) offers descriptively different web-text structures and voice-overs. Stating that both textual content and voice interact to influence customer behaviour (Tan 2010), this interaction can possibly occur in the form of music and type of content in videos. In comparison to the two previously listed creative aspects, namely celebrity endorsement and music, the effect of voice-over might occur as a negative influencer in the interaction while the direct effect is described as a decreasing tool (Harnett et al. 2016). Consequently, the next hypothesis is introduced:

H8: Use-of-Voice in a video advertising video mitigates

the effect of type of content on customer engagement.

(16)

15

expected. The relation of both of these on the outcome of customer engagement is not clearly researched, but since literature states a relation (Martín-Santana, Reinares-Lara & Muela-Molina 2015; Alexomanolaki, Loveday & Kennett 2007; Zander 2006), it is mandatory to further investigate. Additionally, Harnett et al. (2016) mention the reduced opportunity for music while using voice-over, underlining the relation between both aspects and describing the need for further understanding. Consequently, both appeals are supposed to be somehow related to each other. Based on previous literature, a possible interaction within both variables is expected, leading to the following hypothesis:

H9: Music in a video advertising enhances the effect of use-of-voice on customer engagement.

3. Methodology

3.1 Data and Sample

Information about brand posts was collected on the social media platform YouTube (www.YouTube.com) between 1st and 24th of November 2017 (N=149). Data was manually captured using two volunteering assistants and the author of this thesis. Following Ashley and Tuten (2014) who state that top brand videos are more useful to gain data, four top brands of the sports category were decided on beforehand: Nike, Adidas, Under Armour and Reebok. Nike and Adidas showed appearance in the Forbes top 100 brand ranking worldwide (Nike #16, Adidas #75, Forbes Media 2017), while all four brands were listed in the top 10 sports brands worldwide in a ranking of 2010 (Forbes 2010). The focus is on previously listed sports brands to increase comparability of the videos without expanding the conceptual model nor adding further variables about content-categories. Following table 1 sums up primary information about the brand and their YouTube channels (YouTube Nike 2017; YouTube Adidas 2017; YouTube Under Armour 2017; YouTube Reebok 2017):

Tab. 1: YouTube Brand Channel Information

Nike Adidas Under Armour Reebok

YouTube Joining 07 March 2005 29 October 2005 03 March 2006 23 December 2005

Number of Videos 276 93 504 1,482

Number of Subscribers 692,258 486,738 156,033 111,189 Number of Views 132,852,455 122,578,989 86,985,292 81,556,471

Additionally, it is important to mention how the platform YouTube and its engagement-process works. For a customer to be able to engage with the video in the form of likes, dislikes or comments, they must first sign in. In comparison to that, only watching a video does not require any form of previous engagement with the platform.

(17)

Effectiveness of Branded Digital Content ½ Master Thesis by Maike Holland 16 Milkman 2012). This gives users time to react to brand posts within the first weeks since most reactions of videos are collected in the first three to four weeks (Cheng, Dale & Liu 2008). Therefore, the customer engagement is pictured clearer than posts that are posted days or hours before the data collection. The next step was to watch the video and code according to the following operationalisations (see table 2). Watching videos multiple times to determine the correct coding was possible. For the data collection Excel was used, which includes the insertion of formats to eliminate double selected videos, missing values and outliers before transferring cleaned data into SPSS-23. Since this study relied on human coders to classify the extent of advertising videos, a random subsample for coding (decisions = 147, N = 7) was selected. Each video was coded separately and according to the coding scheme (see 3.2 Data Measurements). The author of this study introduced the coders to the topic and the coding. Afterwards, the inter-rater reliability was tested. The results of the Kippendorff’s Alpha show high values on all dimensions (a = .979), indicating that the content tends to evoke a similar coding across all coders.

3.2 Data Measurements

The measurement scales of variables are based on previously used scales (see table 2). For informational and entertaining content, the coding scheme of de Vries, Gensler and Leeflang (2012) is adapted. During a pre-analysis of YouTube videos by sports brands, the use of sports teams or many celebrities in one video is perceived as a common. Consequently, the scale of celebrity endorsement using more than one celebrity is added. The use of voice’s, also called voice-over’s, content is not conducted, since the first two variables informational and entertaining already focus on the video’s content.

Tab. 2: Scale Items

Variable Item Operationalisation Source

(18)

Effectiveness of Branded Digital Content ½ Master Thesis by Maike Holland 17 Music No Music (0/1) Background Music (0/1) Mood Music (0/1) Does the video use music? Is the music playing passively in the background or is it in the form of more obvious ‘mood music’? ‘Mood music’ is defined as actively forcing to touch emotions, whereas background music is defined as a subtler form. Harnett et a. (2016) Use –of Voice Character Speaking Narrator Speaking No (0) Yes (1) Character Voice (0/1) Narrator Voice (0/1) Is someone talking in the video? If yes, is the person who is talking a character in the video or a narrator? ‘Someone talking’ describes the use of voice, not just short conversations. Character Speaking is defined as a person in the video who is talking, as opposed to a narrator speaking, who is, for example, a third person. Influenced by: Harnett et a. (2016); Chatman (1999); Tan (2010) Note: (0/1): 0 = does not appear; 1 = does appear

The number of views on YouTube is counted when the video was watched. The so called ‘click-rate’ describes only the click on a video but does not confirm that a user watched it. This ‘click-rate’ is not collected on YouTube nor the study, since it would not reflect customer engagement (YouTube Help 2017). Purely a click on YouTube does not confirm any form of commitment and cannot be used for counting engagement numbers. This also describes, that the previously listed variables, for example, type of content did not commit the user to the video since the click does not support that the user actually watched the video. Consequently, every counted view on YouTube means that a user watched every single second of a video (YouTube Help 2017).

Control Variables

The number of channel subscribers influences the scope of the advertising and consequently might lead to differences in customer engagement across brand posts. Therefore, the number of channel subscribers is controlled for. Furthermore, literature states that message length conceivably influences study outcomes by the number of views (de Vries, Gensler & Leeflang 2012; Baltas 2003; Robinson, Wysocka & Hand 2007). Hence, the control variable of video length in seconds is added. To generate high comparability, an approach by Chang, Dale and Liu (2008) is adopted: during their study about statistics on the social network of YouTube videos, the authors define the current lifetime as the day of upload until the day of retrieved data. This approach is also used by Garg et al. (2015) to define video-viewer interaction parameters. Therefore, the video lifetime as a third control variable is included (day of video upload – day of data collection = video lifetime in days).

(19)

Effectiveness of Branded Digital Content ½ Master Thesis by Maike Holland 18

3.3 Method of Analyses

To test the hypotheses, multiple regression and moderation in SPSS-23 was conducted. To set the models, all control variables needed testing whether they influence customer engagement significantly or not. Therefore, a regression with all control variables (number of subscribers, video Length, video lifetime) was performed on all four dependent variables (views, likes, dislikes, comments). Accordingly, number of views is controlled by subscribers (p=.022***; r2=0.054; B=1.842; standardised B=.196) and video lifetime (p=.024**; r2 =.054; B=-5,236.077; standardised B=-.193). Furthermore, the control variable number of subscribers showed significance for likes (p=.002**_{; r}2_{=.088; B=.003; standardised B=.262), dislikes} (p=.048*; r2=.034; B=0.001; standardised B=.171) and comments (p=.041*; r2=.042; B=0.000;

standardised B=.176).

The table 3 presents the first 12 models out of the total of 24 to test previously described hypotheses. All models are split up into four different blocks, each block representing one dependent variable.

Tab. 3: Regression and Moderation Models for Dependent Variables Views and Likes

Dependent Variable: Views

Model 1 V = b0 + b1EC + b2CE + b9M + b10VO + b7CS + b8CL + e1

Model 2 V = b0 + b1EC + b2CE + b3BM + b4MM + b5VOC + b6VON + b7CS + b8CL + e1

Model 3 V = b0 + b1EC + b2CE + b9M + b10VO + b7CS + b8CL + b11EC*CE + e1

Model 4 V = b0 + b1EC + b2CE + b9M + b10VO + b7CS + b8CL + b12EC*M + e1

Model 5 V = b0 + b1EC + b2CE + b9M + b10VO + b7CS + b8CL + b13EC*VO + e1

Model 6 V = b0 + b1EC + b2CE + b9M + b10VO + b7CS + b8CL + b14VO*M + e1

Dependent Variable: Likes

Model 7 L = b0 + b1EC + b2CE + b9M + b10VO + b7CS + e1

Model 8 L = b0 + b1EC + b2CE + b3BM + b4MM + b5VOC + b6VON + b7CS + e1

Model 9 L = b0 + b1EC + b2CE + b9M + b10VO + b7CS + b11EC*CE + e1

Model 10 L = b0 + b1EC + b2CE + b9M + b10VO + b7CS + b12EC*M + e1

Model 11 L = b0 + b1EC + b2CE + b9M + b10VO + b7CS + b13EC*VO + e1

Model 12 L = b0 + b1EC + b2CE + b9M + b10VO + b7CS + b14VO*M + e1 Where: V = Views L = Likes CS = Control Variable Subscribers CL = Control Variable Video Lifetime e = Error Term EC = Entertaining Content in % CE = Celebrity Endorsement BM = Dummy variable Background Music M = Dummy Music Appearance MM = Dummy Mood Music VO = Dummy Voice-Over Appearance VOC = Dummy Voice-Over Character VON = Dummy Voice-Over Narrator

(20)

19

control-variables are the same for the variables of likes, dislikes and comments, the only difference between the models for likes, dislikes and comments is the dependent variable itself. Therefore, following table 3 presents models 1 to 12 to describe regression and moderation models, whereby models 7 to 12 represent the same structure as models 13 to 24 (see Appendix A: Models 13-24). Hypotheses 1 to 5 test direct effects, represented in model 1, 2, 7, 8, 13, 14, 19, 20. Further testing needs the use of moderation roles, namely of celebrity endorsement, music and voice-over and possible moderating effects on the type of content towards customer engagement. The supplementary models test whether creative aspects influence customer engagement indirectly. To test these indirect effects for hypotheses 6 up to 9, the following models represent the testing: 3-6, 9-12, 15-18, 21-24.

All analyses are marked as significant with p-values lower than .05, the marginal significance is marked to gain information about possible trends (*_{p-value <.1). Significant} results are for a better visualisation coloured in green. To achieve reliable results, standardisation while analysing interaction effects is mandatory. Furthermore, the Variance Inflation Factor (VIF) was tested during all analyses. VIF scores below 10 are described as harmful regarding multicollinearity; higher values might need the elimination of variables if possible (O’Brien 2007).

4. Results

4.1 Descriptives and Correlations

The dataset encompasses information of 149 brand posts (N=149) by four internationally leading brands, posted within 365 days prior October 1st, 2017. The following Table 4 indicates descriptive statistics.

Tab. 4: Descriptive Statistics

Variable Mean SD Min Max

Video Lifetime in Days 188.80 84.69 48 365 Video Length in Seconds 66.79 60.58 7 439 Number of Subscribers 321,954.39 245,583.82 692,258 111,189 Number of Views 646,667.96 2,309,509.77 410 19,966,260 Number of Likes 906.17 2,499.31 5 24,703 Number of Dislikes 171.05 949.94 0 11,009 Number of Comments 76.71 458.98 0 5,530 Informational Content in % 39.08 43.07 0 100 Entertaining Content in % 60.94 43.05 0 100 N = 149

(21)

20

days (SD = 84.69). Table 4 sums up all means, SDs, minimum (min) and maximum (max) values.

Besides, 49% of all videos (N=149) show entertaining content (39% informational content; 17% both) and 50% use background music (42% mood music, 8% no music). The high amount of entertaining content is reasonable since entertainment serves as a rapidly growing platform for advertisers to communicate brand messages (Foutz 2017). In 56% a character is speaking, while only in 31% a narrator is speaking (13% no voice-over). The distributions of creative aspects are presented in figure 2. Fig. 2: Distribution Message Appeal, Music, Voice-Over

Furthermore, correlations within the independent variables are tested. As earlier discussed, there are two types of models: One version uses the general presence of music and voice-over, whereby the supplementary version uses the specification of both variables. Following table 5 demonstrates the general appearance of the creative aspects voice-over and music, while no significant correlation for both of them is measured. Tab. 5: Pearson Correlation Table of Music and Voice-Over Music Voice-Over Music 1 Voice Over .074 1

N = 149; Note: ***_{p-value < .01;}**_{p-value < .05;}* _p-value <.1

The specification of music as background and mood music correlates significantly with the specifications of voice-over, namely in the form of character and narrator, as following table 6 describes. The correlation of voice-over appearance and mood music correlates negatively (r=-.850; p=.000), which is expected since voices and music were previously described as opponent factors (Martín-Santana, Reinares-Lara, Muela-Molina 2015). As expected, background music is significantly correlated with both forms of voice-over (r=.399;

(22)

Effectiveness of Branded Digital Content ½ Master Thesis by Maike Holland 21 Tab. 6: Pearson Correlation Table of Music and Voice-Over Specifications Background Music Mood Music Voice-Over Character Voice-Over Narrator Background Music 1 Mood Music -.850*** 1 Voice-Over Character .276*** -.207** 1 Voice-Over Narrator .138* -.159 * -.429*** 1 N = 149; Note: *** p-value < .01; ** p-value < .05; * p-value <.1

Also, the correlations between the dependent variables are tested. Results show high correlations between dislikes and comments (r=.973; p=.000). Furthermore, likes and dislikes (r=.844; p=.000), as well as likes and comments (r=.868; p=.000) are highly and significantly correlated. Tab. 7: Pearson Correlation Table of Dependent Variables Views

Likes Dislikes Comments

Likes .355*** 1

Dislikes .410*** .844*** 1

Comments .243*** .868*** .973*** 1 N = 149; Note: *** p-value < .01; ** p-value < .05; * p-value <.1

Correlation-results presented in table 7 possibly explain future analyses, especially the high correlations of likes, dislikes and comments. Even though the number of views is positively and significantly correlated to all other variables, namely likes, dislikes and comments, the correlation is not as high as the others. One possible explanation for that is the fact that YouTube requests a sign-in to like, dislike or comment on content. To react towards a video, a log-in is requested. Hence, the correlation between these variables is reasonably higher.

4.2 Data Analyses

(23)

22

In the next step, the dependent variable customer engagement is measured with the four metrics for customer engagement: views, likes, dislikes and comments. The following sub-chapters sum up the results per metric. In the last section of this chapter, a summary of hypothesis-testing is presented. 4.2.1 Customer Engagement: Views Table 8 shows (standardised) coefficients for the dependent variable views. Celebrity endorsement is throughout the regression models highly significant and positively influencing the number of views. As presented in model 3, celebrity endorsement does not only influence customer engagement directly (b=.298; p=.003) but also indirectly (b=.259; p=.002). Therefore, H3 and H6 for the metric views are supported. Figure 3 presents the moderation effect, while the appearance of celebrity endorsers in interaction with entertaining content increases views, the reverse of the results of informational content and celebrity endorsement (b=-.259; p=.002). Hence, the type of content itself does not influence customer engagement, but it does in combination with celebrities. In addition, the interaction includes a positive and significant r2-change (R2-change=.055; p=.002).

Fig. 3: Interaction Results for Type of Content and Celebrity Endorsement on Views

Furthermore, H5 with a marginally significant negative effect of a character’s voice is supported (b=-.163; p=.093). The interaction of entertaining content and voice-over appears significantly (b=-.168; p=.041), thus H8 is held. The appearance of any form of voice-over in interaction with entertaining content shows higher standardised coefficients. Consequently, the interaction of both shows a more negative effect. The reverse interaction including informational content and voice-over does lead to higher customer engagement (b=.168;

p=.041). In addition to previously mentioned hypotheses, the control variables show constant

(24)

23

Tab. 8: (Standardised) Coefficients for Dependent Variable Views

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

(Constant) 1,122,039.91 -936,198.61 536,063.335** 1,141,492.17** 1,302,233.64** 1,046,438.07* Entertaining Content H1H2 .075 .075 .184* .051 .123 .075 Celebrity Endorsement H3 .292*** .298*** .276*** .292*** .304*** .296*** Music Background H4 .043 Music Mood H4 .119 Music Appearance H5b .030 .024 .024 .042 .041 Voice-Over Character H5a -.163* Voice-Over Narrator (H5a) .010 Voice-Over Appearance H5b -.149* -.157** -.150* -.166** -.082 Control Variables Subscribers .154* .145* .166** .155* .134* .151* Video Lifetime -.176** -.155* -.143* -.175* -.194** -.164* Interaction Effects Entertaining Content x Celebrity Endorsement H6 .259*** Entertaining Content x Music H7 .027 Entertaining Content x Voice-Over H8 -.168** Music x Voice-Over H9 -.088 R2 (adj. R2) .169 (.134) .184 (.137) .224 (.186) .169 (.128) .194 (.154) .172 (.131) R2-change .055*** 0 .024** .002 F-Value 4.822*** 3.948*** 5.820*** 4.108*** 4.837*** 4.174***

(25)

24

Subscribers show a positive effect towards customer engagement (b=.166; p=.037), whereas the video lifetime decreases the number of views (b=-.143; p=.075). This might be surprising since it is negatively related to the number of views, but consistent with the findings by Cheng, Dale and Liu (2008) stating that customer engagement numbers increase mostly during the first three to four weeks. Due to missing significance, further hypotheses cannot be confirmed. 4.2.2 Customer Engagement: Likes Table 9 presents the results for the dependent variables of likes and underlines the significance of celebrity endorsement on customer engagement within all six measured models, leading to a second confirmation of H3 that customer engagement increases celebrity endorsement (b=.208; p=.033).

Tab. 9: (Standardised) Coefficients for Dependent Variable Likes

(Constant) -73.314 -1687.037* _-369.493 _-14.142 _-73.314 _-80.103 Main Variables Entertaining Content H1, H2 .008 .015 .074 -.044 -.003 .008 Celebrity Endorsement H3 .208** .222** .198** -.207** .205** .209** Music Background H4 .109 Music Mood H4 .052 Music Appearance H5b .031 .026 .017 .028 .034 Voice-Over Character H5a -.060 Voice-Over Narrator (H5a) .094 Voice-Over Appearance H5b .000 .000 .000 .005 .018 Control Variable Subscribers .262*** .245*** .275*** .264*** .268*** .262*** Interaction Effects Entertaining Content x Celebrity Endorsement H6 .158* Entertaining Content x Music H7 .058 Entertaining Content x Voice-Over H8 .038 Music x Voice-Over H9 -.022 R2 (adj. R2) .126 (.095) .145 (.103) .147 (.111) .126 (.089) .127 (.090) .126 (.089) R2-change .021* .001 .001 0 F-Value 4.112*** _3.421*** _4.066*** _3.423*** _3.443*** _3.407***

Notes: ***_{p-value < .01;}**_{p-value < .05;}*_{p-value < .1/ R}2_{-change based on Model 7}

(26)

25

(here: likes) indirectly while presented with entertaining content, which leads to the confirmation of H6.

The control variable of subscribers shows significance in all measured models (model 9: b=.275; p=.001) which indicates a positive effect towards the number of likes.

4.2.3 Customer Engagement: Dislikes

Table 10 continues with the third metric for customer engagement, dislikes. While customer engagement is defined as a customer’s activity towards a company (Pansari & Kumar 2017), previous metrics of views and likes are positive metrics for customer engagement, the number of dislikes is perceived as a negative form of customer engagement. Nevertheless, the correlation in chapter 4.1 stated positive correlations within these metrics.

Tab. 10: (Standardised) Coefficients for Dependent Variable Dislikes

(Constant) -49.699 -678.833** -219.714 -23.570 -54.414 -55.804 Main Variables Entertaining Content H1, H2 .005 .015 -.106 -.054 .000 .005 Celebrity Endorsement H3 .215** .239** .201** .215** .214** .218** Music Background H4 .103 Music Mood H4 .045 Music Appearance H5b .019 .011 .003 .017 .025 Voice-Over Character H5a -.109 Voice-Over Narrator (H5a) .159* Voice-Over Appearance H5b -.016 -.017 -.017 -.014 .026 Control Variable Subscribers .161** _.134* _.180** _.152** _.163** _.161** Interaction Effects Entertaining Content x Celebrity Endorsement H6 .239*** Entertaining Content x Music H7 .067 Entertaining Content x Voice-Over H8 .019 Music x Voice-Over H9 -.052 R2 (adj. R2) .079 (.047) .132 (.088) .127 (.090) .080 (.041) .080 (.041) .080 (.041) R2_change _.048*** _.001 _.000 _.001 F-Value 2.462** 3.051*** 3.442*** 2.062* 2.046* 2.062*

Note: ***_{p-value < .01;}**_{p-value < .05;}*_{p-value < .1 / R}2_{-change based on Model 13}

(27)

26

Moreover, the interaction of customer engagement and celebrity endorsement is significant (b=.239; p=.006). Thus, the appearance of celebrity endorsement interacting with entertaining content appears engaging, confirming H6. Moreover, R2 improved with the interaction of entertaining content and celebrity endorsement (R2=.127; R2-change=.048; p=.006). Compared to the positive effect of celebrity endorsement in previously described

models, the appearance of voice-over narrators shows marginal significance and a positive effect towards the number of dislikes (b=.159; p=.0.85). Hypothesis 5a expected to show effects of character speaking, however, narrator speaking showed a positive effect. Therefore, this result disproved previously formulated hypothesis, but still gives further insights into the use-of-voice. Even though a positive effect is presented, the increasing number of dislikes can be listed as decreasing tool for a successful customer engagement. Other analyses for the dependent variable dislikes did not result in significant values. 4.2.4 Customer Engagement: Comments Table 11 shows the standardised coefficients for the dependent variable of comments. Tab. 11: (Standardised) Coefficients for Dependent Variable Comments

(Constant) -43.797 -303.856* _-111.902 _-32.111 _-49.808 _-44.401 Main Variables Entertaining Content H1, H2 -.028 -.020 .055 -.084 -.042 -.028 Celebrity Endorsement H3 .192* .215** .180* .191* .188* .192* Music Background H4 .094 Music Mood H4 .031 Music Appearance H5b .015 .008 .000 .011 .016 Voice-Over Character H5a -.074 Voice-Over Narrator (H5a) .168* Voice-Over Appearance H5b .015 .015 .015 .021 .024 Control Variables Subscribers .178** _.152* _.194** _.180* _.186** _.178** Interaction Effects Entertaining Content x Celebrity Endorsement H6 .198** Entertaining Content x Music H7 .062 Entertaining Content x Voice-Over H8 .049 Music x Voice Over H9 -.011 R2 (adj. R2) .070 (.038) .117 (.074) .103 (.065) .071 (.032) .072 (.033) .070 (.031) R2 change .033** .001 .002 0 F-Value 2.162* 2.679** 2.719** 1.809 1.847* 1.790

(28)

27

Celebrity endorsement shows a significant effect on the number of comments (b=.215; p=.029) compared to previously listed metrics. This is the lowest coefficient for celebrity endorsements appearance. Furthermore, the interaction of entertaining content and celebrity endorsement increases the number of comments significantly (b=.198; p=.024) and confirms H6. In line with the metric of dislikes, the appearance of voice-over narrators shows a marginal significance and a positive effect towards the number of dislikes (b=.159; p=.0.85). 4.2.5 Hypotheses Testing

According to previous results, H1 and H2 do not show significant results and therefore are not supported. However, H3 - the direct effect of celebrity endorsement on customer engagement is confirmed while testing it on all metrics for customer engagement (model 1, 7, 13, 19). H4 does not show significant results. Consequently, no conclusions about any effects of music result out of this study. H5a, the negative effect of a character speaking is supported by the metric number of views (model 2). Due to missing significance, H5b cannot be supported. Following table 12 sums up the testing of all hypotheses and (non-)confirmations.

Tab. 12: Hypothesis (No-)Confirmation Listing

Views Likes Dislikes Comments

H1 Informational content in a video advertising influences customer

engagement positively.

û û û û

H2 Entertaining content in a video advertising influences customer

engagement more positively than informational content.

û û û û

H3 Celebrity endorsement in a video advertising influences customer

engagement positively.

ü ü ü ü

H4 Background music in a video advertising positively influences

customer engagement compared to mood music and absence of music.

û û û û

H5a Use-of-voice (character speaking) in a video advertising negatively

influences consumer-engagement. ü û û û H5b The effect of music on customer engagement is more positive than the effect of use-of-voice in video advertising. û û û û H6 Celebrity endorsement in a video advertising enhances the effect of type of content on customer engagement. ü ü ü ü H7 Music in a video advertising enhances the effect of type of content on customer engagement. û û û û H8 Use-of-voice in a video advertising mitigates the effect of the type of content on customer engagement. ü û û û H9 Music in a video advertising enhances the effect of use-of-voice on customer engagement. û û û û

(29)

28

music on the type of content for H7 is not supported (Model 4, 10, 16, 22). Furthermore, H8 is partly supported, as for only one significant model (5). Lastly, H9 is not supported, no models show significant results in the interaction of music and voice-over (Model 6, 12, 18, 24).

5. Conclusion and Discussion

This study provides insights into the effectiveness of brand posts on social media and determines the effectiveness of specific aspects in video-advertisement. The role of celebrity endorsement, the use of voice-over and moderating roles of message appeals, define a new view for the relation between online content and customer engagement. Thus, the research question ‘To what extent do the type of content, celebrity endorsement, appearance of music

and voice in commercial videos (or video ads) influence customer engagement?’ is partially

answered. In particular, the effectiveness of celebrity endorsement was confirmed multiple times and in multiple online metrics. Furthermore, the appearance of character-voices influences the number of views negatively. This implies that especially instead of focusing on the content itself, the appearance of celebrity endorsers is perceived as one main reason for a customer to engage with videos that are posted by brands. The results of this study indicate a trend while YouTube is representing social media platforms (Kaplan & Haenlein 2010), other networking platforms which offer video-uploads may result in comparable outcomes. In the past, literature gave insights about the effectiveness of branded digital content on customer engagement of brand posts (de Vries, Gensler & Leeflang 2012), creative strategies (Ashley & Tuten 2014) and user-generated content (Berger & Milkman 2010). This study took up the issue and is built upon previous studies, especially on the study by de Vries, Gensler and Leeflang (2012). Through adding variables and changing the platform, an extended view of customer engagement and its influencers is created.

(30)

29

Another interesting point is that any type of content does not yield significant results. Although the informational content was described as the key reason for customers to respond (Lin & Lu 2011), it does not indicate a direct influence on customer engagement. Users might not watch these videos to inform nor entertain themselves, but to see celebrity endorsers, since the direct effect of celebrity endorsement is confirmed. This explanation may also be applicable for the appearance of music and background music, while no form of music resulted in significant results. This outcome could mean that customers do not value the appearance of music with engagement on social media platforms. Besides, music was previously described as hard to differentiate (Galan 2009; Zander 2006). Harnett et al. (2016) described music in the context of sales effectiveness, while further literature only focused on the effect of music on its own (Galan 2009). Existing research described the deficiency of research in the field of music in advertising several times (Abolhasani, Oakes & Oakes 2017; Oakes 2007; Hung 2000). Consequently, further research in this field is still necessary.

Regarding the appearance of voice-over, a negative effect only regarding views was significantly influencing customer engagement, confirming results by Harnett et al. (2016). Moreover, the interaction of entertaining content and voice-over mitigating customer engagement is confirmed (Tan 2010). The similar analysis using informational content and the use of a voice corroborates a positive outcome of the number of views. Controversially, since the direct effect of voice-over is negative, this outcome needs to be taken into consideration as well since it did result in high significance. Surprisingly, the appearance of narrator-voice increases dislikes and comments. No hypothesis was formulated regarding the appearance of a narrator's voice. Although voice-over appears as a positive influencer for customer engagement, it is perceived as negative, meaning that dislikes are increasing while using a narrator's voice. Due to very high correlations between comments and dislikes, a similar trend for both of these metrics is noticed. This also indicates that the content of comments is not purely measurable as positive customer engagement since an undefined number might contain negative content (Choudhury & Breslin 2010). Literature described the positive effect of a spokesperson in combination with music to increase advertising (Martín-Santana, Reinares-Lara & Muela-Molina 2015). Results outline that the comparison of radio and video advertisements is not appropriate in this case. However, the negative and independent outcome of a narrator’s voice is a new implication for advertising’s effectiveness. Consequently, any form of voice-over in combination with entertaining content, but also the use of character- and narrator-voices is perceived as a destructive tool and not successful in online video advertising.

(31)

30

Of note is that multiple models did not show significance in the metrics of likes, dislikes and comments. It goes without saying that requesting sign-up to interact on these levels with content on YouTube (YouTube Help 2017) can be seen as a hurdle which customers have to overcome first. Previous studies in the field of user privacy state that users exhibit the so-called ‘publicly private behaviour’ (Lange 2007). This describes that when identities are revealed, users seem less involved and less willing to share on social media than in the situation where the user is not named (ibid). Hence, results regarding the difference of dependent variables corroborate with previous findings of social media platforms. It is the logical outcome for fewer engagement numbers while most significant results are measured in the models for the numbers of views.

More subscribers resulted in higher customer engagement metrics. Besides, this can be described by the fact that attitude towards advertising has the most engaging effect when being highly involved, such as in the form of a subscription (Galan 2009). The video lifetime had a negative effect on customer engagement. As literature described earlier, most interaction by users is recorded during the first three up to four weeks (Cheng, Dale & Liu 2008). Consequently, this result confirms their findings and leads to the suggestion of not extending the stay of video advertisings online. The variable video length did not change customer engagement’s outcomes. This might be based on the low median of 66.79 (SD 60.58), but also on the type of media, since previous literature focused on written words and this study on video content (de Vries, Gensler & Leeflang 2012; Berger & Milkman 2012).

To relate findings to managerial implications, the use of celebrity endorsement is highly recommended. Additionally, using more than one celebrity will increase online metrics even more. Thus, the appearance of many celebrities during entertaining content while using no form of voice-over is according to this study’s design the most appealing option for customers. Furthermore, important to consider while planning advertisements is that neither a character nor narrator speaking increase engagement. Besides, the interaction of any type of voice-over while promoting with entertaining content is not recommended as well, leading to the consequence to avoid voice-overs, as long as it is not a highly informational video. One suggestion is to add information about the video’s content in the video-title, for instance, the appearance of a celebrity. This might increase customer engagement, especially views. Of course, this is only working when it is not already done, which leads to a limitation of this study: the title’s content was not conducted and thus results about the initial impression to decide whether to watch a video or not is not researched.

(32)

31

a reason to subscribe to their online platform, which then is related to the increase in multiple metrics of customer engagement.

In conclusion, these findings imply to be thoughtful when selecting an online marketing strategy. When online advertising in the form of videos on social media is used, managers should use celebrity endorsers, no voice-overs and keep engagement-numbers into consideration to increase online metrics, such as views and likes.

Moreover, some limitations of this study must be acknowledged. First of all, a limitation due to the use of the platform YouTube must be considered. Although literature defines YouTube as a social media platform (Kaplan & Haenlein 2010), every social media platform is special on its own. While YouTube’s users are 18 up to 34 years old (YouTube Press 2018), Facebook-users are statistically older and between 25 and 34 years (Zephoria 2018). In a further analysis, these statistics might influence results regarding the effectiveness of branded digital content. Also, as previously suggested, the content of video’s titles may influence customer engagement, such as the appearance of a celebrity’s name. Especially on the platform YouTube, the title is one of the first indicators for videos. Based on the results of celebrity endorsement’s success on digital content, the title and its appearance on social media is part of the limitations and should be a next step investigated further.

This study was unable to encompass the entire content of comments. However, the content of comments might change results in such a way to understand how and possibly why customers engage with a brand (Campbell et al. 2011). Previously studies worked on a sentiment analysis to gain insights about comments on social media (Choudhury & Breslin 2010) but also stated a lack of understanding due to the special slang on social media platforms. Thus, future research should focus further on comments and the content concerning the use of music, voice-over and celebrity endorsement to gain more insights about the success of branded digital content and the reasons why customer engage with individual videos.