Analyzing Retweeting Intention In Microblog Marketing

Published Date: 02 Nov 2017

Abstract

Microblogs has been adopted widely by enterprises for business purposes. Retweeting is the key mechanism for information diffusion on microblogs and is especially important in microblog research. Users’ retweeting behavior has been studied in previous work. However, most of the existing studies are just simple qualitative or quantitative analyses of popular microblogs; few of them have proposed a framework to predict users’ retweeting behavior. In our work, we proposed a relatively general model of factors influencing the development of a user’s retweeting intention. We suggest that retweeting is similar to information adoption. The intention to retweet is predicted by perceived usefulness of the microblog content and subjective norms. Perceived usefulness is hypothesized to be influenced by informativeness, hedonism, timeliness and relevance of the information content. The research model was tested based on a survey. Pre-test results suggest that both perceived usefulness and subjective norms are positively associated with the intention to retweet. Neither timeliness nor hedonism has a significant effect on perceived usefulness.

Keywords: microblog, microblog marketing, content value, perceived usefulness, subjective norms, retweeting intention.

INTRODUCTION

With the development of Web 2.0 technology, Internet users have involved from simple unidirectional information consumers to bidirectional roles who are content producers as well as consumers (Cheung et al. 2012). Among various Web 2.0 applications, the microblog has become increasingly popular among Internet users. The microblog is a mini-form of the traditional blog. It allows users to tweet short text segments, normally less than 140 characters, that other registered users can see, retweet or comment on. While the microblog is the same as blog in essence, it has its own characteristics. Microblogs are commonly used for real time news and instant communication. Users can log in to a microblog through various channels, i.e. Web, mobile phone, third party applications. The widespread and multiple access methods provide mobility for microblog use, which is one of the key reasons for its popularity. The large user base and active updates from users ensure the existence of abundant and timely information on microblogs.

As both a social network site and an information source, microblogs have attracted a lot of users. Due to the characteristics of microblogs, information can spread very fast to a huge number of users with very little cost. These characteristics are appreciated by enterprises (Sui et al. 2010). Companies in many countries have realized the benefit of applying microblog technology in their marketing activities (Sui et al. 2010). On a microblog, companies can obtain useful information to optimize their production and sales operations and to enhance their communication with customers, partners, and suppliers (Xu et al. 2012). Enterprises post various kinds of messages on microblogs according to different aims. The posts may aim to raise public awareness, build brand image, advertise, manage customer relationships and so on (Zeng et al. 2011).

In any marketing project, information dissemination is always an important issue. On microblogs, the diffusion of messages is mainly based on retweeting. Retweeting is important not only because it facilitates information diffusion, but it also helps information to be better received by readers. When users receive marketing messages directly from firms, they would often regard it as advertisements. The same message would be more convincing if it comes from a credible person who is trusted by the recipient (Malhotra et al. 2012).

Since retweeting is especially significant for companies in microblog marketing, it is meaningful for us to research users’ retweeting intentions. The previous works related to retweeting are mostly based on analysis of existing popular microblogs. Few of them have proposed a framework to predict the formation of users’ retweeting intention. Furthermore, the microblogs posted by enterprises have their specific characteristics, with the ultimate goal being making sales. So the properties of marketing information should be recognized. Our research would be meaningful for understanding retweeting intention under enterprises’ microblog usage context. We will also provide practical reference for enterprises carrying out microblog marketing projects.

The rest of our work will be organized as follows. Firstly, we will introduce research background. Secondly, research model and hypotheses development will be presented. At the end, we will present our research methodology and the pre-test result.

RESEARCH BACKGROUND

Enterprises’ Usage of Microblog

As microblog becomes increasingly popular, it is a great platform for enterprises to promote sales, construct and maintain their business network and build brands (Zhang 2011). With the utilization of social network resources, companies can achieve better marketing effect. Because of retweeting mechanism, the messages from enterprises can reach a lot of targeted receivers in and out of their adjacent microblog network circle. This mechanism is very helpful for raising public awareness, which means "an understanding of the activities of others" (Dourish et al. 1992). Awareness is one of the significant foundations for successful marketing. By information dissemination, messages from the company reach consumers and consumers can get informed of the company’s activities. The principle of microblog marketing is based on word-of-mouth (Cheung et al. 2012). Word-of-mouth is defined as recommendation offered by current customers to potential customers of a good product or service. It is considered to be one of the most effective marketing strategies. The characteristic of microblog as social net working site makes it a great platform for word-of-mouth marketing (Bernard et al. 2009).

Various aims can be achieved through microblog. The company can build brand image by guiding user communication. Social relationship can be expanded based on the social network resources on microblog. By handling word-of-mouth, customer loyalty can be enhanced and new customer being attracted. Based on the communication with customers, company can learn about their opinions and feedbacks, so that the company would be able to improve transaction experience in future business. Sales promotion and customer activation can also be carried out on microblog platform. In addition, microblog is useful for crisis management. The real-time property of microblog enable the company to be informed immediately when crisis emerge. Also, during the crisis handling phase, the company can guide public opinion and be aware of the trends via microblog (Zeng et al. 2011). Microblog gathers valuable first-hand data from users directly. This information would be helpful for companies to make better strategies to cooperate with partners and cater for customers. Many companies have set up an official account on microblog as a communication channel. The enterprise’s microblog is a place for existing customers to express feelings, provide feedbacks, ask questions and get answers. It is also a place where potential customers can be drawn in (Bernard et al. 2009). Another fascinating advantage of using microblog is that there is no restriction to the company’s size, capital base or fame. Hence it is a good tool for small and middle sized company development.

2.2 Retweeting Intention

When a user receives a message, he/she can copy the original text, adding comments and reposting it, with an indication of quoting the original source. The username of the original author as well as the endorsed authors will be shown in the new microblog as a "@username" format. The reposted text can be retweeted again in the same manner. As this action carries on, the original message can be tracked by more and more users.

One characteristic of microblog marketing is that it is more reliant on social network resources, while blog marketing is more reliant on authors’ personal strengths (Zhang 2011). The basis for successful microblog marketing is that the information issued by the organization reaches targeted users effectively. There are two ways that a message issued by an enterprise can reach a user: the user followed the enterprise; the user’s followee retweet the message. Retweeting is the key mechanism for information diffusion on microblog. Like virus, one message gets disseminated widely and fast via repeated reposts.

Furthermore, when companies send messages directly to users, it may not receive much attention, because the message would be regarded as making sales. Readers would be more interested in the same message if it comes from their friends. This phenomenon is also common in word-of-mouth marketing, in which consumers perceive words from other consumers as more trustworthy and persuasive than advertisement of the company (Cheung et al. 2012). Also, after forwarding, the forwarder’s name will be added to the message, which means message get endorsement from the forwarder. Hence, if a message get retweeted by a creditable person, it will become more convincing (Malhotra et al. 2012).

Since retweeting is especially significant for companies in microblog marketing, analysis of users’ retweeting behavior became necessary. The motivation of retweeting has been studied in existing literature. Boyd et al. (2010) have studied retweeting as a conversational practice, after analyzing several datasets, they listed an incomplete set of motivations for retweeting, such as to spread tweets to new audiences, to entertain or inform audiences, to comment on someone’s tweet, to maintain friendship, to save tweets for future personal access (Boyd et al. 2010). This method was followed by Nagarajan et al., who also made an unexhausted list of characteristics of most-retweeted Tweets, including calling for some social action, collective group identity-making, crowdsourcing and information sharing (Nagarajan et al. 2010). Suh et al. have studied the characteristic of microblogs which are mostly retweeted. They divided the influencing factors into two categories — factors with content features and factors with contextual features. Content factors relate to the content of the microblog, while contextual factors relate to the person from whom a microblog is posted (Suh et al. 2010).

To sum up, the previous work related to retweeting are mostly simple qualitative or quantitative analysis of popular microblogs. Few of them have studied microblog user’s retweeting behavior based on survey. Furthermore, the microblogs posted by enterprises have their specific characteristics. So retweeting in the case of business is special. A research recognizing these differences would be meaningful for enterprises.

RESEARCH MODEL AND HYPOTHESIS DEVELOPMENT

Figure 1. Research Model

Perceived Usefulness

Useful means "capable of being used advantageously" (Davis 1989). The perceived usefulness of a microblog refers to user’s perception of the content being useful to him/her in any aspects. For example, the product news released by enterprise may help the users’ decision of purchase. The retweeting behavior is coined with information adoption in previous research (Sussman et al. 2003). The prominent theories regarding adoption behavior are Theory of Reasoned Action (TRA) (Ajzen et al. 1980; Fishbein et al. 1975), Technology Acceptance Model (TAM) (Davis 1989), Theory of Planned Behavior (TPB) (Ajzen 1985), and Unified Theory of Acceptance and Use of Technology (UTAUT) (Venkatesh et al. 2003). In the above-mentioned theories, individuals’ intention of adoption is formed based on evaluation of the targeted technology as well as the consequences of adoption. According to TAM, the intention to adopt is decided by perceived ease of use and perceived usefulness. Perceived usefulness refers to the extent to which a person believes that using a particular technology would enhance performance (Davis 1989). Retweeting is in essence an information adoption behavior. Since there are similarities between the intention of adoption and retweeting, factors that influence the adoption behaviors can also be used to predict the forwarding behaviors as well. Following this logic, the formation of users’ intention to adopt/forward a microblog would also rely on the usefulness of the content.

Among the numerous messages on microblog, there are both useful and useless contents. Andréde et al.’s research showed that, nearly 36% of the tweets are worth reading, 25% are not worth reading, and 39% are neutral (André et al. 2012). Individuals join microblog community to looking for new and interesting information (Java et al. 2007). The valuable messages are more-liked by users (André et al. 2009). Boyd et al. suggested that one reason for retweeting is to keep for future reference, because the retweeted posts will be stored in the forwarder’s microblog page (Boyd et al. 2010). In this way, it is reasonable that useful information being retweeted. What’s more, retweeting is an act of endorse, a person’s post is associated with his/her reputation, so useless microblogs are not likely to get forwarded in avoidance of ruining names. As a result, we hypothesise that:

H1. Perceived usefulness of a microblog is positively associated with a user’s intention to retweet it.

Informativeness

The basic function of Microblog was stated as sharing information and communicating with each other. With the high adoption rate, microblog has become a platform with huge user-generated-content database. Users primarily value microblog as an information source (André et al. 2012; Cheng et al. 2011; Kwak et al. 2010). Java et al. identified three categories of Twitter users: information sources, friends, and information seekers. Most enterprises fall into the category of information sources, whose main task is to post news and retain a large base of followers (Java et al. 2007). In contrast, individual users tend to act as information seekers. Some users even utilize microblog as a search engine to find answers to their question. The posts with informativeness are valued as more helpful. Customers expect tweets to be informative so that the information can help them (Black 2010). There are different categories of messages on microblog, some are self-centered and some are informative. The information sharing type of updates have been proved to be more welcomed and better valued by receivers (André et al. 2012; Naaman et al. 2010). Hence we hypothesise that:

H2. Informativeness is positively associated with the perceived usefulness of a microblog.

Timeliness

Timeliness, defined as the extent to which the age of the information is appropriate for the current task, has been one of the criteria to evaluate information quality for a long time (Lee et al. 2003; McKinney et al. 2002; Wang et al. 1996). One of the special characteristics of microblog compared with other social network sites is its emphasis on real-time information. The high frequency of updates on microblog ensured that the newest information can always be traced. Studies have suggested that microblog is not only a social net work sites, but rather a news media (Kwak et al. 2010). People log in microblog to obtain fresh information. As one of the measurements of information quality, timeliness is of great importance for data. The topics on microblog never stop renovating. Researchers have showed that topic is usually active only in a short period after first being published (Kwak et al. 2010). Some messages are often considered interesting only when it is new (Chen et al. 2010). Generally speaking, as the time grows, the news loses its value. Previous study have suggested that users found the old, repeated information not worth reading (André et al. 2012). In addition, as a platform providing faster mode of communication, the value of microblog is its ability to provide real-time information (Java et al. 2007). Hence we hypothesis that:

H3. Timeliness is positively associated with the perceived usefulness of a microblog.

Hedonism

Hedonic use of information technology has attracted more and more attention from researchers recently (Heijden 2004; Kim et al. 2005; Teo et al. 1999; Zhao et al. 2012). It has been regarded as a significant factor in personal adoption and continuance usage of information technology. Microblog, as social networking site, has the gratification of entertainment (Xu et al. 2012). Hence hedonic experience (playfulness/enjoyment) is especially expected during usage. Being humorous and interesting is an effective way to share knowledge. In André et al.’s research of ratings for tweet, content that are "funny" or "exciting" are frequently rated worth reading (André et al. 2009). Based on afore mentioned points, we hypothesis that:

H4. Hedonism is positively associated with the perceived usefulness of a microblog.

Relevance

Relevance means the extent to which the microblog posted by an enterprise is related to its followers (Agarwal et al. 2002). A microblog can relate to readers in many aspects such as job, age, health statues and personal interests.

In existing literature, the notion of relevance has been an inseparable dimension of information quality (Agarwal et al. 2002; Lee et al. 2003; Wang et al. 1996; Zmud et al. 1990). The relevant information is considered with higher quality. Microblog webpage is a platform on which the company communicates with current/potential customers (Smith et al. 2012). Through this kind of communication, the enterprises advertise about their products and brands, and customers obtain valuable information they need. Boyd et al. viewed microblog as a conversational platform (Boyd et al. 2010). A conversation would be more useful to a person if he/she is involved in the topic. Users engaging in communication with enterprises on microblog wishes to seek for solutions to their questions. If the enterprise recognize the audiences’ characteristics and post relevant messages, the information would be perceived as more useful. Hence, we hypothesise that:

H5. Relevance is positively associated with the perceived usefulness of a microblog.

Subjective Norms

Subjective norms are one of the main constitutions of TRA (Ajzen et al. 1980) and TPB (Ajzen 1985). It is defined as "the person’s perception that most people who are important to him think he should or should not perform the behavior in question" (Teo et al. 2003). An individual’s subjective norm is determined by his/her "perceived expectations of specific reference individuals or groups, and his or her motivation to comply with these expectations" (Linden 2011; Teo et al. 2003). In UTAUT (Venkatesh et al. 2003), subjective norm is included in social influence. It is demonstrated by an individual’s internationalization of the reference community’s culture and how much he/she cares about the enhancement of his/her image and belongingness to this community when making a decision.

With reference to previous research, users on social net working sites are not necessarily looking to meet strangers, instead, they are primarily communicating with acquaintance in their existing social networks (Boyd et al. 2008). Participants engage in online social network for practical needs as well as emotional needs (Xu et al. 2012). Identification is an important component of community. Members who want to increase identification should behave by norms, which means users pay attention to other people’s thoughts and feelings when they conduct behaviors (Hsu et al. 2008) .

An individual’s behavioral intention is influenced by his/her perception of whether people who are important to him/her think he/she should perform the behavior. Subjective norms influence individual’s evaluation of behavior and the consequences followed by conducting such behavior. When a microblog is retweeted by members of the a person’s reference group, we can believe that the person is more willing to conduct the same behavior so that he/she can comply with the norm, hence intention of this person to retweet this microblog is stronger.

H6. Subjective norms is positively associated with people’s intention to retweet.

RESEARCH METHODOLOGY

Measurements and Data Collection

Considering the nature of our research model, it was appropriate for us to conduct a survey to collect data to test our hypotheses. To ensure the validity of the scale, we developed our questionnaire primarily from previously validated measures. The final scales and the related references are presented in the Appendix A. Seven-point Likert scales ranging from ‘strongly agree’ to ‘strongly disagree’ were used to measure all items.

Users of the microblog were selected as the subjects of this study. Users of any microblogs (Sina, Tencent...) who have followed more than one enterprise microblog are qualified as respondents. Since we collected data in China, it was appropriate to use a Chinese language questionnaire. We used a forward–backward translation process to ensure equivalence of meaning. This was done by one of the authors and another invited PhD candidate. Before the formal survey, we conducted a pilot test to ensure the reliability and validity of the scale. We posted the Chinese questionnaire online and provided a hyperlink to the survey web page. The respondents would be directed to the online version of the questionnaire when they clicked on the hyperlink (http://www.sojump.com/jq/2145628.aspx). As a result, we received 77 responses. After deleting the invalid questionnaires, including those with the same answers to all questions, we got 75 valid responses.

Data analysis and results

We used a structural equation modeling technique, partial least squares (PLS), to test our research hypothesis. The Smart PLS 2.0 software package was used for our estimation.

Results of the measurement model

Based on two-stage analytical procedures (Hair et al. 1998), we first conducted confirmatory factor analysis to obtain the measurement model. Second, we examined the structural relationships. To validate our research model, content validity, convergent validity and discriminant validity were assessed. Content validity could be achieved if the measurement items were consistent with the previous research. Convergent reliability could be assessed by examining composite reliability and average variance extracted from the measures (Hair et al. 1998).

Construct reliability was measured by Cronbach’s alpha, Fornell’s composite reliability, and AVE. As shown in Table 1, the Cronbach’s alpha of all constructs ranged from 0.86 to 0.95, which was higher than the recommended minimum cutoff of 0.70. As to composite reliability, all the values were higher than the benchmark of 0.70. In addition, all of the AVE values in our study were higher than the standard of 0.50. Table 2 shows the correlation matrix and the square roots of the AVEs. The square root of the AVEs for each construct were greater than the correlations between constructs, hence we confirmed the discriminant validity.

Items

Factor analysis

Average Variance Extracted

Cronbachs Alpha

Composite Reliability

HN

IF

PU

RI

RV

SN

TM

HN2

0.8657

0.4415

0.3385

0.2577

0.1068

0.3091

0.2182

0.8211

0.8904

0.9322

HN3

0.9027

0.5869

0.3761

0.2145

0.3037

0.2415

0.4292

HN4

0.9482

0.5418

0.3818

0.179

0.2312

0.2238

0.3152

IF1

0.4796

0.8525

0.5682

0.3275

0.3769

0.2303

0.4951

0.7459

0.8868

0.9215

IF2

0.5103

0.8436

0.4574

0.1892

0.1451

0.0867

0.4908

IF3

0.5191

0.9014

0.5235

0.2847

0.1167

0.2446

0.4441

IF4

0.4992

0.856

0.4291

0.2879

0.1593

0.1676

0.4841

PU1

0.3189

0.526

0.9212

0.3866

0.5728

0.2754

0.3797

0.8016

0.9172

0.9417

PU2

0.4032

0.5704

0.8993

0.3438

0.553

0.2666

0.361

PU3

0.4324

0.5383

0.9024

0.322

0.5823

0.2068

0.3587

PU4

0.2928

0.4391

0.8572

0.3734

0.6293

0.3224

0.3501

RI1

0.2297

0.3542

0.3865

0.8968

0.3007

0.7175

0.3908

0.8250

0.9294

0.9496

RI2

0.2088

0.3767

0.4501

0.912

0.278

0.6677

0.342

RI3

0.1673

0.192

0.2932

0.9111

0.2755

0.6614

0.2012

RI4

0.2559

0.2228

0.3086

0.9133

0.3233

0.6602

0.1832

RV1

0.1445

0.1867

0.6305

0.3256

0.9198

0.2745

0.2208

0.8808

0.9322

0.9568

RV2

0.2968

0.2645

0.6054

0.2941

0.9532

0.2188

0.1977

RV3

0.2374

0.2253

0.6003

0.2915

0.9422

0.2347

0.1922

SN1

0.2399

0.2011

0.258

0.6738

0.2133

0.9305

0.4166

0.8710

0.9506

0.9643

SN2

0.259

0.2032

0.2449

0.6877

0.23

0.9547

0.4063

SN3

0.3071

0.2535

0.3242

0.7162

0.2688

0.9317

0.4061

SN4

0.2462

0.1503

0.2873

0.7058

0.2531

0.9158

0.3205

TM1

0.362

0.5128

0.3355

0.1447

0.231

0.3042

0.8904

0.7917

0.8681

0.9193

TM2

0.3267

0.5499

0.3955

0.3882

0.217

0.4371

0.9246

TM3

0.267

0.4088

0.3455

0.2807

0.1304

0.3555

0.8528

Table 1 Descriptive statistics of reliability indices and factor loadings for constructs.

IF

TM

HN

RV

PU

SN

RI

IF

0.864

TM

0.552

0.89

HN

0.575

0.349

0.906

RV

0.238

0.216

0.231

0.939

PU

0.575

0.404

0.402

0.654

0.895

SN

0.213

0.412

0.287

0.259

0.302

0.933

RI

0.316

0.30

0.241

0.324

0.398

0.746

0.908

The bold numbers are the square roots of the regarded average variance extracted.

Table 2 Correlation between constructs

Results of the structural model

After examining the measurement model, we tested the proposed hypotheses with PLS. The results of the analysis are shown in Figure 2.

The results show significant relationship between perceived usefulness and retweeting intention (b=0.192, P < 0.05), hence Hypothesis 1 is supported. Meanwhile, informativeness (b=0.41, P < 0.01) and relevance (b=0.538, P < 0.01) are found to influence perceived usefulness, hence both Hypothesis 2 and Hypothesis 5 are supported. However, there is no significant relationship between timeliness and perceived usefulness, hedonic value and perceived usefulness, showing that Hypothesis 3 and Hypothesis 4 are not supported. Subjective norms has a significant effect on retweeting intention (b=0.689, P < 0.01), hence Hypothesis 6 is supported.

Figure 2. Results of the PLS analysis

Discussion

This study aims to understand the factors influencing individuals’ retweeting intention. We suggested that retweeting intention is predicted by perceived usefulness of the microblog content and subjective norms. Perceived usefulness is hypothesized to be influenced by informativeness, hedonism, timeliness and relevance of the information content. Based on a pre-test of the research model, we get the following result (Table 3):

As expected, both H1 and H6 are supported. The relationship between subjective norms and retweeting intention is much stronger than the relationship between perceived usefulness and retweeting intention. This may be caused by the low variance of the sample. Most of the respondents are between 21 to 30 years old. Hence age may have positively moderated the relationship between subjective norms and retweeting intention. In future research, we will collect data with higher variance so as to avoid this potential error.

Both informativeness and relevance are proved to influence perceived usefulness. However, there is no significant relationship between timeliness and perceived usefulness (H3) or hedonism and perceived usefulness (H4). This may be caused by a shortage of samples. In the main study, we expect to collect more data (at least 300) and test our research model.

HYPOTHESES

RESULTS

H1. Perceived usefulness of a microblog is positively associated with user’s intention to retweet it.

Supported

H2. Informativeness is positively associated with perceived usefulness of a microblog.

Supported

H3. Timeliness is positively associated with perceived usefulness of a microblog.

Not Supported

H4. Hedonism is positively associated with perceived usefulness of a microblog.

Not Supported

H5. Relevance is positively associated with perceived usefulness of a microblog.

Supported

H6. Subjective norms is positively associated with people’s intention to retweet.

Supported

Table 3 Results of Hypothesis Testing