• No results found

Using Twitter : differences between industries

N/A
N/A
Protected

Academic year: 2021

Share "Using Twitter : differences between industries"

Copied!
46
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Running head: USING TWITTER: INDUSTRIES’ DIFFERENCES 1

Using Twitter: Differences between Industries

Georgina Camps Rusiñol 11103604

Master’s Thesis

Graduate School of Communication

Master’s programme Communication Science: Corporate Communication Supervisor: Mw. Dr. Iina Hellsten

(2)

Abstract

Twitter is a social media network used by the majority of organizations to engage with their consumers. It is also a powerful tool as it is economically profitable for for-profit

organizations, and may influence consumers’ purchase intentions and brand attitudes. Therefore, for social media managers it is important to know how to use Twitter in order to get consumers to endorse with their brands. The aim of this explorative study is to know how different types of industries use Twitter, and see if there is a relationship between the industry type and consumers’ endorsement with a brand on Twitter. By using content analysis, this research analyzes the relationship between the industry type and the level of consumers’ endorsement, which includes followers, retweets (RT) and favorites, taking into account factors that can affect this relationship, such as the tweet function, the use of links, hashtags, mentions, human voice, presence and type visuals. Results show that Twitter is used by organizations to build a community, followed by giving information and calling to action. Moreover, this research also shows that industries use Twitter in different ways, and that the tweet function affects the level of consumers’ endorsement. Tweets that try to build a community are the ones that get more endorsed by consumers followed by the ones that call to action and give information.

Keywords: Twitter, consumers’ endorsement, industry, for-profit organizations, RT,

(3)

USING TWITTER: INDUSTRIES’ DIFFERENCES 3 Using Twitter: Differences between Industries

The emergence of social media (SM), a set of “internet-based applications that build on the ideological and technical foundations of Web 2.0, and that allow the creation and exchange of user-generated content” (Colliander, Dahlén & Modig, 2014, p.181) has given the opportunity to consumers to engage more among them and with brands. Moreover, the number of consumers using social media has risen over the last years (Colliander et al., 2014). Studies suggest that social media influences consumers’ purchase intentions (Sonnier,

McAlister & Rutz, 2011) and that the same communication in social media has a greater effect on consumers compared to traditional media (Colliander & Dahlén, 2011). Social media networks have become a new and powerful tool for social media marketers to gain customer engagement, but also a challenge as they propose a change from traditional public relations and marketing strategies. Therefore, it is important to know which communication strategies for-profit organizations use to gain endorsement on Twitter.

The social media umbrella includes various online formats such as Facebook, a social networking site, Youtube, a creativity-work sharing site, Wikipedia, a website for

collaboration, and Twitter, a microblogging site (Colliander et al., 2014). The last one,

Twitter, is one of the most popular network for companies and it is used by the majority of the Fortune 50 firms (Case and King, 2011). Kaplan and Haenlein (2011) showed Twitter’s potential profitability finding that microblogging activities of Dell’s on Twitter generated in the United States $5 million in sales. In this study, I will use Twitter to analyze how

companies use this social media network to increase consumers’ engagement. Evidence suggests that companies should be present in social media and should send resources to social media campaigns, as it increases brand attitudes and purchase intentions (Colliander et al., 2014). Furthermore, the high number of companies present on social media is also an indicator that there are returns on the investment in social media. Moreover, according to

(4)

Edosomwan et al., (2011) when a company engages in social media activities on Twitter, it becomes more attractive to consumers and strengthens the company’s brand. So, it is

important for organizations to be present on social media sites and to engage in social media activities in order to gain endorsement.

Brand endorsement is a form of consumer engagement. It is defined as “a behavioral manifestation toward a brand or firm that goes beyond purchase and results from motivational drivers” (Van Doorn et al., 2010, p.01). In Twitter, to endorse with a brand includes following a brand, re-tweeting a tweet and liking a tweet. Consumers’ online brand endorsement has an advertising function as it uses consumers’ to distribute persuasive communications. This communication is more persuasive and trustworthy as users do not see the brand spreading the content but their peers, other users. Therefore, it is important for a brand to gain endorsement. Consumer engagement, “consumers’ non-transactional interactions with a brand or with other consumers in a brand context (Schamari & Schaefers, 2015. p.21) is considered a key element of successful social media activities. Schamari and Schaefers (2015) suggested that it is generally treated as something positive as high levels of consumer engagement with brands lead to improved attitudes and favorable behaviors.

For social media managers, it is important to know which communication strategies they should use on social media, as communication strategies allow brands to present themselves to consumers, and consumers can evaluate brands. Therefore, Twitter can help organizations gain endorsement from external stakeholders and customers in order to manage their corporate brand and reputation. On the one hand, the corporate brand is defined as the expressions and images of an organization’s identity. It is the interface between

organization’s stakeholders and its identity (Abimbola et al., 2012). It is composed by the corporate expression, which includes the brand promise, brand personality and brand communication, the ways in which the organization communicates; and brand image, the

(5)

USING TWITTER: INDUSTRIES’ DIFFERENCES 5 current and immediate reflection that stakeholders have toward an organization (Abimbola et al., 2012). On the other hand, corporate reputation is the stakeholders’ overall evaluation of an organization over time, and it is based on performance, product, services, citizenship, service, innovation, workplace, and governance (Abimbola et al., 2012). Therefore, this research aims to give managers some guideline in how to use Twitter to engage with stakeholders by

gaining endorsement, which can have an effect on brand reputation and on the creation of a corporate brand.

From an academic perspective, this research wants to participate in social media research by analyzing differences in the communication strategies and expanding the research in social media for for-profit organizations and from the organization perspective, not from the user perspective (Lovejoy & Saxton, 2012; Ashley & Tuten, 2015; Kelleher, 2009; Park & Lee, 2013).

Therefore, after seeing the benefits that using social media and gaining endorsement have for companies, it is important to know which communication strategies organizations use to gain endorsement on Twitter. Thus, the aim of this research is to answer the following question, which are the differences on the use of Twitter among industries? To be able to answer this question, in this research I will analyze the tweets organizations’ send. To do so, I will look into the tweet characteristics, such as the use of hashtags, mentions, presence of visuals (Lovejoy, Waters & Saxton, 2012); the tweet function, which analyzes if tweets aim to build a community, call to action, or give information (Lovejoy & Saxton, 2012); the use of human voice (Kerkhof et al., 2011; Kelleher, 2009); and the type of visuals, (Meyers-Levy & Malaviya, 1999; Mindiard et al., 1991).

Theoretical Framework

The aim of this research is to analyze which communication strategies different industries use to attract endorsement on Twitter. There are few studies on this topic, and

(6)

especially on how different industries use Twitter, thus, this is an exploratory study. In the next pages I will use communication theories, such as the excellence theory (Grunig & Grunig, 1992), and theories from other fields, for example, the affordance theory (Gibson, 1978), the existentialism theory (Cummings, 1978), the uses and gratification theory (Ashley & Tuten, 2015), and the dual coding model (Singh et al., 2002) to be able to answer my question.

The excellence theory (Grunig & Grunig, 1992) is a normative theory of how public relations (PR) should be practiced to be ethical and effective. The authors of this theory argue that in the practice of PR there are four different models (Table 1).

Table 1

Grunig and Grunig (1992) Public Relations Models

Model Characteristics

Press agentry (one-way asymmetrical)

Uses persuasion and manipulation to influence audiences to behave as the organization desires.

Public information (two-way symmetrical)

Uses press releases and other one-way communication techniques to distribute organizational information.

Two-way asymmetrical Uses persuasion and manipulation to influence audiences to behave as the organization desires.

Two-way symmetrical Uses communication to negotiate with stakeholders, resolve conflicts and reach mutual understanding and respect.

According to Grunig and Grunig (1992), the perfect model of PR is the two-way symmetrical, and due to the characteristics of Twitter, it should be the model used by companies. Moreover, they also argue that every organization practices a different model depending on their nature. For example, media and communities use the press agentry, government, uses two-way symmetrical and customers the two-way asymmetrical.

(7)

USING TWITTER: INDUSTRIES’ DIFFERENCES 7 Moreover, the affordance theory (Gibson, 1978) suggests that although the features of an object are common to everyone, the affordances are not. This means that the same object can be used in different ways depending on what is perceived good for. If I adapt this theory to SM, I can argue that each organization may use Twitter in a different way, depending on what they perceive is useful for. Furthermore, existentialism theory (Cummings, 1978) argues that not everyone will be impacted by the environment in the same way and that people are free to use behaviors based on the meaning they assign to the environment. This can also happen on Twitter, each organization may assign a different meaning to the environment, in this case, Twitter, and act in different ways in this social media network.

Taking into account the consumers’ perspective, Ashley and Tuten (2015) suggest that consumers will engage with a brand in a self-relevant way if the information shared by the brand is relevant to the consumer. So, each organization should adapt the content to the interest of their consumers. Thus, I would like to know if there are differences between industries and the level of consumers’ endorsement. In more detail, I aim to answer how and to what extent the industry type affects consumers’ endorsement with the brand, and how and to what extent the industry type affects the different types of consumer endorsement.

RQ1: Does and to what extent the industry of an organization affect the amount of consumers' endorsement?

RQ2: Does and to what extent the industry of an organization affects the type of consumers' endorsement?

The relationship between the industry type and consumers’ endorsement can be affected by different factors, the tweet function, tweet characteristics, use of human voice and presence and type of visuals.

(8)

Tweet Function

The first factor that can affect the relationship between an organization’s industry and the level of consumers’ endorsement is what is called social media content. Social media content is defined as creative strategies, “the executional factors and message strategies used to bridge the gap between what the marketer wants to say and what the consumer needs to hear” (Ashley & Tuten, 2015, p. 18). These strategies are important because they are related to advertising results as they can enhance receiver’s opportunity, motivation or ability to process the information of the post. Moreover, uses and gratification theory (Ashley & Tuten, 2015) can also be used to explain people’s needs for communication. This theory suggests that these communication needs are oriented to relationships, self, and content. Relationship-orientation is about the achievement of social interaction provided by the media, in this study, Twitter; self-orientation refers to the needs of the individual, and content refers to the

information delivered.

Literature on social media has categorized social media messages in different ways. In this study I will use Lovejoy and Saxton (2012) message categorization. On their study of how non-profit organizations use social media, they concluded that the tweets can have three main functions, to give information, to create a community or to call to action. They found out that non-profit organizations use Twitter mostly to give information, followed by creating a community and calling to action. This tweet categorization can also be used for for-profit organizations. Therefore, I expect that brands, depending on their industry, can use different tweet functions, and that the tweet function can have an effect on the level of consumers’ endorsement.

Expectation 1: The industry type will be associated with the tweet function.

Expectation 2: The tweet function will be associated with the level of consumers’ endorsement.

(9)

USING TWITTER: INDUSTRIES’ DIFFERENCES 9

Tweet Characteristics

The second factor that can affect the relationship between the industry type and consumers’ endorsement is the tweet characteristics. Tweet characteristics refer to the use of hyperlinks, hashtags, public messages, retweets (RT) (Lovejoy et al., 2012) and the presence of visuals. Therefore, I can expect that tweets, depending on their industry, can have different tweet characteristics, and therefore, brands can use tweets for different purposes when

communicating, taking into account the content and the needs of their consumers.

Expectation 3: The relationship between the industry type of an organization and consumers’ endorsement with the brand will be moderated by the tweet

characteristics. Human Voice

The third factor that can affect the relationship is the use of conversational human voice. It is defined as “an engaging and natural style of organizational communication perceived by an organization’s publics based on interactions between individuals in the organization and individuals in publics” (Kelleher, 2009, p.177). It includes providing immediate feedback, being open to have conversations and welcoming dialogue (Kelleher & Miller, 2006), admitting mistakes, treating others as humans, communicating with humor providing links to competitors (Searls & Winberger, 2001), message personalization and informational speech (Kelleher, 2009).

Using conversational human voice has positive effects for the organizations as it increases commitment, trust, satisfaction, and control mutuality (Kelleher, 2009; Park & Lee, 2013); word of mouth intentions (Park & Lee, 2013); influences attitudes and intentions, it is a driver for consumer engagement (Schamari & Schaefers, 2015); and gives credibility to the brand (Kerkhof et al., 2011). Moreover, the use of conversational human voice humanizes the

(10)

brands. This humanization increases and creates consumer reactions like positive emotions, patronage intentions, product likability and positive experiences toward a brand. If I take into account that people create a perception of firms (Aaker, Vobs & Mogliner, 2010), I can also assume that people have perceptions of industries. Therefore, the use of conversational human voice may be positive for some industries while negative for others when gaining engagement because people have different perceptions of industries. Thus, I expect that the use of

conversational human voice can moderate the relationship between the industry type and consumers’ endorsement with a brand.

Expectation 4: The relationship between the industry type of a company and

consumers’ endorsement with a brand will be moderated by the use of conversational human voice.

Visuals

The fourth factor that can affect the relationship is the visuals. Research shows that pictures are used in advertising to create a brand image, to show how a product can be used and who uses it, to get the attention of the consumer and to inform about the brand (Singh et al., 2000). Visuals are used because they are easier to memorize than words as they have the capacity to arouse mental images. The dual coding model (Singh et al., 2002), which studies differences between how humans process pictures and text, states that pictures are encoded in imaginal codes and texts in verbal codes, so pictures are labeled more spontaneously as the creation of the code is more likely for pictures than for words. Thus, pictures will have a greater number of codes that will act as multiple retrieval routes in the memory, which means that pictures will be memorized faster than text. McQuarrie and Phillips (2005) also support that pictures are easier to memorize than text and add that pictures can transmit specific beliefs. According to the authors, pictures are ambiguous and open, characteristics that make them more persuasive as they can lead to different interpretations, so people can create their

(11)

USING TWITTER: INDUSTRIES’ DIFFERENCES 11 own story about the picture and give them different meanings. Joffe (2008) adds that the main difference between the effects of texts and visuals is on their emotive impact. He argues that pictures are more emotive while texts are more rational, logical and linear. On his study, he shows that visual elements inculcate stronger personal concerns and evoke stronger emotional engagement than texts as they represent life situations and are subjectively relevant. The emotions transmitted by the pictures make them more vivid and they leave a stronger memory trace than text due to their salience effect. When attention is directed to a part of the

environment, it tends to be remembered when making judgments.

Pictures also have an impact on the persuasion process. Joffe (2008) suggests that when the audience is brought to an emotional state, the persuasion is more likely to occur. Meyers-Levy and Malaviya (1999) also support the influence of pictures on persuasion and explain it in two ways. They use the systematic processing theory, which states that message recipients may use a systematic processing strategy if they have the opportunity to process the message in an extensive and critical way, are motivated and able and can form accurate views. When this happens, the persuasive effect of the message will be influenced by the receivers’ perceived strength and diagnostic of the information on the message, which in turn, may be determined by the recipients’ perceived unique benefits of the object or issue the message is about. Thus, pictures can help identify the unique benefits of the object or issue for different reasons. First, pictures can directly and explicitly show the uses, benefits, and

characteristics of a product. Second, pictures give viewers reasons to substantiate the claim as they carry visual testimony. Miniard et al., (1991) show that pictures may influence post message attitudes, make the verbal information more memorable and evoke affective

responses. Powell et al., (2015) argue that visuals are highly salient and are perceived quickly as they are easier to understand than text, and therefore, the frame provided by the image may be stronger than the frame provided by the text. Thus, images are expected to have stronger

(12)

effects than text and to have a stronger effect on behavioral intentions than text. Therefore, visuals have a strong effect on advertising and persuasion, and companies should use them to persuade. Moreover, visuals can help companies gain endorsement as pictures are more persuasive than text.

Expectation 5: The relationship between industry type and consumers’ level of endorsement will be moderated by the use of visuals.

Furthermore, in this research, I also aim to know if different types of visuals, such as images, videos, GIF or graphs have an effect on the level of consumer endorsement.

RQ3: To what extent does the type of visual affect the level of consumer endorsement?

To sum up, I aim to know if there is a relationship between an organization’s industry and the level of consumers’ endorsement, and between industries and all the types of consumers’ endorsement: number of followers, number of RT and number of favorites per tweet. I also want to know if there is a relationship between the type of visual used in a tweet and the level of consumers’ endorsement. Moreover, I expect that industries use different tweet functions, that the tweet function is associated to consumers’ endorsement, and that tweet

characteristics, such as the use of links, visuals, hashtags and mentions, the use of

conversational human voice and the presence of visuals moderate the relationship between an organizations’ industry and consumers’ endorsement.

(13)

U S IN G T W IT T E R: IN D U S T RIE S ’ D IF F E RE N CE S 13 F igur e 1 . Conc ept ua l m ode l. Indus try C ons um ers ’ endor se m ent w ith the br and T w ee t f unc tion C onve rsa tiona l hu m an voi ce U se o f v isua ls T w ee t c ha ra cte ris tic s E 1 E 2 R Q 1/ R Q 2 E 3 H4 H5 V isu al type RQ 3

(14)

14 Methodology

In order to be able to answer the proposed research questions, and test the expectations of my explorative research, I used quantitative content analysis as according to Lombard et al., (2002), is a convenient method to analyze messages spread through mass communication media, such as Twitter.

The first step of my analysis was to select the sample, followed by collecting the data. After that, I created a codebook in order to be able to answer my research questions and test my expectations. To finish, I created the variables and ran the analysis.

In my analysis, there are ten variables, the level of consumers’ endorsement, which is the dependent variable, the tweet functions, which works as dependent in E1 and as

independent in E2. The industry type and the visual type, which are independent variables. The tweet characteristics, the conversational human voice, and the presence of visuals, which are moderators.

Sample

Since the aim of this study is to analyze the level of consumer endorsement between different industry types, the first step was to select the industries. To do so, I used the ranking of social media today (Carranza, 2015) about the industries that benefit most from social media, as in general, they include social media on their marketing strategies and use social media to get in touch with consumers, promote products and research. Those industries are entertainment, real state, marketing, retail, education, restaurants, and fashion. After selecting the industries, the second step was to choose the organizations for each industry type. I decided to select the five organizations that had the higher number of followers on Twitter from each industry according to Twittercounter (2016). As data from the marketing industry could not be found, I deleted that industry from my study. Moreover, only the organizations that tweet in English or Spanish were included as they are the languages known by the

(15)

USING TWITTER: INDUSTRIES’ DIFFERENCES

15

researcher. So the final sample included 30 different organizations, 5 from each type of industry (Table 2).

Table 2

Organizations Selected from Each Industry

Industry Companies

Entertainment Sugarhut, Pacha, El Teatro, Club One, XS Las Vegas Real Estate Zillow, Lennar, Rent Seeker, Sayfco Holding, Domain

Retail Whole Foods Market, Sephora, Duane Reade, Target, ToysRUs Education Sismologico Nacional, UNAM, Cultura UNAM, Education.com,

Harvard Heath

Restaurants Starbucks Coffe, McDonalds, Subway, Taco Bell, Nando’s Fashion Chanel, Victoria’s Secret, H&M, Marc Jacobs, Burberry

Note. The organizations are ordered from left to right according to their number of followers,

UNAM = Universidad Nacional Autónoma de México.

I collected the last 100 tweets sent per organization before November 22nd, but for some organizations, the program could not collect 100 tweets, so in the end, I collected 2966 tweets. Once tweets with missing data or in a language that was not Spanish or English were excluded, the final sample had 2719 tweets, 482 from the restaurant industry, 467 from the retail industry, 460 from the entertainment industry, 456 from the fashion industry, 446 from the real state industry and 408 from the education industry.

Procedure

The tweets included in the sample were collected using Python, a software that allows grabbing data from the Twitter Application Programming Interface (API). Data was gathered on the 22nd of November 2016 and it included the last 100 tweets sent by each organization. In order to analyze the content of the messages, I created a codebook (see Appendix A). It had 18 questions, some of them created by the researcher in order to be able to answer

(16)

16 the research question and test the expectations and some questions were grabbed from

previous research on data analysis on Twitter (Lovejoy & Saxton, 2012). The content categories were based on: consumers’ endorsement, tweet characteristics, tweet function: a call to action, give information, community building, the presence of visuals, type of visuals and the use of conversational human voice, as proposed in the literature.

In content analysis, the lack of reliability is a risk. Therefore, inter-coder reliability was tested using Krippendorff’s Alpha. I coded all the tweets, and an external coder coded a 10% (272) of the total tweets in order to test the reliability. The average Krippendorff’s Alpha of the variables is .81

Measures

Industry. To measure this variable each tweet was coded with a number from 1 to 6 depending of the industry the organization was from.

Tweet function. To measure this variable, I used Lovejoy and Saxton (2012) tweet functions. Each tweet had to be coded according to its main function, calling to action, giving information, community building or other. The category of others includes all the tweets that do no belong to any of the other functions. The mode was 3, which means that tweets that try to build a community are the most used. This variable has a Krippendorff’s Alpha of .79.

Tweet characteristics. This variable was measured by different questions. It includes the presence of links in the tweet, if the tweet is a retweet, if it has hashtags, the number of hashtags, if it has mentions, the number of mentions and the use of visuals. A new variable was created by summing the presence of links, hashtags, mentions, visuals and if the tweet was a RT (M=1.89, SD=1). Then, according to the median of my sample, I divided the variable into low presence of tweet characteristics and high presence of tweet characteristics, creating a new variable.

(17)

USING TWITTER: INDUSTRIES’ DIFFERENCES

17

Presence of visuals. This variable was measured by answering if the tweet had a visual or not. The mode is 0, which means that most of the tweets don’t include a visual.

Visual type. This variable was measured by analyzing the type of visual. The coder had to chose between image, which included pictures, photos, drawings, illustrations, collage, montage, GIF, graph or video. The mode for this variable is 2, which means that images are the most used visuals.

Human voice. To measure this variable four questions were created from Kelleher (2009) definition of human voice. It included questions such as “Does the tweet include or

use emotions?”, “Does the tweet includes message personalization?”. Coders had to answer

yes or no to each question. A new variable was created by summing the 4 questions (M=.33,

SD=.54). The new variable has a Krippendorff’s Alpha of .82. Then, by using median split I

created a new variable by diving into low use of conversational human voice and high use of conversational human voice.

Consumers’ endorsement. This variable was measured using Van Doorn et al., (2010) definition of endorsement and it includes the number of followers of the brand, the number of RT of each tweet and the number of favorites of each tweet. A new variable was created summing the number of followers, number of RT and favorites (M=3037791.76,

SD=3631469.42).

Number of followers. This variable was measured by checking the number of followers the organization had the day the data was collected. The mean is 3036526.59 (SD=3630863.44).

Number of RT. This variable was measured by counting the number of RT of each tweet. The mean is 72.51 (SD=522.135).

Number of favorites. This variable was measured by counting how many times a tweet was marked as favorite. The mean is 129.11 (SD=810.15).

(18)

18 Analysis

To test the expectations and research questions I used different methods of analysis. To test the RQ1, RQ2 and RQ3 I used ANOVA as this type of analysis allows you to compare differences between groups. To test E1, I used a crosstab to see the correlation between the industry type and the tweet function. For E2, to test the effect of the tweet function to the level of endorsement I ran an ANOVA. For E3, E4 and E5, as I was looking for a moderator effect and the independent variable and the moderator were categorical I ran a two-way

ANOVA as it allows to compare differences between groups and test for an interaction effect. Results

The means and standard deviations of the variables explained before are added in the following table (Table 3).

Table 3.

Descriptive Statistics of the Variables Included in the Study

Mean Std. Deviation N

Consumers’ endorsement 3037791.76 3631469.42 2718

Number of followers 3036526.59 3630863.44 2719

Number of RT 72.51 522.135 2719

Number of favorites 129.11 810.15 2718

Note. Categorical variables are not included.

Industry Type

To answer the RQ1, does and to what extent the industry type affects the amount of consumer endorsement I ran a one – way ANOVA as the independent variable is categorical and the dependent continuous. The industry type is the independent variable and consumers’ endorsement the dependent variable (Figure 2).

(19)

USING TWITTER: INDUSTRIES’ DIFFERENCES

19

I found a significant, large effect among the type of industry and the level of consumers’ endorsement, F (5,2717) = 1387.08, p < .001, η2 = .72. As the assumption of equal variances in the population has been violated, Levene’s F (5,2712) = 538.815, p <.001, a Tamhane’s post hoc test was used to test if there were significant differences between industries. Results indicated that there were significant differences between all types of industry (Table 1, Appendix B). It should be noted that the data for the consumers’

endorsement variable is not normally distributed. Therefore, I can say that the industry type affects the level of consumers’ endorsement and that the industry with highest consumer engagement is fashion.

Figure 2. Consumers’ endorsement per industry type.

Note. X-axis, industry type. Y-axis. Level of consumers’ endorsement.

To answer the second research question, does and to what extent the industry type is associated with each different type of endorsement three different one-way ANOVA were conducted. In the first analysis I used the industry type as independent variable and the number of followers as the dependent variable.

0 1000000 2000000 3000000 4000000 5000000 6000000 7000000 8000000 9000000

Entertainment Real State Retail Education Restaurants Fashion

(20)

20 Results show that the industry with more followers is fashion (M = 9131622.91, SD = 2043597.31), followed by restaurants (M = 4237008.58, SD = 3886108.25), retail (M = 2534188.62, SD = 1194079.66), education (M = 1652567.32, SD = 765331.09), real state (M = 237878.16, SD = 82922.42) and entertainment (M = 187500.88, SD = 24561.29).

I found a significant, large effect among the type of industry and the number of followers, F (5,2718) = 1388.05, p < .001, η2 = .72. As the assumption of equal variances in the population has been violated, Levene’s F (5,2713) = 539.255, p <.001, a Tamhane’s post hoc test was used to test if there were significant differences between industries. Results indicated that there were significant differences between all types of industry (Table 2, Appendix B). It should be noted that the data for the number of followers’ variable is not normally distributed. Therefore, I can say that the industry type affects the number of followers, and that differences exist between all types of industries.

For the second analysis I used the industry type as independent variable and the number of RT as dependent variable.

Results show that the industry with more RT per tweet is fashion (M = 293.44, SD = 827.70), followed by real state (M = 90.41, SD = 943.288), education (M = 30.43, SD = 58.38), entertainment (M = 15.56, SD = 59.85), retail (M = 6.31, SD = 47.87) and restaurants (M = 1.04, SD = 14.42).

I found a significant, small effect among the type of industry and the number of RT, F (5,2718) = 22.20, p < .001, η2 = .04. As the assumption of equal variances in the population has been violated, Levene’s F (5,2713) = 40.43, p <.001, a Tamhane’s post hoc test was used to test if there were significant differences between industries. Results showed that there are significant differences only between some industries (Table 3, Appendix B). It should be noted that the number of RT’s variable is not normally distributed. Therefore, I can say that

(21)

USING TWITTER: INDUSTRIES’ DIFFERENCES

21

the industry type affects the number of retweets and that differences between industries are only significant for some of them.

In the third analysis I used the industry type as independent variable and the number of times a tweet was added as favorite as dependent variable.

The tweets from the fashion industry are the ones that are added more as favorites (M = 693.89, SD = 1872.69), followed by education (M = 31.52, SD = 54.64), entertainment (M = 18.27, SD = 69.36), retail (M = 12.53, SD = 77.98), real state (M = 9.61, SD = 43.69), and restaurants (M = 6.50, SD = 108.77).

Results show a significant moderate effect among the type of industry and the number of favorites, F (5,2717) = 59.00, p < .001, η2 = .09. As the assumption of equal variances in the population has been violated, Levene’s F (5,2712) = 94.26, p <.001, a Tamhane’s post hoc test was used to test if there were significant differences between industries. Results showed that the only significant differences are between some industries (Table 4, Appendix B). It should be noted that the dependent variable, number of favorites, is not normally distributed. Therefore, I can say that the industry type affects the number of retweets and that differences between industries are only significant for some of them.

Tweet Function

To see which is the main tweet function used by organizations I ran a frequency analysis for the variable tweet function. Results (Table 4) show that the main function is to build a community, followed by give information, and call to action.

To test expectation one, which stated that there is a relationship between the industry type and the tweet function I ran a correlation using a crosstab as both variables are nominal. A Chi-Square Test was performed to examine the relationship between the industry type and the tweet function. The relationship between these variables is significant χ 2 (15) = 1463.38,

(22)

22 Table 4.

Frequencies for the Tweet Function Variable

Tweet function Frequency Percent Call to action 472 13,7% Give information 983 36,2% Community building 1264 46,2% Other 99 3,6%

To test expectation one, which stated that there is a relationship between the industry type and the tweet function I ran a correlation using a crosstab as both variables are nominal. A Chi-Square Test was performed to examine the relationship between the industry type and the tweet function. The relationship between these variables is significant χ 2 (15) = 1463.38,

p < .001.

There is a significant moderate association between the industry type and the tweet function, λ = .37, p <.001. The results (Figure 3) show that the industry that uses more tweets to call to action is education (34.7%), followed by entertainment (31.7%), fashion (20.7%), retail (7.5%)and real state, while restaurants (0%) don’t use it. The tweets to give information are mostly used for the fashion industry (25.5%), followed by education (24.7%), real state (23.3%), entertainment (22.4), retail (4%) and restaurants (0.1%). The tweets directed to build a community are mainly used by restaurants (37.9%), followed by retail (30.2%), real state (15%), fashion (8.5%), entertainment (5.9%) and education (2.6%). Therefore, I can meet our E1 and say that the industry type is associated to the tweet function.

(23)

USING TWITTER: INDUSTRIES’ DIFFERENCES

23

Figure 3. Use of tweet functions per industry.

In order to test the second expectation, which stated that the tweet function is associated to the level of consumers’ engagement I ran a one-way ANOVA, with tweet

function, as the independent variable, and consumers’ endorsement as the dependent variable. The tweets that try to build a community are the ones that get more consumers’

endorsement (M = 3288461.07 SD = 3466602.77), followed by the ones that call to action (M = 2944501.68, SD = 3283094-07), that give information (M = 2944501.68, SD = 3283094.07) and the ones in the category of others (M = 2258668.74, SD = 3631723.88).

Results show a significant small effect among the type of industry and the number of favorites, F (3,2716) = 4.68, p = .003, η2 = .01. A Tamhane’s post hoc test indicated that the only significant differences are between the tweets that give information and try to build a community (Mdifference = -456411.7, p = .026), and between the tweets that try to build a community and the category of others (Mdifference = -1029792, p =.012). It should be noted that the assumption of equal variances in the population has been violated, Levene’s F (3,2713) = 22.12, p <.001 and that the consumers’ endorsement variable is not normally distributed. Therefore, E2 is partially met, because there is a relationship between the tweet function and

0,00% 10,00% 20,00% 30,00% 40,00% 50,00% 60,00% 70,00% 80,00% 90,00% 100,00% Call to action

Give Information Community building Other

(24)

24 the level of consumers’ endorsement but it only exists between the tweets that give

information and the ones that try to build a community, and between the ones that try to build a community and the tweets in the category of others.

Tweet Characteristics

To test if the tweet characteristics moderate the relationship between the industry type and the level of consumers’ endorsement (E3) a two-way ANOVA was conducted with the industry type, the tweet characteristics and the interaction between the industry type and the tweet characteristics as independent variables and the level of consumers’ endorsement as dependent variable (Table 5).

Table 5.

Descriptive Statistics between Consumers’ Endorsement, Tweet Characteristics and Industry

N M SD

Industry Tweet characteristics

Entertainment Low presence 289 194532.64 18720.42

High presence 167 175456.15 28460.71

Real State Low presence 278 234807.37 94447.43

High presence 168 243225.10 59143.79

Retail Low presence 437 2560243.56 1207506.80

High presence 30 2154948.17 911498.83

Education Low presence 282 1854259.11 838722.87

High presence 126 1201362.48 169262.57

Restaurants Low presence 398 4729263.74 4081039.37

High presence 84 1904700.01 1090285.52

Fashion Low presence 274 9477602.33 2042226.86

(25)

USING TWITTER: INDUSTRIES’ DIFFERENCES

25

Results (Table 5, Appendix B) show a significant moderate effect among industry type and consumers’ endorsement and a significant small effect between tweet characteristics and consumers’ endorsement. Moreover, there was a significant small interaction effect between industry type and tweet characteristics to consumers’ endorsement. A post-hoc test could not be performed. In order to see the interactions a plot was created and it showed that for the fashion industry, education, retail and restaurants high presence of tweet characteristics decreases consumers’ endorsement, while for entertainment and real state the presence of tweet characteristics has a minimum effect on consumers’ endorsement.

It should be noted that the assumption of equal variances in the population has been violated, Levene’s F (11, 2703) = 341.48, p < .001, and that the consumers’ endorsement variable is not normally distributed. Therefore, expectation 3 is not met.

Human Voice

To test expectation 4, which stated that the presence of human voice moderates the relationship between industry type and consumers’ engagement I ran a two-way ANOVA, as the independent variable, industry, and the moderator, human voice, are categorical and the dependent variable, consumers’ engagement, is continuous (Table 6).

Results (Table 6, Appendix B) show a significant large effect between the industry type and consumers’ engagement, and a very weak significant effect between the presence of human voice and consumers’ engagement. There is also a small interaction effect between the industry type and the presence of human voice on consumers’ engagement. In order to see the effect a plot was computed and it showed that for fashion, retail and education the presence of human voice on a tweet has negative effects on consumer engagement, while for restaurants the higher use of conversation human voice increases consumers’ engagement. For

entertainment and real state industries the use of human voice has almost no effect on consumers’ engagement. It should be noted that the assumption of equal variances in the

(26)

26 population has been violated, Levene’s F (11,2705) = 467.53, p <.001, and that consumers’ endorsement variable is not normally distributed. Therefore, I can not meet expectation 4. Table 6

Descriptive Statistics between Consumers’ Endorsement, Human Voice and Industry

N M SD

Industry Human voice

Entertainment Low presence 317 187330.70 23401.95

High presence 142 188283.46 26790.91

Real State Low presence 379 231899.24 80866.70

High presence 66 273758.67 86771.13

Retail Low presence 234 27104043.30 1297794.48

High presence 233 23353599.78 1052136.68

Education Low presence 369 1668861.24 770955.72

High presence 12 1652629.27 765303.86

Restaurants Low presence 251 2949349.11 2466203.69 High presence 231 5636169.46 4603917.53

Fashion Low presence 328 9248453.03 2253066.50

High presence 128 8835763.10 1335804.11

Visuals

To test expectation 5 which stated that the presence of visuals moderates the

relationship between the industry type and the level of consumers’ endorsement a two-way ANOVA with the industry type as the independent variable, the presence of visuals as a moderator and the level of consumers’ endorsement as a dependent variable was conducted.

(27)

USING TWITTER: INDUSTRIES’ DIFFERENCES

27

Results (Table 7, Appendix B) show that there is a small significant effect between industry type and consumers’ endorsement and a weak significant effect between presence of visuals and consumers’ endorsement. Moreover, it also shows a small significant interaction effect between industry type and presence of visuals on consumers’ endorsement. A plot shows that the presence of visuals has a negative effect on consumers’ endorsement for restaurants, education and retail industry, while for fashion, real state and entertainment the presence of visuals does not effect consumers’ endorsement.

It should be noted that the assumption of equal variances in the population has been violated, Levene’s F (11,2704) = 377.55, p <.001, and that the consumers’ endorsement variable is not normally distributed. Therefore, I can not meet expectation 5.

To answer the third research question to what extent does the type of visual affects the level of consumers’ endorsement a one-way ANOVA was conducted with the visual type as independent variable and consumers’ endorsement as dependent.

The analysis shows that tweets that include a video are the ones consumers engage more (M = 5781190.23, SD = 5200037.07), followed by the ones that include a GIF (M = 3149435.71, SD = 3353261.79), the ones that include an image (M = 3099495.39, SD = 3999869.81), and the ones that include a graphic (M = 200446, SD = 16.97).

Moreover, data also shows a small significant effect between the visual type and consumers’ endorsement F (3, 1035) = 15.06, p < .001, η2 = .04. A Tamhane’s post hoc test indicated that the only significant differences are between the tweets that include a GIF and the ones that included a video (Mdifference = -2631755, p <.001), between GIF and graph (Mdifference = 2948989.7, p <.001), the ones that include an image and graph Mdifference = -2899049.4, p <.001), an image and a video (Mdifference = -2681695, p <.001) and graph and video (Mdifference = -5580744, p <.001). Therefore, I can answer to the research question

(28)

28 saying that the type of visual affects consumers’ engagement and that tweets that include a video are the ones that are more endorsed by consumers.

It should be noted that the assumption of equal variances in the population has been violated, Levene’s F (3, 1032) = 36.51, p <.001, and that the consumers’ endorsement variable is not normally distributed.

Discussion

Consumers endorsement leads to consumers’ improved attitudes and favorable

behaviors towards a brand (Schamari & Schaefers, 2015). Thus, organizations aim to endorse with consumers in social media. In this research, I set out to gauge into depth the factors that can impact consumer’s endorsement on Twitter depending on the industry of the organization. The aim of this paper was to explore if there are differences on the level of consumers’

endorsement between industries, and which factors affect this relationship. Therefore, I expected that tweet function, tweet characteristics, the presence of conversational human voice, the presence and type of visuals would have an effect on consumers’ endorsement with a brand on Twitter. I measured all the concepts doing a content analysis.

Results suggest that there is a difference between the industry type and the level of consumers’ endorsement, between the number of followers and the industry, between some of the industries and the number of RT, and between some of the industries and the number of favorites per tweet. Moreover, I also found out that the industry type and the tweet function are associated, and also the tweet function and the level of consumers’ endorsement. It is also important to point out that the most used function is community building, followed by giving information and calling to action. When it comes to the tweet characteristics results show that any of the industries includes a lot of links, visuals, hashtags or mentions on their tweets as tweets with low presence of those characteristics are predominant in all the industries. I did not find any effect between the presence of characteristics and consumers’ endorsement, and

(29)

USING TWITTER: INDUSTRIES’ DIFFERENCES

29

tweet characteristics do not moderate the relationship between the industry type and the level of consumers’ endorsement. Results also show that five of the industries do not use

conversational human voice, while the other one, retail, uses it in half of his tweets. I did not find a relationship between the use of conversational human voice and consumers’

endorsement, and conversational human voice did not moderate the relationship between industry type and consumers’ endorsement. Looking into the use of visuals, on one hand, I did not find any relationship between the presence of visuals and consumers’ endorsement, and the presence of visuals did not moderate the relationship between industry type and consumers’ endorsement. On the other hand, results showed that the type of visual affects the level of consumers’ endorsement. It is important to highlight that videos are the visuals that get more consumers’ endorsement followed by GIF, images, and graphics.

Taking a closer look at the results, I can say that the fashion industry is the one with a higher level of consumers’ endorsement, followed by restaurants, retail, education, real state, and entertainment. Fashion industry also leads the number of followers, number of RT and number of favorites per tweet. The predominance of the fashion industry and the high differences on the consumers’ endorsement compared to the rest of industries can be

explained by the fact that the companies from the fashion industry are known worldwide and have locals around the world. For the rest of the industries, most of the selected organizations are based on a specific country, and, their level of consumers’ endorsement is lower.

When it comes to the tweet function, results show that the tweets are used to build a community, followed by giving information, and call to action. This contradicts previous studies which stated that non-profit organizations mainly use Twitter to give information, followed to build a community and to call to action (Lovejoy & Saxton, 2012) and that the Fortune 50 firms commonly use Twitter to distribute news (Case & King, 2011). Therefore, results show that in the last years there has been a shift in the use of social medi.

(30)

30 Organizations want users to engage and participate in organizations’ social media activities. This change can be explained by the importance that users play in social media. For

organizations is important to engage with consumers’ as they can create a community, and after time, a strong fan network, that can help organizations respond to online firestorms as fans can be a trusted source of information for other users (Pfeffer, Zorbach & Carley, 2014). Moreover, when brands try to build a community, they also create dialogues with consumers, and therefore, organizations can improve consumers’ attitude toward the brand, increase purchase intentions and influence the perceived expense and caring (Colliander et al., 2014).

According to Grunig and Grunig (1992), the two-way symmetrical is the perfect model to use, as it is the more ethical and effective. On Twitter, the two-way symmetrical would correspond to the community building function, and according to Grunig and Grunig (1992) should be the most used function. Therefore, my results show that brands are using Twitter in the way proposed by Grunig and Grunig (1992).

The excellence theory (Grunig & Grunig, 1992) also suggests that organizations practice different models, and results also support this statement. Twitter is used to build a community mostly by restaurants, followed by retail, real state, retail, fashion, entertainment and education, while is used to call to action mainly by the entertainment industry, followed by education, fashion, retail and real state, and is used to give information by fashion, education, real state, entertainment, retail and restaurants. Moreover, the differences in the Twitter usage can also be explained by the affordance Theory (Gibson, 1978) which stated that the same object can be used in different ways depending on what is perceived good for, and by the existentialism theory (Cummings, 1978), which suggests that people, in this case, organizations, are impacted by the environment, Twitter, in different ways, and therefore, assign different meaning. Consequently, organizations act in different ways depending on the

(31)

USING TWITTER: INDUSTRIES’ DIFFERENCES

31

meaning they assign to Twitter. Moreover, these theories can also explain the differences in the use between for-profit and non-profit organizations.

When it comes to the level of consumers’ endorsement, the tweet function also plays a role. Results show that tweets that try to build a community and give information have

different levels of consumers’ endorsement and that the ones that try to build a community get more consumers’ endorsement, followed by the ones that call to action and to give

information.

Results also show that including a visual, using conversational human voice, and including links, hashtags, and mentions, decreases consumers’ endorsement for some of the industries. This can be explained by the fact that people can have different perceptions of the brands (Aaker et al., 2010), and therefore expect the brand to act in one way or another, and when those expectations are not met, consumers’ do not endorse.

Furthermore, the type of visual affects the level of consumers’ endorsement. The tweets that include a video are the ones that consumers endorse the more, followed by the ones that include a GIF, an image and a graphic.

The rest of the expectations could not be met, although results suggest that the

industry type has an effect on the use of visuals, conversational human voice and the presence of different tweet characteristics.

Limitations

The analysis in my research showed that some of the expectations were not met. This can be explained by the characteristics of the sample as data showed that there are big differences on the level of consumers’ endorsement, causing SD bigger than the means, and therefore, this can have affected the results. Moreover, the groups, in this case, the industries, did not have the same amount of cases, and therefore, the assumption of equal variances could not be met. Thus, I had to reject some of the expectations even if I found significant results.

(32)

32 This could be improved in future research by selecting another sample or selecting it in

another way. For example, by selecting the data from a specific point in time, and therefore, tweets would be all from the same timeline, and more comparable ones to each other. Moreover, the industries selected were from a different range, some of them were based worldwide, while others were local based, so future research could improve the sample by selecting only worldwide or local companies.

Another problem in the study is the characteristics of social media data. There are big differences in the number of followers between different accounts, and also in the number of RT and favorites between tweets, but this is the nature of the data, and can not be modified. This variance in the data breaks statistical assumptions needed in order to test data using parametrical test, therefore, social media data does not fit some parametrical statistical analysis. Therefore, further studies could use non-parametrical test to analyse social media data.

All these limitations lead me to assume that some of the results can only apply to this data set and can not be generalized.

Practical Implications

Although some of the variables included in the study were not good predictors of consumers’ endorsement, they do give an insight on what organizations can do on Twitter.

The most important finding is that the tweets that aim to build a community, and tweets that include a video, are the ones that get more endorsed by consumers. Therefore, social media managers that aim to gain consumers’ endorsement, should include videos, and use Twitter to build a community. This also indicated that consumers like to participate in Twitter, so organizations should let users participate on social media and influence, create or participate on brands social media campaigns.

(33)

USING TWITTER: INDUSTRIES’ DIFFERENCES

33

Results also show that including a visual, using conversational human voice, and including links, hashtags, and mentions, decreases consumers’ endorsement for some of the companies. Therefore, companies should look at their consumers’ behavior on social media toward the brand, create a social media strategy and include different characteristics and elements on their tweets according to what their consumers like.

Conclusion

This study aimed to explore the differences on the level of consumers’ endorsement on Twitter depending on their industry, and which factors can affect this relationship. Following the results, I can see that the industry affects the level of consumers’ endorsement and that industries use twitter in different ways, to build a community, to give information, and to call to action. Moreover, results also show that tweets that try to build a community are the ones that get a higher level of consumers’ endorsement.

Future studies should move on analyzing this results, exploring other industries and other factors that could affect the relationship between the industry type and the level of consumers’ endorsement in order to have a bigger picture on the use of Twitter among for-profit organizations.

References

Aaker, J., Vohs, K. D., & Mogilner, C. (2010). Nonprofits are seen as warm and for-profits as competent: Firm stereotypes matter. Journal of Consumer Research, 37(2), 224-237. Abimbola, T., Trueman, M., Iglesias, O., Abratt, R., & Kleyn, N. (2012). Corporate identity,

corporate branding and corporate reputations: Reconciliation and integration. European

Journal of Marketing, 46(7/8), 1048-1063.

Ashley, C., & Tuten, T. (2015). Creative strategies in social media marketing: An exploratory study of branded social content and consumer engagement. Psychology & Marketing,

(34)

34 Carranza, A. (2015, June 16). 7 industries that benefit most from social media. [Web article].

Retrieved from

http://www.socialmediatoday.com/social-networks/anthonycarranza/2015-06-16/7-industries-benefit-most-social-media

Case, C. J., & King, D. L. (2011). Twitter usage in the fortune 50: A marketing opportunity?

Journal of Marketing Development and Competitiveness, 5(3), 94-103.

Colliander, J., Dahlén, M., & Modig, E. (2015). Twitter for two: Investigating the effects of dialogue with customers in social media. International Journal of Advertising, 34(2), 181-194.

Cummings, L. L. (1978). Toward organizational behavior. Academy of Management Review,

3(1), 90-98.

Edosomwan, S., Prakasan, S. K., Kouame, D., Watson, J., & Seymour, T. (2011). The history of social media and its impact on business. Journal of Applied Management and

Entrepreneurship, 16(3), 79.

Gibson, J. J. (1978). The ecological approach to the visual perception of pictures. Leonardo,

11(3), 227-235.

Grunig, J. E., & Grunig, L. A. (1992). Models of public relations and communication. In J. E. Grunig (Ed.), Excellence in public relations and communication management (pp. 285-325).

Hillsdale, NJ: LEA.

Joffe, H. (2008). The power of visual material: Persuasion, emotion and identification.

Diogenes, 55(1), 84-93.

Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! the challenges and opportunities of social media. Business Horizons, 53(1), 59-68.

Kelleher, T. (2009). Conversational voice, communicated commitment, and public relations outcomes in interactive online communication. Journal of Communication, 59(1), 172-188.

(35)

USING TWITTER: INDUSTRIES’ DIFFERENCES

35

Kelleher, T., & Miller, B. M. (2006). Organizational blogs and the human voice: Relational strategies and relational outcomes. Journal of Computer-Mediated Communication, 11(2), 395-414.

Kerkhof, P., Beugels, D., Utz, S., & Beukeboom, C. (2011). Crisis PR in social media: An experimental study of the effects of organizational crisis responses on facebook.

Presentation at SWOCC Conference in Amsterdam, 13

Lombard, M., Snyder-Duch, J., & Bracken, C. C. (2002). Content analysis in mass communication: Assessment and reporting of intercoder reliability. Human

communication research, 28(4), 587-604.

Lovejoy, K., & Saxton, G. D. (2012). Information, community, and action: How nonprofit organizations use social media. Journal of Computer-Mediated Communication, 17(3),

337-353.

Lovejoy, K., Waters, R. D., & Saxton, G. D. (2012). Engaging stakeholders through twitter: How nonprofit organizations are getting more out of 140 characters or less. Public

Relations Review, 38(2), 313-318.

McQuarrie, E. F., & Phillips, B. J. (2005). Indirect persuasion in advertising: How consumers process metaphors presented in pictures and words. Journal of Advertising, 34(2), 7-20. Meyers-Levy, J., & Malaviya, P. (1999). Consumers' processing of persuasive

advertisements: An integrative framework of persuasion theories. The Journal of

Marketing, 45-60.

Miniard, P. W., Bhatla, S., Lord, K. R., Dickson, P. R., & Unnava, H. R. (1991). Picture-based persuasion processes and the moderating role of involvement. Journal of

(36)

36 Park, H., & Lee, H. (2013). Show us you are real: The effect of human-versus-organizational

presence on online relationship building through social networking sites.

Cyberpsychology, Behavior, and Social Networking, 16(4), 265-271.

Pfeffer, J., Zorbach, T., & Carley, K. M. (2014). Understanding online firestorms: Negative word-of-mouth dynamics in social media networks. Journal of Marketing

Communications, 20(1-2), 117-128.

Powell, T. E., Boomgaarden, H. G., De Swert, K., & de Vreese, C. H. (2015). A clearer picture: The contribution of visuals and text to framing effects. Journal of

Communication, 65(6), 997-1017.

Twittercounter. (2016). Twitter Top 100 Most Followers. Retrieved from: http://twittercounter.com/pages/100/organization

Schamari, J., & Schaefers, T. (2015). Leaving the home turf: How brands can use webcare on consumer-generated platforms to increase positive consumer engagement. Journal of

Interactive Marketing, 30, 20-33.

Singh, S. N., Lessig, V. P., Kim, D., Gupta, R., & Hocutt, M. A. (2000). Does your ad have too many pictures? Journal of Advertising Research, 40(1-2), 11-27.

Sonnier, G. P., McAlister, L., & Rutz, O. J. (2011). A dynamic model of the effect of online communications on firm sales. Marketing Science, 30(4), 702-716.

Van Doorn, J., Lemon, K. N., Mittal, V., Nass, S., Pick, D., Pirner, P., & Verhoef, P. C. (2010). Customer engagement behavior: Theoretical foundations and research directions.

(37)

USING TWITTER: INDUSTRIES’ DIFFERENCES 36

Appendix A Codebook Q0: Tweet id:

Q1: From which organization is the tweet? 1- Sugarhut 2- Pacha 3- El Teatro 4- Club One 5- XS Las Vegas 6- Zillow 7- Lennar 8- Rent Seeker 9- SAYFCO Holding 10- Domain 11- Whole Foods Market 12- Sephora 13- Duane Reade 14- Target 15- ToysRUs 16- Sismologico Nacional 17- UNAM Mexico 18- Cultura UNAM 19- Preescholers 20- Harvard Health 21- Starbucks 22- McDonalds 23- Subway 24- Taco Bell 25- Nando’s 26- Chanel 27- Victoria’s Secret 28- H&M 29- Marc Jacobs 30- Burberry Write the answer in numbers.

Q3: How many retweets does the tweet have? Write the answer in numbers.

Q4: How many favorites does the tweet have? Write the answer in numbers.

Tweet Characteristics

Q5: Is it a retweet? 0=No – 1=Yes.

Code yes if it starts with RT, or if the data collected indicates it is a RT. Q6: Does the tweet include a link to another website? 0=No – 1=Yes.

(38)

38 Code yes if the tweet includes a link to another website. If the link leads to a picture or video posted in youtube, Instagram, vimeo, or any other video/picture platform also code yes. Q7: Does the tweet include hashtags? 0=No – 1=Yes.

Code yes if the tweet includes a hashtag, #xxxx. If the answer to Q7 is yes, answer Q8, if not, skip to Q9. Q8: How many hashtags the tweet has?

Count the number of hashtags and write the answer in numbers (1,2,3, etc.) Q9: Does the tweet include a mention? 0=No – 1=Yes.

Code yes if the tweet mentions another user, @UserId. If the answer to Q9 is yes, answer Q10, is not, skip to Q11. Q10: How many mentions does it have?

Write in numbers.

Q11: Does the tweet include a visual? 0=No – 1=Yes.

Code yes if it includes a picture, image, graphic, infographic, drawing or video. If the visual is not in the tweet, if you have to press a link to go to the visual, code no. Pictures from Facebook, Instagram, or any other social media that is not Twitter are coded as no. Videos from Youtube, Vimeo or any other social media that is not Twitter are coded as no.

If the answer to Q11 is yes, answer Q12, if not, skip to Q13.

Q12: Which type of visual is it? 1-GIF, 2-Image, 3-Graph, 4-Video.

Code one if the visual is a GIF. Code 2, if the visual is a picture, illustration, photo, drawing, collage, montage. Code 3, if the visual is a graphic. Code 4 is the visual is a video. Tweet Function

Only analyze the text, do not analyze hashtag, links, images or videos even if they include text.

(39)

USING TWITTER: INDUSTRIES’ DIFFERENCES 39 Q13: Which is the main tweet function? 1=Call to action – 2=Give information –

3=Community Building – 4=Other.

Code call to action if the tweet calls people to react, to do something. Aims to get followers to do something for the organization. It also involves promotion. This category includes tweets that promote an even, try to sell a product or call for volunteers or employees, if the tweet includes a job offer or asks for volunteers to participate in an event, promotion, advertisement or campaign (all synonyms of this words, included). For example, “New

exhibition about pop art, come and discover Andy Warhol”, “Sales start today! Run to your nearest store”, “Don’t miss the match, today at 21:00”, “New Coca-Cola strawberry flavor! Try it!”, “Come and support your team. Tickets available”, “We have limited spaces

available on the Guest list for this Saturday night! Have you added your names??”, “Next week I’ll be playing in NY. See you all”, “Can’t wait to see you on Thursday on the dance floor”,”Online sales start at 22.00. Take advantage of it today.”, “Open from Monday to Friday. We are waiting for you”, “@UsedID burned the dance floor! Relieve the night on Snapchat”, “New video available on Youtube. Watch it now”.

Code give information is the tweet is only giving information about the brand, a product or an event, do not call to action or ask people to respond. For example, “We

launched a new campaign”, “Our summer campaign spot is out”,” New exposition in the museum”, “Concert today at 4:30 at the beach!”, “Sales season, starting from 9”, “Take a look at what's happening this weekend @UserID .... For VIP tables call 01277200885”, “Get ready! Ticket sales for Coldplay start tomorrow”.

Code community building if the tweet tries to create or build a community feeling, to interact, share and converse with stakeholders. This category includes all tweets that are giving recognition and/or thanks, that are answering the question of a follower, asking people to participate in a dialogue, to share their experiences/stories, to vote in some contest/quiz or

(40)

40 wishing anything to their followers. For exemple: “Thanks @UserID for telling us about your

experience at our restaurant. Hope you had a great time”, “Check what @UserID did with our products. We love it”, “We love this tweet! RT: @UserId: This is why I love partying in this club”, “@UserID if you have any other question, do not hesitaet to contact us”, “You can win a smartgirl box! Participate in: http://samsung.com/es/smartgirl”, “Have you tried our new pumpkin coffe? Share your experience with us”, “Our actress Lucy Hale is

nominated to teeens choice award. Vote her at :http://xxxxxx.com”. “Which flavor would you like to try in our coffe? Pumpkin, Spicy or Mint?”, “Help us decide our new t-shirt model. Vote which is your favorite”, “Looking forward to see you @UserID”.

Code other for the rest of the tweets. If they can’t be included in the previous functions. For example, “Good morning”, “People is having fun tonight”, “Our shops are

full of people”, “Always say yes to dancing”.

Conversational Human Voice (Kelleher, 2009) Q14: Does the tweet admit a mistake? 0=No – 1=Yes.

Answer yes if the tweet admits a mistake. For example, “Sorry for any inconveniences

we may have caused to our consumers. We’ll keep working hard to satisfy your needs”, “There have been a problem with the delivery. We are sorry and apologize for it”.

Q15: Does the tweet include or use emotions? 0=No – 1=Yes.

Code yes if it includes emoticons, emojis or sound mimicking (haha, hehe, hihi, hoho, huhu or similar), a lot of exclamation/question symbols together or uses words that refers to emotions, such as happy, sad, tired, bored, mad, angry, glad, proud, disappointed, crazy, furious, excited (also include synonyms). Also include sentences link we can’t wait, unbelievable, what a surprise, so cute, and expressions like hurrah, yupi, oh no, we are looking forward. For example: “@UserId ha ha it is funny! Thanks for sharing” “We love

Referenties

GERELATEERDE DOCUMENTEN

It is found that higher slack levels of human resources impact positively on internal innovation outcomes, whereas the squared terms of the independent variables are not

The night/day accident ratio for different classes of road and light levels (Amsterdam-West and Leeuwarden).. Classification of

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of

Om de doelstellingen en de vraagstelling te kunnen beantwoorden is gekeken naar verschillende soorten invloeden op werkstress: persoonskenmerken zoals leeftijd,

The importance for more private sector agricultural development in rural areas is clear but there are still some challenges when using agricultural PPPs for

An additional finding was that levels of parenting stress have strong associations with child psychopathology, and that different associations for mothers and fathers came to

Because of its importance, however, we will again mention that even though there are great differences between them, the Anderson model Hamiltonian matrices do have the

of the three performance indicators (return on assets, Tobin’s Q and yearly stock returns) and DUM represents one of the dummies for a family/individual,