Starbucks: Investigating User-Generated
Content Patterns Across Platforms
& Across Countries
Joyce Anna Paulina Broekhaar
11232617
Amsterdam Business School
MSc. Business Administration – Marketing Track Master Thesis – Final version
Supervisor: W.M. van Dolen 27 January 2017
Statement of originality
This document is written by Joyce Anna Paulina Broekhaar, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This article investigates the effect of country differences on cross-platform UGC patterns on three different social media sites. Three contemporary platforms that are currently most popular in terms of consumer usage and marketer interest were selected for this research: Facebook, Instagram and Twitter. A content analysis was performed on 450 UGC posts for a popular, international company (Starbucks). Consumer-produced brand communications that were posted by consumers from countries in which the Starbucks company has a low, medium and high stage of development (or maturity) were analyzed and coded on six UGC content dimensions. The codebook of the research of Smith et al. (2012) was used, after which UGC comparisons could be drawn across the countries and across the social media platforms. This research provides a new and contemporary perspective on how a country differences and different social media channels may influence how brand-related user-generated content is posted online. Furthermore, it provides a critical perspective on the research of Smith et al. (2012) by generalizing and extending their research. A better understanding of how consumers engage with different social media channels across different countries might help marketers that try to co-create their brand on different social media platforms in different countries.
Keywords: User-generated content (UGC), social media, company maturity, Facebook,
Table of contents
Statement of originality ... 1
Abstract ... 2
Introduction ... 4
Literature review ... 8
Theoretical framework ... 8Social media & user-generated content ... 8
Social media platforms ... 12
Cross-country differences ... 15
Hypotheses ... 16
Promotional Self-presentation ... 17
Brand Centrality ... 18
Marketer-directed Communication ... 19
Response to Online Marketer Action ... 20
Factually Informative about the Brand ... 20
Brand Sentiment ... 21
Method and data ... 23
Sampling ... 23
Data collection ... 25
Coding & coding categories ... 27
Results and discussion ... 29
Effects of company’s maturity ... 31
Variability brand-related UGC on the different platforms ... 33
Promotional Self-presentation ... 34
Brand centrality ... 35
Marketer-Directed Communication ... 36
Response to Online Marketer Action ... 37
Factually Informative About the Brand ... 38
Brand Sentiment ... 39
Conclusion ... 42
Theoretical contributions and implications ... 44
Limitations ... 45
Directions for future research ... 46
Acknowledgements ... 48
References ... 49
Introduction
Currently, the concept of Social Media is on top of the agenda for many business executives. Online content is not only posted by companies anymore: Consumers write texts, make photos and videos and share these with other people, and these people in turn react to this content. Much brand-related, user-generated content (UGC) can be found on different social media platforms. This UGC has the potential to shape consumers’ perceptions of brands and even influence consumer behavior (Keller & Libai, 2009). Companies do not only use social media channels for advertisements and promotions, but also to engage with consumers (Solis, 2010). Kaplan & Haenlein (2010) argue that there is considerable diversity among social media channels (blogs, social networking sites, content communities). Also, the social media context is extremely dynamic and complex: social media platforms are ever evolving (Isosomppi, 2015). New platforms arise and existing platforms introduce new features. Much uncertainty exists on how companies should deal with this complexity; managers are uncertain on how they should allocate their marketing tools and resources most effectively among different social media channels (Stelzner, 2016). Which aspects should be considered when determining the most effective social media strategies? Many social media channels have been studied in isolation to make sense of this ambiguity. The study of Smith, Fisscher & Yongjian (2012) has investigated the variability of brand-related UGC among three different social media platforms. However, the dynamic environment of social media and the evolution of the platforms themselves also change the behavioral norms and rules that exist within the different social media platforms over time (Isosomppi, 2015). This results in a constant need for contemporary research of social media platforms, especially for platforms that are high in user engagement and influential in purchase decisions. Understanding how consumers engage with different social media platforms provides managers with insights on how they can co-create their brands with consumers on different platforms. Smith et al. (2012) have made a significant contribution
to research on this topic by investigating cross-platform UGC patterns across Facebook, Twitter and Youtube. Furthermore, they have investigated whether the extent to which a brand is more, versus less, proactive in its social media strategies relates to these cross-platform UGC patterns. Also, they have provided us with a content-dimensions codebook that can be used to replicate and generalize their research. Facebook, Instagram and Twitter are currently the most popular social media platforms in terms of consumer usage and marketer interest (Greenwood, Perrin & Duggan, 2016; Chaffey, 2016). This research will partly replicate and extend the research of Smith et al (2012) for these more-contemporary social media sites by investigating how brand-related UGC differs across them. YouTube is thus ‘replaced’ by Instagram. Furthermore, the research will be conducted for another product category, as different product categories often elicit different consumer responses (Munoz, Chambers, Hummer, 1994).
As Smith et al. (2012) have already indicated, the degree to which a brand is proactively managed on social media plays a role in how brand-related UGC is posted across countries. It could be expected that other moderators also exist. Research found that consumers do not respond to brands in general ways on different social media platforms (Wilson, Murphy, Fierro, 2012). For traditional advertising, much research has already focused on differences in consumer responses across countries and overwhelming support for impact of these differences has been reported (e.g. Tansey, Hyman, & Zinkhan, 1990; Alden, Hoyer & Lee, 1993). It would be interesting to see whether these cross-country effects also exist for consumer responses on the Web. Therefore, this article will also investigate the effect of country differences (“company’s maturity”) on cross-platform UGC patterns.
Therefore, the two main research questions of this article are:
1. How does brand-related UGC differ across Facebook, Instagram and Twitter?
2. Do these cross-platform UGC patterns also hold for UGC from countries in which a company is more, versus less, mature?
To answer these research questions, a content analysis was performed on 450 UGC posts for a popular, international company. This was done by analyzing consumer-produced brand communications that were posted by consumers from a low-, medium- and high-maturity country in which that company operates. The codebook with six content dimensions of the research of Smith et al. (2012) was used, after which comparisons could be drawn across the countries and across the social media platforms. By using this codebook, our results could be (partly) compared with the research of Smith et al. (2012), providing a critical perspective on whether their findings are also generalizable over time, and generalizable for another product category.
This research has two main contributions: 1. It generalizes and extends the research of Smith et al. (2012) by partly repeating their research with more contemporary social media platforms in another, broader product category. This improves our understanding of how users engage with these channels. 2. It provides a new and critical perspective by highlighting similarities and differences in cross-platform UGC patterns for countries in which a company is less, versus more, mature.
By providing these insights, it can inform marketers that are concerned with co-creating their brand with consumers on different social media platforms.
The rest of this article will be structured as follows; Firstly, this article will review previous literature on social media, user-generated content, the selected social media channels and cross-country differences. Next, the content dimensions will be discussed in more detail,
and accompanying hypothesis will be presented. Furthermore, our methodology and data collection will be explained, followed by an overview of the results of our content and statistical analyses. Finally, the results will be discussed, and this article will close with a conclusion, including limitations and direction for further research.
Literature review
Theoretical framework
Social media & user-generated content
According to Kaplan & Haenlein (2010, p. 61), social media can be defined as “a group of Internet-based applications that build on the ideological and technological foundations of Web 2.0, and that allow the creation and exchange of User Generated Content (UGC)”. User-generated content is content that is published on either on a website that is publicly accessible, or on a social media platform that is at least accessible to a certain number of people. Furthermore, UGC is “created outside of professional routines and practices” (Kaplan & Haenlein, 2010; OECD 2007, p. 61). Companies are not the only ones that post content online: they are not in control anymore. Consumers write texts, make photos and videos and share these with other people, and these people in turn react to this content. This social interaction is a typical characteristic of social media. Boyd and Ellison (2008) argue that consumers use UGC to express themselves and to communicate with others. UGC can be compared with electronic word-of-mouth (eWOM), but it is not identical to this concept. Hennig-Thureau, Gwinner and Gremier (2004, p. 39) define eWOM as “any positive or negative statement made by potential, actual or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet”. Research has already shown that word of mouth has a significant impact on consumer behavior. Keller & Libai (2009) found that it generates more than 3.3billion brand impressions every day. Furthermore, word of mouth is the primary factor between 20-50% of all purchasing decisions, and it generates more than double the sales of paid advertising (Bughin, Doogan & Vetvik, 2010). In general, UGC is broader than eWOM, but when UGC is brand-related as it is in our research, the two overlap considerably. Companies do not only talk tó consumers, but they also talk wíth them. Consumers want to talk with them and be heard. Thus, social media are not only used for advertisements and promotions, but also
for companies to come up with innovative ideas, handle customer service issues and to engage with their customers (Solis, 2010). An examination of previous literature also suggests that there are meaningful differences between traditional media platforms and social media platforms. Luo et al. (2013) and Kumar et al. (2013) argue that these points of differences exist in the strength of the relationship that is formed with the consumer, the measurement of return on investment, and in the level of customer influence. Therefore, the traffic on traditional media platforms goes one-way (from the firm to the consumer), while social media platforms allow consumers, firms and other consumers to constantly and instantly interact.
Currently, 2.2 billion people are using social media, and this number is expected to grow to 2.7 billion in 2019 (Statista, 2016). Therefore, social media is something that many companies cannot simply neglect. Research has shown that UGC on social media platforms affects consumer choices (Trusov et al., 2009; Rui et al. 2013; Hennig-Thureau, Wiertz & Feldhaus, 2014) that it relates to significant managerial outcomes such as sales (Dhar & Chang, 2009; Ghose & Ipeirotis, 2010). In 2008, Christodoulides found that more and more branded companies were aiming to participate actively in these social media platforms, instead of only monitoring them. And this was proven to be right: In 2008, only 42% of companies used social media for marketing (Williamson, 2010), while this has grown to over 90% in 2016 (Stelzner, 2016). Marketers are devoting larger shares of their marketing budget to social media strategies. The CMO survey of 2015 discovered that CMOs expect that they will spend 10.7% of their marketing budget on social media in the next year, and that they expect that amount to rise to 23.8% over the upcoming 5 years (CMOsurvey.org 2015). In a survey of marketing professionals, Stelzner (2016) found that the vast majority of marketing professionals (90%) believes that social media is even crucial for their businesses. However, most of these marketers (92%) were unsure about how they should utilize their social media strategies in the most effective way, in order to engage their customers and to earn a return on investment. Hoffman
and Fodor (2010) also argue that marketers insufficiently understand how value creation may be different on different social media platforms, and that this causes much ambiguity on how they should utilize different social media platforms.
In order to understand these different platforms better, many social media channels have been studied in isolation. However, not many articles have investigated multiple social media channels in one study. Cheong and Morrison (2008) investigated the perceptions of brand-related UGC versus the types of content that were generated on YouTube, blogs and forums, but they did not consider the variability of UGC across these sites. Smith et al. (2012) have incorporated multiple social media channels into one single study for comparative purposes, to enlarge our understanding of the variability between social media. They compare how users engage with three different social media platforms: Twitter (microblogging site), Facebook (social network) and YouTube (content community). They do so by looking at how brand-related, user-generated content (UGC) is posted across these sites by comparing UGC on six dimensions: Promotional self-presentation, brand centrality, marketer-directed communication, response to online marketer action, factual informative about the brand and brand sentiment. This research has investigated and proven that variability in UGC among Facebook, Twitter and YouTube exists, at least for two US-based retails brands. Furthermore, they also argue that these findings are not robust; they found this variability to differ for brands that were less, versus more, proactive in their social media strategies.
Social media use and marketing are always evolving. The report of Chaffey (2016) identifies Facebook, Twitter and Instagram as the current most popular social media sites used worldwide in terms of size, penetration and engagement. The report of Greenwood et al. (2016) confirms these findings. According to the study of Stelzner (2016), marketing professionals also cited Instagram among the most used and most important platforms for marketers to consider at this point in time. Currently, 44% of marketers use Instagram for their social media
activities, while this share was 36% in 2015 and only 28% in 2014. The importance of Instagram is even more visible in business-to-consumers (B2C) markets. In this market, 51% of marketers use Instagram in their social media strategy. When B2C marketers were asked to list only one platform as their number one choice in importance, Instagram was even surpassing YouTube. Equally to the previous year, Facebook remains by far the most used platform with 93% of marketers using Facebook in their social media strategy in 2016. However, all other most commonly used platforms – Twitter, LinkedIn, YouTube and Google+ - are decreasing in importance compared to the previous year. Twitter remains to be the second most used social media platform with 76% of marketers running some sort of promotion on it (Stelzner, 2016). Even though their smaller size, Instagram and Twitter have proven to be more influential than Facebook in purchase decisions of teens and young adults (Piper Jaffray, 2013). Instagram even appeared to be the most influential. Krallman et al. (2016) investigated the effect of the social media platforms Facebook, Instagram and Twitter on intentions for consumer co-creation and usage, and they found that consumers mainly use Instagram for entertainment purposes, followed by Twitter and Facebook respectively. Regarding information purposes, Twitter was the most used platform. Also, for social interaction purposes, Instagram and Twitter scored highest. For brand involvement and social media co-creation, Instagram scored highest, followed by Facebook and Twitter respectively. Krallman et al. (2016) underline the importance of considering Instagram in future social media research. In should be noted that their results are not in line with the results of the research of Smith et al. (2012); These results will be discussed in more detail in the next chapter. This may indicate that the research of Smith et al. (2012) is becoming outdated, which would not be surprising since social media platforms are constantly emerging and evolving (Isosomppi, 2015). This calls for the need of investigating more contemporary social media platforms; social media channels that are currently most popular among both consumers and marketers, and have the highest user penetration and
purchase influence. Therefore, this research will do so by party replicating and confirming the research of Smith et al. (2012) by investigating the variability of UGC on Facebook and Twitter and extending this research by adding the social media platform Instagram to this analysis. Furthermore, this research will be conducted in another, broader product category. As Munoz et al. (1994) indicate, different product categories often elicit different consumer responses. This was also argued by Smith et al. (2012): they call for the replication of their study in a broader product category, to see whether their results hold for other product categories. Therefore, I choose to conduct this research in another product category. A detailed description and review of the investigated social media platforms will follow in the next paragraph.
Social media platforms
Facebook was founded in 2004 and currently it is by far the most popular social media platform for both marketers and consumers (Stelzner, 2016). On its Newsroom website (2016), Facebook states that its mission is to “give people the power to share and make the world more open and connected”. Currently, Facebook has nearly 1.8 billion active users of which 1.18 billion log onto the platform every day. According to Facebook, its users use the platform to connect with their friends and family, to see what is happening in the world and to share and express what matters to them. It is a social networking site in which people create their personal profile that features all kinds of personal information: personal details, photos, interests, (attended) events, and pages that people ‘like’. All kinds of content can be shared on this platform: photos, videos, status/text, links and events. People can connect to (“friend”) other people that are also using the platform. Furthermore, people can engage in all kinds of activities. For example, they can post messages on other people’s pages (“walls”), they can like brands or other authorities, they can comment on links and they can participate in discussions. All of this allows people to build and maintain social relations: people can communicate with others and keep up with each
other’s life. Therefore, Facebook pages can be seen as online word of mouth platforms, because they enable customer initiated social interactions. Word of mouth is generated by interactions on Facebook pages, but also by commenting, liking and sharing content that is brand-related (Berger & Schwartz, 2011; Hennig-Thurau et al., 2004; Woinicki & Godes, 2008). Previous research has investigated the platform’s norms and functionality (Paparcharissi, 2009), consumer’s usage motivations (Krallman et al. 2016; Debatin et al. 2009; Ellison et al, 2007) and self-presentation issues (Labrecque et al, 2011; Papacharissi, 2009; Zywica & Danowski, 2008; Tom Tong et al, 2008). According to a CMO study in 2016 (CMOstudy.org 2016), the highest shares of social media spending is dedicated to Facebook. When we look at the size of the platform, this may not come as a surprise. However, firms actually know very little about the value of interacting with consumers through Facebook. Stelzner (2016) found that many marketers are unsure about their marketing efforts on Facebook: 35% of marketers are not even sure whether their Facebook marketing efforts are effective.
Instagram is a relatively young social media platform that was launched in October 2010. It is owned by Facebook Inc. and has over 600 million monthly active users (Statista 2016). It started as a simple photo-sharing application among friends, and currently it has grown to a global community of brands and consumers. Instagram allows people to post, view like and comment on photos and short videos that are posted on this platform. People can easily share updates by uploading a picture or video, editing this picture or video (e.g. with a filter) and sharing it with their followers. Also, other people can be tagged in this content, and a geolocation can be attached. People can add captions to the pictures, in which hashtags (#) and mentions (@) can be used. With these hashtags, other people can find Instagram posts that are related to the specific hashtag more easily. By using a mention, people can link their post to the referenced user account. Content can be watched, liked and commented on via both mobile and desktop devices, but content can only be shared with mobile devices. Since it is relatively young
social media channel that has experienced rapid growth in the last couple of years, literature regarding its norms and functionality is scarce. However, Krallman et al. (2016) have investigated consumers’ motivations to engage with this platform. They find a major involvement and co-creation potential for this platform (even bigger than for Facebook), and they underline the importance for marketers to also focus significantly on Instagram in their social media efforts.
Twitter is a micro-blogging site, founded in 2006, on which users can posts so-called “tweets”: posts that do not exceed 140 characteristics in length. According to Kaplan & Haenlein (2011), microblogging is a contemporary phenomenon that refers to a person broadcasting brief messages to some or all members of its social network through a specific web-based service. Tweets can include hyperlinks to other websites, blogs, pictures, etc., and often they ask for the sharing of information, opinions, complaints or daily life details. People can follow other users or companies, by which they allow the posts of the followed account to show up in their stream. Most posts are publicly visible. Besides posting tweets, users can reply to and forward (“retweet”) other member’s tweets. Currently, Twitter has 313 million monthly active users (Twitter Usage Facts, 2016). Previous research has already investigated norms and behaviors on Twitter (Boyd et al, 2010), the way people present themselves (Marwick & Boyd, 2011) and the motivations (what and why) for posting on Twitter (Krallman et al. 2016; Jansen et al. 2009; Java et al. 2009; Naaman et al. 2010). Hennig-Thureau et al. (2014) conducted their research on Twitter, investigating the effect of microblogging word of mouth (MWOM) on the early adoption of new products. They found proof for this effect, indicating that sharing post-purchase quality perceptions on Twitter had an influence on sales. When looking at specific brand-related research on Twitter, it was found that approximately 19% of tweets are related to brands (Jansen et al. 2009). For nearly half of these tweets the brand is the primary focus. In these tweets, users especially share their opinions and seek information about the brand.
As the above review shows, Facebook, Instagram and Twitter represent different types of social media platforms. Each of these platforms has its own characteristics with its own unique architecture. Users have different motivations for visiting and using these sites. The content that is produced on these sites appears to be quite unique for each site, and people interact in different ways on each platform. As was already mentioned before, one of the main contributions of this article is that it will investigate the variability of brand-related UGC across Facebook, Twitter and Instagram.
Cross-country differences
For traditional advertising, much research has been done on differences in consumer responses across countries and overwhelming support for impact of these differences has been reported (e.g. Tansey et al., 1990; Alden et al., 1993). Usunier (1993) found many contrasting consumer attitudes towards different advertising strategies across countries, such as attitudes toward comparative advertising. Albers-Miller (1996) argues that for traditional advertisements, local idiosyncrasies call for different advertisement strategies since a standardized approach does not have the same effect, not even on a regional basis. However, it is difficult to test for differences across countries and cultures because we do not even know what we know: many elements that are used to compare countries or cultures are not even recognized without a basis of comparison (Hall, 1976).
Regarding social media, Wilson, Murphy and Fierro (2012) have already argued that people are not responding to brands in a universal way on different social media platforms. However, the effect of country differences on social media strategies has never been investigated to the best of our knowledge. Literature on this topic is very scarce. However, it appears that many companies have difficulties with finding the most appropriate and suitable social media strategies across different regions, because simply implementing a
one-size-fits-all strategy in different countries does not always pay off. In some regions, certain strategies appear to be more effective than in other regions. One way to analyze these social media strategies is by investigating cross-platform UGC patterns on social media for different countries. A large MNE can experience different “stages of development” in different countries. In some countries, it is more developed / mature than in other countries, where it might be in the early development stage (e.g. because it just entered that specific country). Could it be that these differences have an influence on the way that brand-related UGC is posted across different social media channels? Literature on a firm’s maturity is scarce and it has not been investigated before whether this maturity affects online consumer expressions like brand-related UGC. It would be interesting to check whether these differences influence what and how people share, since these differences appeared to influence the effect of traditional advertising strategies. Therefore, the other major contribution of this research is that it will check whether certain cross-platform UGC patterns differ across countries in which a company is less, versus more, mature.
Hypotheses
To investigate whether and how brand-related user generated content varies across Facebook, Instagram and Twitter, the content dimensions of the research of Smith et al. (2012) can be used: Promotional self-presentation, brand centrality, marketer-directed communication, response to online marketer action, factually informative about the brand and brand sentiment. According to Smith et al. (2012), these dimensions are all present in brand-related UGC. All dimensions, except for brand sentiment, are binary. Brand-related UGC posts can be analyzed for each specific dimension, after which comparisons can be made across social media platforms, but also across countries. In the next chapter, a description will follow of each dimension and related hypotheses regarding variability of UGC among social media platforms
will be presented. These hypotheses recognize the technical and the cultural aspects of the three social media platforms that may have an influence on the production of brand-related UGC. However, it should be mentioned that even though the technical aspects of each site may help shape the unique culture and use of each site, still quite similar content can be posted on all sites. Hypotheses regarding differences in a company’s maturity will not be presented due to a lack of theoretical basis for doing so; investigating these differences will be done inductively rather than deductively. The research of Albers-Miller (1996) shows that much other cross-country and cross-cultural research does not have directional hypothesis either, because of the complexity and versatility of these concepts.
Promotional Self-presentation
According to Zywica and Danowski (2008), self-presentation can be seen as making an effort to express a specific identity and image to others. Consumers construct their image by using possessions, brand and other symbols. This can be done in both the online and offline context (Belk, 1988; Schau & Gilly, 2003). Marwick and Boyd (2011) argue that explicit and deliberate self-presentation may create tensions on some platforms, depending on the culture of the platform. Twitter, for example, is recognized for hosting posts on the details of the user’s personal life, but self-promotion was often considered to be inappropriate. Kietzman et al. (2011) agree on these findings by arguing that Twitter is more used to promote conversations than to promote the self. On Facebook, personal profiles and user generated content are linked to the self-presentation concept (Zywica & Danowski, 2008). Papacharissi (2009) proposes that users’ personal profiles (e.g. likes) and “display of friends” are the biggest means by which self-presentation occurs, and this appears to be done more easily on Facebook than on Twitter. Smith et al. (2012) found that UGC on YouTube was significantly more often used to promote the self than Facebook or Twitter. This is not surprising, considering YouTube’s slogan
‘broadcast yourself’. We hypothesize that UGC on Facebook and Twitter would be quite similar and low in featuring self-promotion. To the best of our knowledge, Instagram has not been investigated regarding self-promotion purposes. In the research of Krallman et al. (2016), Instagram was found to be mostly used for entertainment and social interaction purposes. It therefore is likely that some degree of self-presentation would be appropriate on this site, but connecting to other users is also important. Therefore, we posit:
H1. Brand-related UGC on Instagram is more likely than brand-related UGC on Facebook and Twitter to feature consumer self-promotion.
Brand Centrality
This dimension refers to the brand’s role in brand-related UGC. In other words: is the brand central or peripheral to the message? When it is central to the message, it is the focus of the post. When it is peripheral to the message, it has more of a supporting role. Previous research has not discussed brand centrality in UGC extensively, but it often assumes it. Jansen et al. (2009) suggest that brand centrality may vary across content. Smith et al. (2012) found that brands were likely to be more central in brand-related UGC on Twitter than on Facebook. Considering the fact that a tweet has a 140-character limit, this is not surprising. It is difficult to introduce multiple topics with so few words. Further, Smith et al. (2012) argue that Twitter’s technical design and focus on discussion facilitation and information sharing privileges brand centrality. Since Instagram is mainly a photo-sharing application, there is only a small focus on the accompanying text. Most pictures contain only small amounts of text, leaving little space for the introduction of multiple topics. Facebook is more focused towards the facilitation of social connectedness, and therefore Instagram and Twitter are proposed to feature a greater percentage of brand-central posts than Facebook posts.
H2. Brands are more likely to be central in brand-related UGC on Instagram and Twitter than in brand-related UGC on Facebook.
Marketer-directed Communication
Deighton and Kornfeld (2009) illustrated that many social media sites allow consumers to communicate with marketers. Companies often have different social media channels, and these channels allow consumers to pose questions or complaints to marketers. Also, they allow consumers to respond to questions or comments that companies post on these platforms. Facebook allows consumers to post questions to marketers on their Facebook brand pages, which is not possible on Twitter or Instagram. Furthermore, it is possible to reply to specific marketer generated posts on Facebook, and marketers can in turn reply to these replies. Twitter users are especially recognized for asking for information and complaining about things on the platform (Krallman et al. 2016; Naaman et al, 2010). The research of Krallman et al. (2016) has proven that information seeking motivation is higher on Twitter than on Instagram. Furthermore, users can easily send out a tweet that is directed towards a marketer, and marketers can easily reply to these tweets (as was also the case for Facebook). On Instagram, users are only able to reply to specific marketer posts. However, marketers cannot easily reply to these replies (this is not a supported feature on the platform, as it is on Facebook and Twitter). Another possibility for consumers is to mention marketers in their own posts with accompanying picture. We expect this to be a barrier to many users, and that it is highly unlikely for consumers to do this. Therefore, we hypothesize brand-related UGC on Facebook and Twitter to be more directed towards marketers than UGC on Instagram.
H3. Marketer-directed brand-related UGC is most likely to be posted on Twitter and Facebook and least likely to be posted on Instagram.
Response to Online Marketer Action
This content dimension looks at whether brand-related UGC posts are a response to an action of an online marketer. This is not only when consumers reply to questions that were posed by marketers, but also responses to promotions (e.g. coupons, events) are included. These responses also do not necessarily have to be directed to the marketers, but they can also be directed to other users. Consumers can easily follow brands on each platform: on Facebook they can ‘like’ the brand and on Twitter and Instagram they can follow the brand. By doing so, they will be confronted with online marketer actions that appear in their feed. As Smith et al. (2012) already mention, responding to this content is relatively easy on Facebook and Twitter. However, on Instagram it is more difficult to respond to online marketer content. Users’ own posts are not very likely to be a response to online marketers; We think this might again be a high barrier for people. Furthermore, when people reply to a specific marketer-generated post, it is not possible for other people (and marketers) to reply to this specific reply again. This interaction barrier does not make it attractive for people to respond to online marketer posts. Thus, we posit:
H4. Brand-related UGC is least likely to be posted in response to an online marketer action on Instagram.
Factually Informative about the Brand
In UGC, users might mention brands for different reasons like giving their opinion, complaining or talking about it as an object of interest. Another reason could be that users want to share
information about the brand. This can be investigated by analyzing whether a UGC post contains factual information about the brand. Examples of factual information are a store’s location, the color of a specific item, the timing of a sale and the price of a product. Smith et al. (2012) found that brand-related UGC was more likely to feature factual information on Twitter than on Facebook. Krallman et al. (2016) found that people mostly used Twitter for information purposes; More than they use Facebook or Instagram for that same purpose. People often use Twitter to share information and news. Smith et al. (2012) found that more user-generated postings on Facebook contain factual information than user-generated postings on Twitter. Facebook is mostly used for social interactions and connections, but these connections can be formed and fortified by sharing information about brands. Furthermore, marketers post much information about new products, events and coupons on Facebook, providing consumers with resources that they can draw on in their conversations. Instagram is more often used for entertainment purposes, so we do not expect brand-related UGC on Instagram to contain much factual information.
H5. Brand-related UGC is least likely to feature brand-related factual information on Instagram.
Brand Sentiment
Brand sentiment is also called valence. Hoffman and Fodor (2010) argue that this is a popular measure for marketers when they want to evaluate the success of their social media initiatives. Brand sentiment can be measured by differentiated it as positive, negative, neutral or unclear. In their research, Smith et al. (2012) did not only find that brand sentiment was different for each platform, but they also found that the pattern was different for the two brands. They conclude that it is safest to argue that brand sentiment is not predictably different across sites. However, since Instagram is mostly used for entertainment purposes, we hypothesize brand
sentiment in brand-related UGC on Instagram to be mostly positive and least negative than brand-related UGC on Facebook and Twitter.
H6. Sentiment towards brands in brand-related UGC is most likely to be positive on Instagram than on Facebook or Twitter.
H7. Sentiment towards brands in brand-related UGC is least likely to be negative on Instagram than on Facebook or Twitter.
Method and data
This research is of descriptive and exploratory nature. A qualitative study, by means of a social media content analysis, was performed. To be more specific, this was a content analysis of brand-related user-generated posts on different social media platforms for one company that operates in multiple countries. Content analysis allows us to identify the notions and relations that would define the data collection (Yildirim & Simsek, 2004). Kolbe and Burnett (1991) argue that content analysis is a reliable method for systematically comparing content of communications. Besides that, it offers an objective way of comparing content for a large amount/sample of UGC. This makes it an appropriate and suitable method for this research. The overall design of study is a combination of social media research (looking at different social media posts) and a case study (specific in the case of the selected company).
Sampling
The unit of analysis are individual brand-related UGC postings on social media platforms. Postings were considered to be brand-related when the brand was mentioned or displayed in the post. As mentioned before, the focus of this research will be on three different types of social media: Facebook, Instagram and Twitter.
As our research objective is to investigate the potential effect of a company’s maturity on the UGC that is posted about that brand, we selected a brand that is popular and that has different stages of development in different countries: Starbucks. Starbucks is an American coffee company and coffeehouse chain that was founded in 1971. Currently, it operates in 67 countries all over the world with a total number of stores of 22.519 (Starbucks Company Profile, 2015), selling a broad range of products, varying from coffee beverages, smoothies, tea, baked goods and sandwiches to items that are related to the coffee-making and drinking process, like equipment (espresso machines, grinders), drinkware and syrups. It is known for its coffee
quality, taste and customer experience (Sacks, 2014). Starbucks manages its social media slightly differently across countries; It has a general Starbucks account on most social media platforms, which is often relevant for people from different countries because these accounts post general product advertisements, information, etc. online. Furthermore, some countries have their ‘own’ Starbucks account, posting (e.g.) promotions for that specific country. Starbucks does not have an account for each country on each platform. However, the Starbucks’ country accounts that do exist mainly operate in a similar way: posting pictures about new product launches, promotions, etc. Significant UGC about Starbucks can be found on all kinds of social media. Moreover, it thus operates in countries all over the world, representing large differences in number of stores per country. Starbucks posts these numbers online, so these data are accessible (Starbucks Investor Relations. Supplemental Financial Data, 2016). This makes Starbucks an appropriate company to study in this research.
This research looks at the effects of differences in a company’s maturity (or “stage of development”) on variability of UGC on Facebook, Instagram and Twitter. These differences were measured on a country-scale. Heterogeneous countries were selected, so that the possible effects of differences are clearly visible. Differences in maturity for Starbucks are addressed by looking at how ‘developed’ Starbucks is in a specific country: A ‘Maturity Index’ variable was created by analyzing Starbucks’ current number of stores per country (both company-operated and licensed) and the population of that specific country. These data can be found on the Starbucks website (Starbucks Investor Relations, 2016) and via Eurostat (2017). According to Albers-Miller (1996), taking specific country information (as is done here) to make cross-country comparisons is an often-used method in previous literature. The total population of a country can be divided by the total number of stores in that same country which gives us an estimated, average number of people per store. Selected countries for this research are France (low maturity), the Netherlands (medium maturity) and the United Kingdom (high maturity),
because these countries are highly heterogeneous regarding Starbucks’ maturity. The data from these countries can be found in table 1.
As mentioned before, the extent to which social media is managed for each specific country can also be analyzed by looking at ‘national’ Starbucks social media accounts. It was already mentioned that Starbucks has a “general” Starbucks account for Facebook, Twitter and Instagram (language is English) which can be followed by people from all countries. However, it is also interesting to check whether Starbucks also has own high-traffic platform pages for the selected countries on these social media platforms. We find that Starbucks France has its own Facebook, Instagram and Twitter account, and so does Starbucks UK. However, Starbucks the Netherlands only has its ‘own’ Facebook page. When looking at the content that is posted among these accounts, we do not find significant differences: many posts are mainly the same, but with translated texts. Therefore, we argue that the marketer-generated content that is ‘pushed’ from these accounts does not differ significantly. The number of followers for each account can be found in appendix 1.
Data collection
Data was collected by scraping posts on Facebook, Instagram and Twitter that either contained the word “Starbucks” or featured the brand. A web scraper (R) and advanced search engines (e.g. Google Advanced Search, Twitter Advanced Search) were used to get a complete sample
of UGC posts on these platforms. Only posts from the selected countries (United Kingdom, France and the Netherlands) were scraped by using the geo-location of posts when scraping and searching. Only posts between the 1st of September 2016 and the 15th of January 2017 were scraped to ensure that the data was recent and thus more relevant to current marketers. A random sample of 450 posts was taken from this extensive dataset (around 15.000 posts) to keep the sampling scope manageable. This was done by systematically selecting the 30th post. Since an enormous amount of UGC about Starbucks is posted every day, we choose to take a random sample in this specific, restricted time frame. Otherwise, by taking the most recent number of posts, we would only get posts from the last couple of days. This could easily bias the results. Furthermore, all posts were analyzed to ensure that they were indeed posted by consumers of the selected countries and that they did not have an apparent commercial objective. For example, some posts that contained the hashtag #tobeapartner were found, indicating that they were posted by Starbucks employees. If a post was not considered to be a user-generated post, another post was randomly selected. 50 posts for each country on each platform were obtained and further analyzed (appendix 2), resulting in a final sample of 450 posts. These posts represented the complete range of UGC that can be found on each platform; For Twitter, it included tweets, replies and retweets. For Instagram, both replies and personal wall updates were captured. Lastly, for Facebook, it included wall posts, forum discussions, photos and status updates.
Before the actual analysis, the dataset was organized in a SPSS database. This database contained the following variables: Social media platform (Facebook/ Instagram/ Twitter), caption (text of the post), geolocation (France/ the Netherlands/ United Kingdom), username of sender (optional), the date of creation (optional) and the link to the post (optional).
Coding & coding categories
All posts were analyzed and coded. According to Lockyer (2004), coding facilitates the organization, retrieval, and interpretation of data. In turn, the basis of these interpretations lead to conclusions. This time-consuming coding task was done manually, and was accomplished by two different coders, including the author (the other coder had no knowledge about the research objective). The coders analyzed the complete post (including picture, link, attached location, etc. if these were available) and coded that specific post regarding its content on six dimensions. These UGC dimension categories were drawn from the framework created by Smith, Fisher & Yongjian (2012). Smith et al. created this set of content dimensions based on prior literature and on a preliminary, inductive analyses of UGC. Using this codebook allows us to repeat and validate part of their research, but in turn it also allows our research to be repeated and validated. Further, it makes our methods transparent because it records the analytical thinking that is used to devise codes, and it allows our study to be compared with other studies. The coding dimensions, its descriptions and its values are presented in table 2. All dimensions are binary, except for the last dimension (“brand sentiment”). Examples of how posts were coded on these dimensions can be found in Appendix 3. Inter-coder reliability was ensured by running a Cohen’s Kappa test for the complete sample for each dimension after all posts had been coded. The kappa statistic values for each dimension lie between the accepted range of 0.81 and 1.0 of Landis & Koch (1977), indicating an almost perfect strength of agreement. The kappa value for each dimension can be found in appendix 4. Discrepancies in coding were examined and adjusted by an independent third coder, resulting in a final database with one value for each post on each dimension. After coding, the category frequencies were tabulated. Next, statistical differences were investigated by using Chi-square tests. This will be discussed in the next chapter.
Results and discussion
To overcome the limitations of only using one type of analysis (i.e. qualitative analysis), statistical tests were performed on the data. But first, category frequencies were tabulated for the total dataset and for the UGC posts of each country separately. These frequencies can be found in tables 3 (total), 4 (France), 5 (the Netherlands) and 6 (United Kingdom). When analyzing the table with the frequencies for the total dataset, we can find some interesting similarities and differences across the dimensions, platforms and countries.
Effects of company’s maturity
As one of the research questions and main contributions of this research relates to investigates whether the variability in brand-related UGC differs across countries, we first perform Poisson Regression tests to check for the existence of country * platform interactions1. This was done
to see whether there are significant differences among the datasets of the low-, medium- and high-maturity countries in the variability of brand-related UGC across social media platforms. A Poisson Regression test was performed for each content dimension, with the content dimension as the response variable. The outcomes of these tests can be found in table 7. No
1 For the dimension ‘promotional self-presentation’, a small constant was added to each cell for the
calculation of interaction effects. This was done to address convergence difficulties regarding ‘0’ frequency counts in some cells for the Poisson regression test.
statistically significant differences were found for the dimensions ‘brand centrality’ (2.846, 4df,
p<0.584), ‘marketer-directed communication’ (1.763, 4df, p<0.414), ‘factually informative’
(0.554, 4df, p<0.968) and ‘brand sentiment’ (5.928, 4df, p<0.205); This indicates that there are no statistically significant differences in the variability of brand-related UGC on Facebook, Instagram and Twitter. Therefore, these data of these countries could be combined and further analyzed as a ‘total’ database. The variability of brand-related UGC across the selected social media platforms was argued to be universal for these dimensions. For the dimension ‘promotional self-presentation’ (9.642, 4df, p<0.047), statistically significant evidence was found for the existence of cross-country differences (at the 5% significance level). Moreover, for the dimension ‘Response to Online Marketer Action’ (4.719, 2df, p<0.094), statistical significant evidence for these differences was found (at the 10% significance level). For these dimensions, it can thus be argued that the variability of brand-related UGC on Facebook, Instagram and Twitter is not universal across countries, and it would be inappropriate to combine the datasets of these countries for further analysis. For these dimensions, tests on how brand-related UGC differs across the social media platforms will be run separately for each country. It should be noted that I choose to also include this latter dimension in our further inter-country analysis (‘Response to Online Marketer Action’) even though it has a significance level of 10%, which is above the accepted 5% significance (threshold) level that was maintained in the rest of this article. This is because I think it could be interesting to investigate where these country differences lie. However, this 10% significance level should be kept in mind when reading and assessing the rest of this research.
Next, we will present and discuss the findings for each dimension separately by integrating findings about platform differences and/or similarities and findings about whether country differences existed. Our hypotheses will be tested for each dimension regarding whether brand-related UGC differs across Facebook, Instagram and Twitter. Since no
hypotheses were formed regarding the possible effects of company’s maturity, an inductive approach was taken in analyzing and presenting similarities and differences for this moderator. For a summary of the findings regarding the hypotheses of this article, please refer to table 9.
Variability brand-related UGC on the different platforms
First, Chi-square tests were performed across the three platforms for each dimension separately. This way, significant differences among social media channels could be statistically proven for each dimension, giving us an idea on where we can expect interesting differences and similarities. In case these Chi-square test yielded a significant statistic (p ≤ 0.05), additional tests were run across the three social media platforms. This way, specific site relationships that were contributing significantly to that statistic could be established (e.g. Facebook-Instagram or Instagram-Twitter). In case the Chi-square statistic was not significant, we argue that the brand-related UGC is quite similar across the three social media sites, and for that specific dimension no further analysis was carried out. A summary of the test results can be found in table 8.
Please note that in the following paragraphs three percentages will be often presented within brackets. The first percentage refers to the data for France (low maturity), the second percentage refers to the data of the Netherlands (medium maturity), and the last percentage refers to the data of the United Kingdom (high maturity).
Promotional Self-presentation
The Poisson Regression test indicated that it was inappropriate to collapse the data of the three countries for further analysis for this dimension, indicating that significant differences exist for this dimension among low-, medium- and high-maturity countries. Hypothesis 1 proposes that promotional-self presentation was expected to be higher in brand-related UGC on Instagram (8%, 6%, 16%) than on Facebook (20%, 0%, 0%) and Twitter (0%, 6%, 0%). The initial Chi-square test indicated that there were indeed significant differences among the platforms for the United Kingdom, but not for the data of France and the Netherlands (p<0.000). For the data of the United Kingdom, the hypothesis was supported (p<0.003): for these data, promotional-self presentation was featured mostly in UGC on Instagram compared to UGC on Facebook and Twitter. For the data of France and the Netherlands, no statistically significant differences regarding promotional self-presentation were found among the selected platforms. Promotional-self presentation was consistent and low among all social media platforms. The research of Smith et al. (2012) presented different findings; They found that brand-related UGC on Facebook was more likely to feature promotional self-presentation than UGC on Twitter. However, this finding was not supported in our research for either UGC from a low-, medium or high-maturity country. Apparently, these findings do not hold. This could for example be due to the fact that our research was conducted in another product category, or because users engage differently with Facebook and Twitter after 6 years; the platforms could have evolved (e.g. due to the introduction of additional features), which changed the norms and rules and thus
attitudes and behavior for this platform (Isosomppi, 2015). When comparing our findings with the findings of Smith et al. (2012), we think it is safest to conclude that promotional self-presentation is not predictably different across sites and across countries in which companies have different maturity.
Brand centrality
For brand-centrality, no statistically significant differences appeared to exist in the way that UGC about Starbucks was posted on Facebook, Instagram and Twitter in low-, medium- and high maturity countries. It appeared that differences in Starbuck’s maturity did not affect the variability UGC that highlights the brand centrality across Facebook, Instagram and Twitter. Therefore, we could combine the data of the different countries and analyze the total dataset. Significant differences between brand-central UGC on Facebook (76%, 72%, 68%), Instagram (82%, 88%, 92%) and Twitter (80%, 64%, 90%) were found (p<0.023). Consistent with H2, brands were found to be more central in brand-related UGC on Instagram than in UGC on both Twitter and Facebook (p≤0.033). Between Twitter and Facebook, no significant differences were found (p<0.585). Therefore, the second hypothesis is only partly supported. Similar to the findings of Smith et al. (2012), brand centrality was expected also to be higher in UGC on Twitter than in UGC on Facebook. This is not the case in this research, indicating that the Starbucks brand plays more of a peripheral role in UGC on Twitter than it does in UGC on Instagram. One explanation for this could be that Twitter is often used to host details on the personal lives of users (Marwick and Boyd, 2011). After careful analysis of the data, it was found that tweets indeed talk about user’s personal lives in which Starbucks played a more peripheral role than it does in UGC about Starbucks on Instagram (e.g. “The one and only thing I dislike about working from home is not being able to pick up a coffee on my journey. #starbucksuk #caffeinateme)”. Even though brand centrality was not as high in UGC on Twitter
as it was on Instagram, it was still considered to be high. Furthermore, it should be noted that brand centrality does not align with self-promotion like it does in Smith et al. (2012)’s research. They found that the higher the degree of self-promotion, the lower the degree of brand centrality. For this research, this could not be confirmed.
Marketer-Directed Communication
Also for this dimension, the initial Chi-square test indicated significant differences among the three platforms in marketer-directed communications (p<0.000). In line with our third hypothesis, Facebook (14%, 38%, 32%) was the most frequently used platform for content that was directed to a marketer (p<0.000). This is not surprising because an additional country-specific Facebook page exists for each country, on which people can communicate (ask questions, complain about things) directly to a marketer. However, it was proposed that marketer-directed communications would be highest on Facebook ánd Twitter (4%, 6%, 14%); this was not supported. Twitter was found to be the second-most used platform for this purpose (p<0.000), but it still was not often used to communicate with marketers. As was already argued for the brand-centrality dimension, Twitter was often used to host posts on the details of the user’s personal life, or to share factual information (as will also be discussed later). On Instagram (0%, 0%, 4%) rarely any marketer-directed content posts were found. Therefore, hypothesis 3 was also supported. It is remarkable that no significant differences were found between countries, since marketer-consumer interaction per channel differs considerable among these countries. After analyzing the Twitter accounts of the specific countries, it appears that Starbucks France never publicly responds to questions that were posted, while Starbucks UK responds to every single question by (e.g.) providing product information, offering tips. The same was found when analyzing the Facebook accounts of each country: Starbucks UK always provides an answer to questions, while Starbucks the Netherlands and Starbucks France
do not respond to them at all. Apparently, this does not influence consumers that want to reach out, since our findings suggest that no statistically significant country differences exist in the way that marketer-directed communications are posted across different platforms.
Response to Online Marketer Action
Since the Poisson Regression on this dimension indicated that it was not appropriate to collapse the three countries for further analyses, Chi-square test were run for each country separately. The results indicated that no significant differences among the social media platforms could be found for the data of the United Kingdom (p<0.074), but that these differences did exist for the data from France (p<0.000) and from the Netherlands (p<0.000). It was hypothesized that brand-related UGC was least likely to be posted in response to an online marketer action on Instagram (4%, 2%, 16%), but this finding was only partly supported since Instagram and Twitter (4%, 2%, 16%) scored similar on this dimension; no significant differences were found between these platforms for either France (p<0.153) and the Netherlands (p<0.315). However, it was found that Facebook (22%, 22%, 22%) was the platform on which users most often responded to online marketer action for both France (p≤ 0.007) and the Netherlands (p ≤ 0.002). This makes sense for the data of the Netherlands, because Starbucks does not have an additional Starbucks country account for Instagram and Twitter. It does have a country account for Facebook, allowing Dutch people to respond to the marketer-generated content (that was sometimes customized for the Netherlands, like posts about promotions). Further, no differences were found for the data of the United Kingdom, in which Starbucks has a high maturity. The degree to which users were responding to online marketer action was similar across the three social media platforms. This finding is in line with the findings of Smith et al. (2012), who found that Facebook and Twitter were equally likely to feature posts that were a response to online marketer action. It should be noted that French consumers mainly use
Facebook to respond to online marketer action, even though it does have a country-specific page for each platform. So, for a country with low maturity users tend to use Facebook to respond to online marketer action, instead of using Instagram or Twitter for this purpose. This finding also makes sense if we look at the number of followers on the French Twitter account respective to the number of followers on the French Facebook account; This ratio is nearly 1:14, while this ratio for the UK is nearly 1:2. Apparently, for a low-maturity country, more people choose to engage with the brand (responding to online marketer action and directing communications towards a marketer) mostly on Facebook, instead of also using additional channels for brand engagement like users from a high-maturity country do.
Factually Informative About the Brand
As the Poisson Regression test showed, there were no significant differences among the datasets of France, the Netherlands and the United Kingdom for this dimension either. This indicated that differences in Starbucks maturity did not have an apparent effect on how factually informative UGC was posted on Facebook, Instagram and Twitter. Therefore, it was allowed to combine the datasets and to further analyze the complete dataset for differences among the social media platforms. The initial Chi-square test indicated that there were significant differences among the social media platforms regarding the degree to which factual information about the brand was present in the UGC posts. Hypothesis 5 posits that brand-related UGC that contains factual information about the brand would least likely be posted on Instagram (10%, 10%, 12%), and this hypothesis was supported by our Chi-square tests (p<0.000). Facebook (42%, 48%, 52%) and Twitter (58%, 50%, 58%) scored similar on this dimension (p<0.166): brand-related UGC that is factually informative is equally common across these platforms. Users share much information about new product launches (like the yearly Pumpkin Spice Latte launch), product prices and store locations on Facebook and Twitter, while Instagram is mainly used for entertainment purposes. Our findings differ from the findings of Smith et al. (2012),
who found that UGC on Twitter was more factually informative than UGC on Facebook. These findings appear not to hold in our case, which could either be because our research was conducted in another product category, or that it was conducted six years later. As mentioned before, the social media channels could have evolved in these six years (e.g. because new features were introduced), changing the ways that users interact with the platforms (Isosomppi, 2015).
Brand Sentiment
Again, the Poisson Regression test indicated that it was appropriate for the data of the three countries to be combined. The first Chi-square test yielded a significant statistic (p<0.000), indicating that there were significant differences in brand-sentiment across Facebook, Instagram and Twitter. Brand-related UGC on Facebook and Twitter appeared to be quite similar in its brand sentiment (p<0.171), while significant differences were found when comparing the brand sentiment of this content with the brand sentiment of UGC on Instagram (p<0.000). Since this was not a binary variable, we did a more in depth analysis, looking for differences regarding positive, negative and neutral content across the three platforms. This analysis provides statistical proof for the existence of more positive brand-related UGC posts on Instagram (64%, 54%, 74%) than on Facebook (52%, 36%, 34%) and Twitter (30%, 40%, 44%) (p<0.000). Therefore, hypothesis 6 was supported. Twitter and Facebook were found to be statistically similar and consistent regarding the percentage of positive brand-related UGC (p<0.669). For negative posts, Instagram was the platform (0%, 0%, 0%) that scored lowest (p<0.000): not even a single post on Instagram was coded as a negative post. Facebook (18%, 26%, 20%) and Instagram (2%, 14%, 24%) again were found to be similar regarding the amount of negative posts (p<0.070). Thus, hypothesis 7 was also supported. Regarding neutral posts, no statistically significant differences were found across the social media platforms (p<0.065).
It is interesting to note that Smith et al. (2012) presented completely different findings. They argue that brand sentiment may differ, but that it is not predictably different across sites. However, this is the case in our research, since our findings indicate that there are no significant differences across countries regarding the brand sentiment of social media platforms. In our research, Twitter and Facebook were similar regarding the amount of positive, negative and neutral posts. Instagram was the outlier relative to Facebook and Twitter regarding positive tweets (higher amount), but also regarding negative tweets (lower amount). However, considering the other dimensions this is not a surprising finding. Instagram is rarely used to talk to marketers or to responds to marketer action, it does not provide much factual information, and self-presentation is highest on this platform. Therefore, it would be odd to find many negative posts. This in an interesting finding for marketers, since brand centrality was also found to be very high on Instagram. Combining this with the fact that brand sentiment on Instagram is mostly positive (and never negative), this seems like an interesting and effective way to co-create the brand with consumers via UGC on Instagram. However, since the findings of Smith et al. (2012) were not consistent for brands with different social media strategies, it would be interesting to see whether this moderator would also have had an influence on our results.
