Marketing Magazines in a Web 2.0 World
How Content Characteristics Influence Online Content Virality and Success
Master’s Thesis Graduate School of Communication Master’s Programme Persuasive Communication
Danitscha Mylène van Zijverden, 10772308 Supervised by Peter Neijens
2 Abstract
Magazines are marketing their product by achieving virality with their online content through social media. When consumers engage with this content on social media by liking, sharing or commenting on it, this helps disseminating the message to a larger audience. This research aimed to offer more insight on whether the content characteristics category, emotion, media richness and posting moment influence the number of readers of the online articles and on the role of engagement in this process. A quantitative content analysis of a Dutch magazine’s online content (n=100) was conducted. The data from the content analysis was combined with the magazine’s Facebook analytics to measure engagement and with the website analytics to measure the number of readers. The results show that the content characteristics do not influence
engagement or the number of readers. Engagement does significantly influence the number of readers, meaning that virality is important in the success of online content by magazines.
Introduction
The magazine industry is facing tough times as both magazine sales and spending on magazine advertising have been declining since 2008 (www.mediamonitor.nl). The consumption
3 of content has shifted to the online environment, so advertising spending has moved with it. People are spending more time per day consuming media online than they are doing through traditional media such as print magazines (Mander, 2014). This has created the necessity for magazines to have an online presence, in which they offer content to stay relevant to consumers and bind them to their brand. Offering content online also gives magazines the opportunity to create revenue from website advertisements. The online content is a marketing strategy for the print magazine as well as a new revenue model. It is therefore important that the online content published by magazines reaches a large audience of readers.
Through their own website and social media accounts magazines can connect to online consumers. Some offer a (paid) digital version of the magazine, while other magazines upload separate articles on a blog on the magazine’s website. The blog articles offer content about the same type of subjects that would be written about in the print version of the magazine. Magazines attract attention to this blog content by linking to it on the social media channels of the magazine in the hopes that consumers will respond to it. The use of social media in the dissemination of the blog content allows consumers to not just consume the content but also engage with it, by liking or sharing it or commenting on it (Cvijikj & Michahelles; Mills, 2012). This engagement with a post on a social medium means that it can reach an audience beyond the followers of the
magazine itself, because now it is also visible for the followers of the person that shared, liked or commented on the post (Sabate, Berbegal-Mirabent, Cañabate, & Lebherz, 2014). When content is shared a lot in a short amount of time this leads to exponential growth in its reach, and it goes what is called ‘viral’ (Cruz & Fill, 2008; Cvijikj & Michahelles, 2013; Ho & Dempsey, 2010; Mills, 2012; Sabate et al., 2014).
Recent research on virality and sharing behavior has pointed to some predictors of the spread of online content. Certain characteristics of content have proven to affect its viral success. Though Southgate, Westoby and Page (2010) did not find a significant effect of the content’s category on virality, Zhang, Peng, Zhang, Wang and Zhu (2014) and Phelps et al. (2004) found that categories of content do differ in popularity. The reason why certain categories of content are more successful than other could be explained by the fact that they are more general in their topic, and therefore relevant to a larger audience (Botha & Reyneke, 2013). Media richness, which combines the vividness and interactivity of media content, was also found to have an effect on consumer engagement, although there is no agreement on which level of media richness works
4 best (Cvijikj & Michahelles, 2013; De Vries, Gensler & Leeflang, 2012; Sabate et al., 2014). The day of the week and the time of posting influence engagement, but there is no agreement on the best moment to put content online (Cvijikj & Michahelles, 2013; Sabate et al., 2014). One characteristic of content that was generally accepted as a predictor of sharing on different
platforms was the presence of emotions, both for brand messages (Dobele, Lindgreen, Beverland, Vanhamme & Van Wijk, 2007; Dobele, Toleman & Beverland, 2005) and non-brand related messages (Phelps et al., 2004; Zhang et al., 2014), especially positive emotions (Berger & Milkman, 2010; Botha & Reyneke, 2013; Eckler & Bolls, 2011; Nelson-Field, Riebe &
Newstead, 2013). According to Botha and Reyneke (2013) this effect of emotion would only be present if the content were actually relevant for the viewer. Araujo, Neijens and Vliegenthart (2015) found that emotional cues alone do not increase sharing of brand messages on social medium Twitter, they needed to be combined with informational cues to be effective.
These are factors of the content that can influence engagement, however they do not measure the effects of engagement itself on the number of readers. When content is shared, liked or commented on by someone that people know or trust this adds credibility to the content. This electronic word of mouth (eWOM) makes it more likely that people will read it than if it comes directly from the brand (Araujo et al., 2015; Cruz & Fill, 2008; Dobele et al., 2005; Ho & Dempsey, 2010; Kilgour et al., 2015). Ultimately virality is used as a means to attract a large audience for the content, which is why the present research will establish the relationship between content characteristics, consumer engagement and the success of the content on the number of readers. This has led to the following research questions, that will be answered with a quantitative content analysis of the online content on the blog and the website- and Facebook analytics of a well-known magazine that focuses on history, culture and nature:
How do the category, emotion, media richness and posting moment of online content affect the number of readers of the articles on the magazine’s website, and what is the role of social media engagement in this process?
The research is scientifically relevant because it will help in giving more insight in virality, specifically by looking into characteristics that positively affect virality and testing the mediating effects of virality on success of online content. The application of this knowledge to
5 the magazine industry broadens our understanding of virality by linking it to a sector in which content was already the main product. Much research up to now has focused specifically on viral advertising campaigns that were used by brands as a means to engage their audience (Araujo et al., 2015; Botha & Reyneke, 2013; Cruz & Fill, 2008; Cvijikj & Michahelles, 2013; De Vries et al., 2012; Dobele et al., 2005; Dobele et al., 2007; Eckler & Bolls, 2011; Ho & Dempsey, 2010; Phelps et al., 2004), which is different from publishing content online because unlike advertising campaigns it does not focus on actively trying to sell the product by promoting the brand. Instead it tries to bind consumers to a brand by offering them valuable content that is not about the brand or its products. The focus of previous research on virality was often only on sharing content through e-mail (Berger & Milkman, 2010; Dobele et al., 2005; Phelps et al., 2004) or on sharing content on a certain social media platform (Araujo et al., 2015; Cvijikj & Michahelles, 2013; De Vries et al., 2012; Hollenback & Kaikati, 2012; Sabate et al., 2014), but not on using social media as a means to drive traffic to a website. The results of this research will also be useful for practice because the findings will show magazines what type of content they should offer online to get more engagement and readers, which in turn can help them build a stronger brand relationship with consumers and help them increase advertising revenue from their online articles.
Theoretical framework Web 2.0 and social media
The internet has transitioned from web 1.0 to web 2.0, from a medium that was mostly used by organizations to send one-way information to a more participatory platform in which consumers can interact and collaborate with each other and with organizations (Macnamara & Zerfass, 2012; Mills, 2012; Singh, Veron-Jackson & Cullinane, 2008). Social media are web 2.0 tools that facilitate this participation, collaboration and interaction between their users (Kilgour et al., 2015; Singh et al., 2008; Macnamara & Zerfass, 2012). Examples of such social media are social networks such as Facebook, LinkedIn and Twitter, wiki’s or blogs (Castronovo & Huang, 2012; Kaplan & Haenlein, 2010; Kilgour et al., 2015). Even though social media were initially focused on the interaction between consumers, they have also enabled organizations to interact directly with consumers and build a relationship with them (Castronovo & Huang, 2012, Fournier & Avery, 2011; Kilgour et al., 2015; Mills, 2012). Social media offer organizations an efficient way for marketing products because the costs are lower than those of traditional media and there is immediate feedback from the audience (Castronovo & Huang, 2012).
6 Any organization can post content on social media, but its success is dependent on
whether consumers are engaging with it through likes, shares and comments (De Vries et al., 2012; Sabate et al., 2014). Followers of consumers engaging with a brand post on social networks will see the post in their feed because of this interaction, it would not be visible to them without it unless they follow the brand themselves (Sabate et al., 2014) or the brand pays to advertise their post. For Facebook specifically even following a brand does not mean users will always see these posts on their timeline. An algorithm controls which posts show up in users’ newsfeed on
Facebook, depending among other things on how often these followers interact with the brand or similar posts and how much engagement the post receives from others (www.facebook.com). There is an option for brands to pay Facebook to spread content among more followers and advertise it to non-followers (www.facebook.com). Consumers engaging with content through shares, comments or likes on social networks is however crucial for making content go viral (Cvijikj & Michahelles, 2013; Sabate et al., 2014).
Viral marketing
When peer-to-peer sharing is used for the dissemination of messages by brands this is called viral marketing (Cruz & Fill, 2008; Dobele et al., 2005; Singh et al., 2008; Southgate et al., 2010). Reaching a large audience of customers and potential customers is an important goal in viral marketing (Cruz & Fill, 2008; Dobele et al., 2007; Dobele et al., 2005; Mills, 2012; Nelson-Field et al., 2013; Phelps et al., 2004). The messages in viral content can influence opinions about brands (Botha & Reyneke, 2013). It is closely related to word-of-mouth, in which information about brands or products is a part of interpersonal communication (Cruz & Fill, 2008; Cvijikj & Michahelles, 2013; Dobele et al., 2005; Grifoni, D’Andrea & Ferri, 2012; Harvey, Stewart & Ewing, 2011; Phelps et al., 2004). It can also take place online as electronic word-of-mouth (Cruz & Fill, 2008; Cvijikj & Michahelles, 2013), where the information can also be spread to a large group of receivers. Consumers sharing online content by a brand can thus be seen as viral marketing and electronic word of mouth because of the exchange of brand content between consumers (Dobele et al., 2005).
The fact that viral marketing messages reaches consumers through a social source instead of directly from a commercial source makes them more credible (Cruz & Fill, 2008; Dobele et al., 2007; Dobele et al., 2005; Eckler & Bolls, 2011; Harvey et al., 2011; Kilgour et al., 2015; Phelps et al., 2004). Word-of-mouth by peers can influence the consumer decision making
7 process (Berger & Milkman, 2010; Castronovo & Huang, 2012; Cruz & Fill, 2008; Cvijikj & Michahelles, 2013; Zhang et al., 2014). People are known to try to resist being persuaded by marketer messages (Dobele et al., 2005), but fellow consumers are usually perceived to be more objective (Cruz & Fill, 2008; Dobele et al., 2005; Phelps et al., 2004;). They are sharing content to show how they are, to help others and to connect with people, but they do not have commercial incentives (Hollenbeck & Kaikati, 2012; Kilgour et al., 2015; Lewis, Mobilio, Perry, & Raman, 2004; Phelps et al., 2004). The relationship that people have with the person sharing the content reflects on the message, so if a friend shares content it is perceived more positively than if a brand shares that same content (Castronovo & Huang, 2012; Kilgour et al., 2015). Because the viral marketing message reaches its audience through other consumers that are not trying to sell the brand that created the content, the message becomes more credible.
Factors influencing virality
Research aimed at finding what characteristics of content influence whether it goes viral has been carried out for different online channels such as e-mail and several social networks. The characteristics were found to have an effect were category of the content, the evoked emotion, the media richness and the day and time of posting. Each of these factors will be discussed
separately. Category.
Content can fall into many different categories of information. One of the earliest studies into categories of content and its influence on virality was done by Phelps et al. (2004) and looked into e-mail forwarding. Different categories of e-mails elicited different forwarding behaviour in its receivers, e-mails about good deeds and jokes about gender issues for example were forwarded a lot while e-mails about politics and product warnings were not forwarded (Phelps et al., 2004). Many changes have occurred on the web since this study, with the advent of web 2.0 consumers starting sharing content through social media networks. Research by Zhang et al. (2014) on microblogging sites showed that categories influenced sharing behaviour on this social medium as well. Personal interest, social interactions, jokes and gossip had significantly more comments and shares than the social, political and business news category, and utility and personal life had significantly more comments (Zhang et al., 2014). Both researches look into virality of content from any type of sender, so brand content and content sent by contacts are not separated. Southgate et al. (2010) researched the success of television advertisements that went
8 viral on YouTube. The results showed that the videos’ success were not affected by category and brand interest (Southgate et al., 2010). Southgate et al. (2010) however measured category together with brand interest as one construct, to measure consumer interest in both the category of the video’s content and the product category of the brand. Perhaps if category and brand interest were measured separately there would be a significant effect. Because Phelps et al. (2004) and Zhang et al. (2014) found that category influenced virality the following research question was formulated:
RQ1: (a) Does the category of blogposts have an influence on the amount of consumer engagement (b) and the number of readers?
Media richness.
The media richness theory was developed by Daft and Lengel (1986) and was built on the idea that uncertainty and equivocality influence information processing. Uncertainty about
information exists when people do not have the full range of information available to them, for example in a phone conversation there is more uncertainty than a face-to-face conversation because you cannot see the other person’s facial expressions (Daft & Lengel, 1986). Information is equivocal when it can be interpreted in more than one way, for example because a word has multiple meanings. Daft and Lengel (1986) pose that face-to-face communication has the highest media richness because of the immediate feedback, number of cues, use of language and because it is so personal, enabling people to understand complex messages better than through media with a lower richness such as numeric documents.
For online media, richness has been defined as the vividness and interactivity of the content (Cvijikj & Michahelles, 2013; De Vries et al., 2012; Sabate et al., 2014). Vividness is about content features that stimulate different senses; videos for example are more vivid than images because they stimulate both sight and hearing while images only stimulate sight (Cvijikj & Michahelles, 2013; De Vries et al., 2012). Interactivity entails “the degree to which two or more communication parties can act on each other, on the communication medium, and on the messages and the degree to which such influences are synchronized” (Liu & Shrum 2002, p. 54). Content with videos or links is more interactive than content with text and images only because readers have to click videos and links (Cvijikj & Michahelles, 2013; De Vries et al., 2012).
9 Vividness and interactivity influenced popularity of brand posts on Facebook (Cvijiki & Michahelles, 2013; De Vries et al., 2012; Sabate et al., 2014), but the effects on the different types of engagement were different in the two studies. While the results by De Vries et al. (2012) showed that more vividness led to significantly more likes but had no effect on comments, the results by Cvijiki & Michahelles (2013) and Sabate et al. (2014) show that images (which have low vividness) received the most engagement. More interactivity led to more comments but not likes in the research by De Vries et al. (2012) but in the research by Cvijikj and Michahelles (2013) and Sabate et al., 2014) it led to less engagement. Because of the diversity in findings the following research question was formulated:
RQ2: (a) Does the level of media richness have an influence on the amount of consumer engagement (b) and the number of readers?
Posting moment.
The time and day of posting can influence the success of brand posts on Facebook because the newsfeed is constantly updated with new content (Cvijikj & Michahelles, 2013; Sabate et al., 2014). Content that that is engaged with or content that is new is displayed at the top of the Facebook feed. So posting content on Facebook during when users are usually online, which is during peak hours (4 p.m. to 4 a.m.) according to Cvijiki and Michahelles (2013) and business hours according to Sabate et al. (2014), can increase the chance that the content is seen and engaged with by consumers, as it will likely be higher up in their newsfeed (Sabate et al., 2014). Cvijiki & Michahelles (2014) found that content that was posted on weekdays received more comments, but there was no effect on shares and a negative effect on the number of likes. Sabate et al. (2014) did not find a significant effect of posting during a weekday on engagement. When it comes to the time of posting during workhours led to more engagement according to Sabate et al. (2014) and posting between 4 p.m. and 4 a.m. led to less likes and shares according to Cvijikj and Michahelles’ (2013) results. Because existing research has not agreed on the best moment to post content the following research questions were formulated:
RQ3: (a) Does the day at which content is posted have an influence on consumer engagement (b) and the number of readers?
10 RQ4: (a) Does the time at which content is posted have an influence on consumer engagement (b) and the number of readers?
Emotion.
Emotions are important in online sharing behaviour just like they are in other social situations (Dobele et al., 2005). They can make content more interesting and elicit more responses (Zhang et al., 2004). Sharing behaviour is influenced by emotions that are felt when consuming content (Botha & Reyneke, 2013; Dobele et al., 2005; Phelps et al., 2004; Zhang et al., 2014). Botha and Reyneke (2013) found that emotions were the final step in deciding whether to share a video or not and Phelps et al. (2004) also found that emotions were an important reason to forward e-mails. Dobele et al. (2005) found that the primary emotions surprise, joy, sadness, anger, fear and disgust influenced forwarding behaviour in their research of viral marketing campaigns. Messages with more emotional cues such as emoticons and words and punctuation that indicated mood led to more engagement (Zhang et al., 2014). According to research by Araujo et al. (2015) brand messages on Twitter were not retweeted more because of emotional cues alone, but a combination of emotional cues with informational cues, hashtags and links created the highest levels of sharing. The inclusion of emotion increases sharing, but different emotions have different levels of effect. Not only is content with emotion shared more than content without emotion, but positive emotions elicit more sharing than negative emotions (Berger & Milkman, 2010; Botha & Reyneke, 2013; Eckler & Bolls, 2011; Nelson-Field et al., 2013). Positive emotions are thus effective in influencing virality. Hence the following
hypotheses were formulated:
H1: Content that elicits positive emotions receives more engagement than content that elicits negative emotions.
H2: Content that elicits positive emotions has more readers than content that elicits negative emotions.
Engagement as a mediator
The success of viral content is often measured through consumer engagement, which is measured on social media by the shares, likes and comments on the content. However
11 2008; Nelson-Field et al., 2013). The content’s category, media richness, posting moment and evoked emotions are characteristics that in previous research showed to influence engagement. If engagement influences the size of the audience this means there is a mediation effect. Therefore the following hypothesis was formulated:
H3: More engagement with a post on social media leads to a higher number of readers.
Methodology Research design and sample
This study uses a combination of content analyses and Facebook and website statistics to measure content characteristics and measure their influence on engagement and the number of readers, and whether the influence of characteristics on the number of readers can be
explained through the amount of engagement. This method was chosen because it enables the systematic description of different types of content, in this case the content characteristics. By combining it with existing data of actual online behaviour from the Facebook- and website analytics that measure interaction with the Facebook page and the website traffic, the research question can be answered.
The material chosen for this research is the online content created by a well-known Dutch magazine about history, culture and nature. Its print magazine has a circulation of over 100,000 copies per month. It has its own website with blog articles and the magazine is also present on different social media, such as Facebook, Twitter, Google+, Pinterest, Instagram and YouTube. Because over 99% of visitors that arrive on the magazine’s website through social media come from Facebook, this will be the only social medium that will be studied in this research. The magazine’s Facebook page has 120,000 followers and all of their posts on this social network receive engagement through likes, comments and shares. With about three new articles on their website each day, the magazine is actively creating online content on a regular basis. This makes for a large sample of articles that are available for this research. The
Facebook page promotes the blog articles on the website with a short text about the article, an image and a link to the article. This research aims to find the influence of the article
characteristics on engagement with the post on Facebook, and whether this engagement leads to more readers.
12 A random sample of 100 blog articles has been drawn from a sample of 714 articles published on the website between May and October 2015. The specific time frame was chosen because it was the most recent data available after changes in the magazine’s website and online strategy that took place in April.
Measures
A codebook was created in which the guidelines for coding all variables for the content analysis have been defined (see appendix 1).
Category.
The content of the magazine’s website is divided into the categories ‘humans’, ‘food’, ‘wildlife’, ‘earth’, ‘history’, ‘photography’, ‘science’, ‘news’ and ‘advertorials’. While there might be some overlap in the categories the articles could possibly fit into, the magazine puts each article in one category only. Roughly the same categories that the magazine’s website uses, were used for coding the category variable in this research. The exceptions are the contests that on the magazine’s website fall under the categories ‘news’ and ‘photography’. The contests and their results have been coded into an additional category for contests because they are different from other articles in the ‘news’ category as they do not describe current events. Another exception was made for articles that are clearly about the magazine itself, these are categorized in the ‘humans’ category on the website but will be coded under the ‘brand’ category in this research because information about the magazine or the brand does not fit in with articles about human development. Because on the magazine’s website the ‘food’ category was barely used and articles about food were often categorized under the ‘human’ category instead, the same will be done in this research because these articles are about
developments in human food consumption. The ‘photography’ and ‘news’ categories have also been excluded, because so many articles include photos or news but the subject of the articles fell in many different categories, so those articles were coded according to their subject instead.
To ensure that the categories are reliable four extra coders have coded the category on 25 of the articles each, so that all of the articles were coded twice. The initial level of agreement between coders was 96%. After discussing the category of the 4% of cases that were not
13 The most used categories were animals (37%) and science (26%), followed by earth (11%), history (10%), humans and brand (7%). Articles about contests were posted the least (2%) and there were no advertorials in the sample.
Media richness.
Media richness was determined by checking the inclusion of images, videos and links in the articles (Cvijikj & Michahelles, 2013; Sabate et al., 2014). Media richness is made up of the variables vividness and interactivity (Cvijikj & Michahelles, 2013; De Vries et al., 2012; Sabate et al., 2014), which were combined into one variable with four categories (see table 1). Plain text articles are not vivid or interactive because they stimulate the senses the least and because the content is static (Cvijikj & Michahelles, 2013; De Vries et al., 2012). Articles with images have low vividness but no interactivity because they stimulate sight but are still static content (Cvijikj & Michahelles, 2013; De Vries et al., 2012). When links are included in the article it is interactive and has medium vividness because the reader can click them and it leads the reader to other webpages with more text and images (Cvijikj & Michahelles, 2013). Articles with videos are interactive and highly vivid because they have to be clicked to start and they stimulate both sight and hearing (Cvijikj & Michahelles, 2013).
There were no plain text articles in the sample, 27% of the articles contained images but no links or videos, 41% contained links but no videos and 32% contained videos.
Table 1
Overview of media richness divided up into interactivity and vividness for online articles No interactivity Interactivity
No vividness Plain text -
Low vividness Images -
Medium vividness - Links
High vividness - Videos
Source: Cvijikj and Michahelles (2013)
Posting moment.
The moment of posting the articles is divided into two variables; the day of the week and the time the article was posted. This data was retrieved from the Facebook analytics which show when the link to the article was posted on Facebook. The weekdays were not divided into
14 week and weekend as was done in Cvijikj and Michahelles (2013) and Sabate et al. (2014) but all weekdays have been coded, so that the data is as detailed as possible. Posting time was divided into the different parts of the day, which take 6 hours each; morning, afternoon, evening and night.
Most articles were uploaded on Tuesdays (25%) and Thursdays (20%), followed by Mondays (13%), Wednesdays, Saturdays and Sundays (all 11%). Only 9% of the articles was uploaded on Fridays. Articles were uploaded in the afternoon (41%) and evening (40%) most often. 19% of the articles were posted in the morning and none were posted at night.
Emotion.
The most prevalent emotion that the article brings out was initially coded from the dictionary definition of the 16 emotions used by Nelson-Field et al. (2013f) (see table 2). The list includes eight positive emotions and eight negative emotions. Both the positive and negative emotions have four activating emotions. As emotions are subjective four extra coders have coded the emotion of 25 articles so that all of the articles were coded twice. The level of agreement between coders using the 16 emotions was only 28%. To increase the intercoder reliability the variable was simplified. Using four categories of emotions, which were positive activating, positive non-activating, negative activating and negative non-activating raised the agreement between coders to 57%. By measuring emotion in postive and negative the level of agreement between coders increased to 92%, so emotions were coded as positive or negative. The articles for which the emotion was not agreed upon were left out of the analysis.
The majority of the articles evoked positive emotions (85.9%), the remaining 14.1% of articles evoked negative emotions.
Table 2
List of positive and negative activating and non-activating emotions
Positive Negative
Activating Non-activating Activating Non-Activating
Hilarity Inspiration Astonishment Exhilaration Amusement Calmness Surprise Happiness Disgust Sadness Shock Anger Discomfort Boredom Irritation Frustration Source: Nelson-Field et al. (2013)
15 Consumer engagement.
Consumer engagement on Facebook was measured through shares, comments and likes. These are the three ways readers can interact with a post on Facebook, and by doing so the reader’s followers will also be able to see the post. They are disseminating the post by engaging with it. This data can be found in the Facebook analytics which includes data on all interactions on the magazine’s Facebook page, including data on all the separate posts that the magazine has put on Facebook. The number of likes, shares and comments have been added up to measure the total engagement because they are all ways that consumers are spreading
content to a larger audience by interacting with it. There was a strong significant correlation between likes and comments (r(100) = .88, p < .001) and likes and shares (r(100) = .92, p < .001) and between comments and shares (r(100) = .86, p < .001) (see table 1 in appendix 2).
The average engagement on a Facebook post was 561.77 likes, shares and comments (SD = 907.12). The post that received the least engagement had 45 likes, shares and comment and the post with the most engagement had 6709 likes, shares and comments.
Number of readers.
The number of readers on the website is measured through Google analytics. Because each online article has its own webpage the analytics can be retrieved for each separate article. The analytics show the total number of page views, the number of unique page views, the average time on the page and the bounce rate for each separate webpage featuring an article. The total number of page views counts the number of visitors on the article page, and includes returning visitors. The number of unique page views excludes returning visitors in counting the number of visitors on the page. The time on the page is only measured through the average time that all visitors have spent on the page. The best way to measure whether visitors of a webpage are actually reading it is by measuring the time they spend on the page. It is however not possible to exclude page visitors that spent less time on the page than it would take to read the article from the data afterwards (this would require changes in the setup of Google
analytics before the data is collected, and cannot be done in retrospect). Therefore the number of readers will be measured through the number of unique page views instead, because it counts how many unique visitors each article gets.
16 The blog articles had an average of 2496.60 readers each (SD = 4863.36). The most popular post had 33800 readers, the least popular one had 104 readers.
Results Category
To test RQ1(a) which asked whether the category of a blog post influenced consumer engagement a one way ANOVA was conducted. There was no significant effect of category on engagement, F(1, 93) = 1.41, p = .220 (see table 2 in appendix 2). To test RQ1(b) which asked whether the category of a blog post influenced the number of readers of the post another one way ANOVA was conducted. The analysis showed no significant effect of category on the number of readers, F(1, 93) = 1.35, p = .242 (see table 3 in appendix 2). The answer to RQ1 is that the category of a post does not influence engagement and the number of readers.
Media richness
To test RQ2(a) which asked whether media richness influenced engagement a one way ANOVA was conducted. The analysis showed no significant effect of media richness on
engagement, F(1, 97) = 2.26, p = .109 (see table 4 in appendix 2). The one way ANOVA analysis testing RQ2(b) which asked whether media richness influenced the number of readers showed no significant effect of media richness on the number of readers, F(1, 97) = .28, p = .756 (see table 5 in appendix 2). The answer to RQ2 is that the media richness of a post does not influence the engagement and the number of readers.
Posting moment
To test RQ3(a) which asked whether the posting day influenced engagement a one way ANOVA was conducted. The analysis showed no significant effect of posting day on
engagement, F(1, 93) = .67, p = .675 (see table 6 in appendix 2). The one way ANOVA analysis testing RQ3(b) which asked whether posting day influenced the number of readers showed no significant effect of posting day on the number of readers, F(1, 93) = .45, p = .843 (see table 7 in appendix 2). The answer to RQ3 is that posting day does not influence engagement and the number of readers.
To test RQ4(a) which asked whether the posting time influenced engagement a one way ANOVA was conducted. The analysis showed no significant effect of posting time on
engagement, F(1, 97) = 1.64, p = .200 (see table 8 in appendix 2). The one way ANOVA analysis testing RQ4(b) which asked whether posting time influenced the number of readers showed no
17 significant effect of posting day on the number of readers, F(1, 97) = 1.87, p = .160 (see table 9 in appendix 2). The answer to RQ4 is that posting time does not influence engagement and the number of readers.
Emotion
To test H1 which hypothesized that posts that elicit positive emotions receive more engagement than negative emotions a one way ANOVA was conducted. The analysis showed no significant effect of emotion on engagement, F(1, 96) = .33, p = .804 (see table 10 in appendix 2). To test H2 which hypothesized that posts that elicit positive emotions receive a higher number of readers a one way ANOVA was conducted. The analysis showed no significant effect of emotion on the number of readers, F(1, 96) = .29, p = .830 (see table 11 in appendix 2). H1 is therefore rejected.
Engagement as a mediator
To test H3 which hypothesized that more engagement with a post on social media leads to a higher number of readers a correlation analysis was conducted. There was a strong significant correlation between engagement and the number of readers, r(100) = .72, p < .001 (see table 12 in appendix 2). H3 was therefore confirmed. This result is the only one of the three steps of the mediation analysis that is significant, because category, media richness, posting moment and emotion did not significantly influence engagement and the number of readers, therefore engagement is not a mediator.
Discussion and conclusion
The aim of this research was to find out whether a magazine’s online blog content characteristics such as category, emotion, media richness and posting moment affect number of readers of its articles, and what role social media engagement plays in this process. The blog content by a well-known Dutch magazine about history, culture and nature was selected because it publishes articles frequently and because all of their content gets engagement on Facebook. The results were obtained through a quantitative content analysis of the magazine’s blog content and its website- and Facebook analytics. The expected effect of the content characteristics on
engagement and the number of readers was absent in the results, engagement was found to be a significant predictor for the number of readers.
18 The results in this research confirm the findings by Southgate et al. (2010) that category has no significant effect on engagement, while opposing Phelps et al. (2004) and Zhang et al. (2014). The differences in results could have occurred because all of the studies research virality on a different platform, this research focuses on engagement on Facebook, Phelps et al. (2004) study e-mail forwarding, Zhang et al. (2014) research microblogs and Southgate et al. (2010) research videos. These different platforms focus on different types of communication and with it comes different behavior by its users. Microblogging sites for example only allow short text posts while Facebook posts can be very long and e-mails are not publicly visible for everyone while social media content is.
The findings on media richness also opposed previous research (Cvijikj & Michahelles, 2013; De Vries et al., 2012; Sabate et al., 2014), as no significant effect on engagement was found. This might be due to the fact that previous research measured the media richness of the post on social media while this research measured the media richness of the article itself. Possibly richer media attracted more attention on busy social media feeds, and received more engagement because it made them stand out more. This would not have been the same for the blog articles, because after clicking the link the article is the only thing on the webpage demanding attention from the consumer.
The time of posting was found to influence engagement by Cvijikj and Michahelles (2013) and Sabate et al. (2014), but this research found no effect of the time that content was posted on engagement. The difference is that the studies by Cvijikj and Michahelles (2013) and Sabate et al. (2014) both included Facebook posts that were posted at nighttime, and their results showed that people engaged with posts that were posted during business hours (Sabate et al., 2014) and outside of peak hours (4 p.m. to 4 a.m.) (Cvijikj & Michahelles, 2013). This could point their results being explained through the fact that people engage with content more if it is posted at a time that they are usually awake. The magazine used in this research never uploaded content at night, which limits their time of posting to the hours of the day that people are usually awake and can thus be active on Facebook. If the sample had included posts that were uploaded during night-time it is likely time there might have been a significant difference as well.
Opposing the findings by Cvijikj and Michahelles (2013) but confirming those of Sabate et al. (2014) this research found the day of posting on Facebook did not significantly influence engagement. Both operationalized the day of the week by comparing weekdays and weekend
19 days, with the only difference being the type of brands they researched, namely fast moving consumer goods brands (Cvijikj & Michahelles, 2013) and traveling agencies (Sabate et al., 2014). Perhaps the engagement of posts on certain days can vary for different types of brands. In this case it could mean that people engage with brands that they associate with leisure (such as traveling agencies and magazines) the whole week, but products that are for everyday use like fast moving consumer goods are engaged with more during weekdays.
The hypotheses that predicted that articles that elicited positive emotions would receive more engagement and a higher number of readers than negative ones were based on previous research that showed a positive effect of positive emotions on virality (Berger & Milkman, 2010; Botha & Reyneke, 2013; Eckler & Bolls, 2011; Nelson-Field et al., 2013), but were not
confirmed. Perhaps this can be explained through the type of content that the magazine in this research specializes in. The magazines is known to write about serious subjects, such as the effects of human behaviour on the planet and nature, which can evoke negative emotions. It could be that because of the magazine’s image consumers accept content that evokes negative emotions in them, because they expected it beforehand.
As the goal of virality is to reach a large audience (Cruz & Fill, 2008; Nelson-Field et al., 2013) this research tested whether the expectation that more social media engagement led to significantly higher number of readers could be confirmed. Previous research skipped this step, measuring engagement only, without seeing if the engagement affects the size of the audience for the content. The results confirm this hypothesis and show that virality does increase the success of online content.
Practical implications
Magazines should try to achieve virality with their online content, because this increases its audience. By using consumers to market their product by spreading the content through online word-of-mouth it becomes more credible as interesting content, and thus increases the chance that others will consume the content (Cruz & Fill, 2008; Dobele et al., 2007; Dobele et al., 2005; Eckler & Bolls, 2011; Harvey et al., 2011) because consumers have no commercial intentions for engaging with the content (Hollenbeck & Kaikati, 2012; Kilgour et al., 2015; Lewis, Mobilio, Perry, & Raman, 2004; Phelps et al., 2004). This influences the decision making process (Berger & Milkman, 2010; Castronovo & Huang, 2012; Cruz & Fill, 2008; Cvijikj & Michahelles, 2013; Zhang et al., 2014), which in this case is clicking the link and reading the content. Because
20 magazines create the online content to stay relevant as a brand and to bind consumers to them, reaching a large audience is an important goal. To reach it consumers must engage with their content through likes, comments and shares so that it can go viral. Virality is however not
influenced by the category of the content, the media richness, the posting moment or the emotion it evokes, excluding them as predictors of which articles will go viral.
Limitations and further research
Future research should focus on finding more content characteristics that could influence the virality of online content. This research could not confirm that category, media richness, posting time and emotion influenced the amount of engagement and the number of readers. Possibly other content characteristics such as the length of articles, the length of paragraphs, the number of images, videos and links, the number of sub headers or other characteristics that might make an article more easy or enjoyable to read.
One of the limitations of this research is that it only focuses on one magazine with a specific topic (history, culture and nature). Future research should look into magazines about other topics, or combine different types of magazines and compare the success of their online content. Perhaps the amount engagement differs for different types of magazines. Different magazines also have different target audiences, some of which might be more active in engaging on social media than others. Also other magazines might have more traffic coming in from social media other than the magazine that was used in this research. Engagement on other social media besides Facebook could lead to different results. Another limitation is that all articles in this research were coded with an emotion, without taking into consideration that some articles might not evoke any emotions at all. Future research should establish whether content evokes emotions at all, before choosing which emotion it evokes.
As this research has confirmed the effect of engagement on the number of readers of online content through a content analysis, another point of interest for future research could be to look into the underlying processes of this effect. For example by researching the role of
credibility on the effect of engagement and reasons for engaging with magazines on social media through surveys with social media users.
21 Literature
Araujo, T., Neijens, P., & Vliegenthart, R. (2015). What Motivates Consumers To Re-Tweet Brand Content? The Impact of Information, Emotion, And Traceability on Pass-Along Behavior. Journal of Advertising Research, 55(3), 284-295.
Berger, J., & Milkman, K. (2010). Social transmission, emotion, and the virality of online content. Wharton Research Paper, 10-114.
Botha, E., & Reyneke, M. (2013). To share or not to share: the role of content and emotion in viral marketing. Journal of Public Affairs, 13(2), 160-171.
Castronovo, C., & Huang, L. (2012). Social media in an alternative marketing
communication model. Journal of Marketing Development and Competitiveness, 6(1), 117-134. Cruz, D., & Fill, C. (2008). Evaluating viral marketing: isolating the key criteria.
Marketing Intelligence & Planning, 26(7), 743-758.
Cvijikj, I. P., & Michahelles, F. (2013). Online engagement factors on Facebook brand pages. Social Network Analysis and Mining, 3(4), 843-861.
Daft, R. L., & Lengel, R. H. (1986). Organizational information requirements, media richness and structural design. Management Science, 32(5), 554-571.
De Vries, L., Gensler, S., & Leeflang, P. S. (2012). Popularity of brand posts on brand fan pages: an investigation of the effects of social media marketing. Journal of Interactive Marketing,
26(2), 83-91.
Dobele, A., Lindgreen, A., Beverland, M., Vanhamme, J., & Van Wijk, R. (2007). Why pass on viral messages? Because they connect emotionally. Business Horizons, 50(4), 291-304. Dobele, A., Toleman, D., & Beverland, M. (2005). Controlled infection! Spreading the brand message through viral marketing. Business Horizons, 48(2), 143-149.
Eckler, P., & Bolls, P. (2011). Spreading the virus: Emotional tone of viral advertising and its effect on forwarding intentions and attitudes. Journal of Interactive Advertising, 11(2), 1-11.
Fournier, S., & Avery, J. (2011). The uninvited brand. Business Horizons, 54(3), 193-207. Grifoni, P., D’Andrea, A., & Ferri, F. (2012). An integrated framework for on-line viral marketing campaign planning. International Business Research, 6(1), 22.
22 Harvey, C. G., Stewart, D. B., & Ewing, M. T. (2011). Forward or delete: What drives peer‐to‐peer message propagation across social networks? Journal of Consumer Behaviour,
10(6), 365-372.
Ho, J. Y., & Dempsey, M. (2010). Viral marketing: Motivations to forward online content. Journal of Business Research, 63(9), 1000-1006.
Hollenbeck, C. R., & Kaikati, A. M. (2012). Consumers' use of brands to reflect their actual and ideal selves on Facebook. International Journal of Research in Marketing, 29(4), 395-405.
Kaplan, A.M. & Haenlein, M. (2010). Users of the world, unite! the challenges and opportunities of social media. Business Horizons, 53(1), 59-68.
Kilgour, M., Sasser, S. L., & Larke, R. (2015). The social media transformation process: curating content into strategy. Corporate Communications: An International Journal, 20(3), 326-343.
Macnamara, J. & Zerfass, A. (2012). Social Media Communication in Organizations: The Challenges of Balancing Openness, Strategy, and Management. International Journal of
Strategic Communication, 6(4), 287-308.
Mander, J. (2014). Digital vs. Traditional Media Consumption: Summary. Retrieved from Global Web Index website: http://insight.globalwebindex.net/hs-fs/hub/304927/file-1414878665-pdf/Reports/GWI_Media_Consumption_Summary_Q3_2014.pdf
Mills, A. J. (2012). Virality in social media: the SPIN framework. Journal of Public
Affairs, 12(2), 162-169.
Nelson-Field, K., Riebe, E., & Newstead, K. (2013). The emotions that drive viral video.
Australasian Marketing Journal, 21(4), 205-211.
Phelps, J. E., Lewis, R., Mobilio, L., Perry, D., & Raman, N. (2004). Viral marketing or electronic word-of-mouth advertising: Examining consumer responses and motivations to pass along email. Journal of advertising research, 44(04), 333-348.
Sabate, F., Berbegal-Mirabent, J., Cañabate, A., & Lebherz, P. R. (2014). Factors influencing popularity of branded content in Facebook fan pages. European Management
Journal, 32(6), 1001-1011.
Singh, T., Veron-Jackson, L., & Cullinane, J. (2008). Blogging: A new play in your marketing game plan. Business horizons, 51(4), 281-292.
23 Southgate, D., Westoby, N., & Page, G. (2010). Creative determinants of viral video viewing. International Journal of Advertising, 29(3), 349-368.
Zhang, L., Peng, T. Q., Zhang, Y. P., Wang, X. H., & Zhu, J. J. (2014). Content or context: which matters more in information processing on microblogging sites. Computers in
24 Appendix 1. Codebook
Variable Description Values
Category The category of the article on the magazine’s website can be used as a guideline, but choose the category that fits the following description best.
Humans: Articles about humans and their activities and food
Animals: Articles about animals both domesticated and wild are coded as animals (also effects of human behaviour on animals)
Earth: Articles about weather, the planet, and plants are coded as earth (also effects of human behaviour on the earth)
History: Articles about the human past and dinosaurs are coded as history
Science: Articles about scientific research, space and medicine are coded as science Advertorials: As these articles are
sponsored by another brand, articles coded under advertorials on the website will always be coded as advertorials. If a brand is mentioned but the article is not in the advertorial category there is no reason to assume it is sponsored and it will be coded according to its subject.
Contests: Articles about a contest or the results of a contest is coded as contest (often in the news or photography category on the magazine’s website)
Brand: Articles that are mainly about the magazine or the brand are coded as brand (often in the humans category on the magazine’s website) 1 = humans 2 = animals 3 = earth 4 = history 5 = science 6 = advertorials 7 = contests 8= brand Media richness
Media richness is coded on the inclusion of images, links and videos. The images, links and videos have to a part of the article, not a part of the general layout of the website. Articles are coded as high as possible, so if there is a picture and a video, the article will be coded 3.
0 = Articles with text only 1 = Articles with images 2 = Articles with links 3 = Articles with videos
Posting day The day of the week the article was posted. 1 = Monday 2 = Tuesday 3 = Wednesday 4 = Thursday 5 = Friday 6 = Saturday 7 = Sunday Posting time
The time the article was posted. It has been divided up into night (12 p.m. to 5.59 a.m.), morning (6
1 = night 2 = morning
25 a.m. to 11:59 a.m.), afternoon (12 a.m. to 5.59
p.m.) and evening (6 p.m. to 11:59 p.m.)
3 = afternoon 4 = evening Emotion The most prevalent emotion that the articles
evokes, according to the definition of the Oxford dictionary:
Hilarity: Extreme amusement, especially when expressed by laughter.
Inspiration: The process of being mentally stimulated to do or feel something,
especially to do something creative. Astonishment: Great surprise
Exhilaration: A feeling of excitement, happiness, or elation.
Amusement: The state or experience of finding something funny.
Calmness: The state or quality of being free from agitation or strong emotion Surprise: A feeling of mild astonishment
or shock caused by something unexpected. Happiness: The state of being happy. Disgust: A feeling of revulsion or strong
disapproval aroused by something unpleasant or offensive.
Sadness: The condition or quality of being sad.
Shock: A feeling of disturbed surprise resulting from a sudden upsetting event. Anger: A strong feeling of annoyance,
displeasure, or hostility.
Discomfort: A feeling of being anxious or embarrassed.
Boredom: The state of feeling bored. Irritation: The state of feeling annoyed,
impatient, or slightly angry.
Frustration: The feeling of being upset or annoyed as a result of being unable to change or achieve something.
Hilarity, inspiration, astonishment, exhilaration, amusement, calmness, surprise and happiness were coded as positive emotions.
fDisgust, sadness, shock, anger, discomfort, boredom, irritation and frustration were coded as negative emotions.
0 = Negative 1 = Positive
Engagement The number of shares, comments and likes added up as reported by Facebook Analytics
Numerical value = The number of shares + comment + likes
Number of readers
The number of unique visitors to the article’s page as reported by Google Analytics
Numerical value = The number of unique visitors to the page
26 Appendix 2. SPPS output tables
Table 1
Correlations of likes, shares and comments
Likes Shares Comments Likes Pearson Correlation
Significance (2-tailed) N 1 100 .92 .000 100 .88 .000 100
Shares Pearson Correlation Significance (2-tailed) N .92 .000 100 1 100 .86 .000 100
Comments Pearson Correlation Significance (2-tailed) N .88 .000 100 .86 .000 100 1 100 Table 2
ANOVA test of between-subjects effects for category and engagement
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Category Error Total Corrected Total 6780994.13 9055807.16 6780994.13 74681867.58 1130214150 81462861.71 6 1 6 93 100 99 1130165.69 9055897.16 1130165.69 803030.83 1.41 11.277 1.41 .220 .001 .220
27 Table 3
ANOVA test of between-subjects effects for category and number of readers
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Category Error Total Corrected Total 187875745 179924347.20 187875745.20 2153700063 2964876964 2341575808 6 1 6 93 100 99 31312624.19 179924347.20 31312624.19 23158065.19 1.35 7.77 1.35 .242 .006 .242 Table 4
ANOVA test of between-subjects effects for media richness and engagement
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Media Richness Error Total Corrected Total 3633097.36 32730385.69 3633097.36 77829764.35 113021415 81462861.71 2 1 2 97 100 99 1816548.68 32730385.69 1816548.68 802368.71 2.26 40.79 2.26 .109 .000 .109 Table 5
ANOVA test of between-subjects effects for media richness and number of readers
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Media Richness Error Total Corrected Total 13477004.90 622137481.70 13477004.88 2328098803 2964876964 2341575808 2 1 2 97 100 99 6738502.44 622137481.70 6738502.44 24001018.59 .28 25.92 .28 .756 .000 .756
28 Table 6
ANOVA test of between-subjects effects for posting day and engagement
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Posting Day Error Total Corrected Total 3370128.25 27258649.36 3370128.25 78092733.46 113021415 81462861.71 6 1 6 93 100 99 561688.04 27258649.36 561688.04 839706.81 .67 32.46 .67 .675 .000 .675 Table 7
ANOVA test of between-subjects effects for posting day and number of readers
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Posting Day Error Total Corrected Total 66056469.40 470151495.60 66056469.36 2275519339 296487694 2341575808 6 1 6 93 100 99 11009411.56 47015495.60 11009411.56 24467949.88 .45 19.22 .45 .843 .000 .843 Table 8
ANOVA test of between-subjects effects for posting time and engagement
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Posting Time Error Total Corrected Total 2656854.49 27750637.50 2656854.49 78806007.22 11302415 81462861.71 2 1 2 97 100 99 1328427.24 27750637.50 1328427.24 812433.06 1.64 34.16 1.64 .200 .000 .200
29 Table 9
ANOVA test of between-subjects effects for posting time and number of readers
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Posting Time Error Total Corrected Total 86844462.30 551260365.60 86844462.26 2254731346 2964876964 2341575808 2 1 2 97 100 99 43422231.13 551260365.60 43422231.13 23244653.05 1.87 23.72 1.87 .160 .000 .160 Table 10
ANOVA test of between-subjects effects for emotion and engagement
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Emotion Error Total Corrected Total 830879.02 3427570.46 830879.02 80631982.69 113021415 81462861.71 3 1 3 96 100 99 276959.67 3427570.46 276959.67 839916.49 .33 4.08 .33 .804 .046 .804 Table 11
ANOVA test of between-subjects effects for emotion and number of readers
Source Type III Sum of Squares df Mean Square F Significance Corrected Model Intercept Emotion Error Total Corrected Total 21329669.50 58510261.14 21329669.52 2320246138 2964876964 2341575808 3 1 3 96 100 99 7109889.84 58510261.14 7109889.84 24169230.61 .29 2.42 .29 .830 .123 .830
30 Table 12
Relationship of engagement with number of readers
Engagement Number of Readers Pearson Correlation
Significance (2-tailed) N
.72 .000
