Enhancing social media engagement in Japan: An empirical study of design and content factors on brand account

Abstract

What brand post characteristics increase consumer engagement activities (e.g., the number of likes, shares, replies, etc.) on social media? Although many scholars have empirically investigated this issue by conducting content analysis of brand posts on a variety of social media, many of them targeted Facebook as the subject of their study. In addition, the study of brand posts in the context of Japan has received little academic attention. Therefore, this study attempts to contribute to the research stream by conducting an empirical investigation into the key factors that increase the engagement of brand posts on Twitter in Japan. 500 brand tweets published by SHARP, a Japanese electronics manufacturer, were collected during the months of May and August 2021. The factors were examined under a new conceptual framework that extends previous studies. The tweets were coded according to an elaborated coding scheme and analyzed using a Poisson regression model. The results indicate that entertaining content and medium interactivity (i.e., call to action) are the determinants to increase engagement of brand tweets, as suggested by previous studies. The results of this study demonstrated that brand account personality as a design element is a key determinant to increase engagement of brand tweets.

1.  INTRODUCTION

Social media platforms, such as Facebook, Twitter, Instagram, and TikTok, have become powerful communication tools that directly connect brands with consumers. However, some brand posts are more likely to elicit a response from audiences, whereas others are not. Then, the question arises: Which brand post characteristic, such as links, pictures, and information, increase engagement (e.g., the number of likes, shares, replies, etc.) on social media? According to de Vries et al. (2012) and Menon et al. (2019), who investigate the impact of brand post characteristics on brand post engagement, these characteristics can be classified into two categories: content- and design-related. In this study, content refers to the type of information in a brand post, and design is defined as how the brand post is communicated. Researchers have empirically investigated this issue by developing conceptual frameworks that combine design/content characteristics as independent variables and brand post engagement as a dependent variable. However, these studies have produced mixed findings, which can be attributed to differences in the operationalization of design and content characteristics (Tafesse, 2015).

Studies on the effects of brand post characteristics (e.g., design, content, platform type, and posting time) suggest that marketers can improve consumer engagement by strategically designing brand posts. Nonetheless, studies focusing primarily on the design and content of brand posts have yielded mixed, sometimes contradictory, empirical results owing to differences in the operationalization of brand account design and brand post content. For all elements of brand posts addressed in previous studies, some empirical results show a positive impact on consumer engagement, whereas others show a negative impact.

Brand personality (BP) has been a popular topic in the marketing literature for over 50 years. Because brand personality serves a symbolic or self-expressive function, numerous studies emphasize its importance as a set of human characteristics associated with a brand (Aaker, 1997; Maehle et al., 2011; Chen et al., 2015; Lee et al., 2018). The impact of BP on the consumer–brand relationship has attracted interest, and much research has sought to understand BP as a way to build the relationship between consumers and brands. While brand personification strategies in the context of social media have received much attention, particularly the relationship between the BP of brand posts and consumer engagement (e.g., Kwon & Sung, 2011; Chen et al., 2015; Kim et al., 2017; Chu et al., 2022), the impact of brand account personality (BAP), which differs from brand personality (BP), remains underexplored.

While BP refers to a set of human characteristics attributed to a brand name, BAP refers specifically to the brand’s social media presence. Marketers imbue a brand’s social media accounts with human personality through the use of personal pronouns, imperative verbs, or human representatives. In this study, we examine the human representative information, such as a photo, name in the content of a brand’s account, or contact information is displayed. The use of personal pronouns or the inclusion of non-verbal cues (abbreviations, emoticons, or repeated punctuation) are the criteria for identifying BAP.

The implicit assumption that the BP is consistent with BAP is still open to question, and at least some studies show that BP differs from BAP in several respects (Cruz & Lee, 2014; Mizukoshi, 2019). Although offline BP is not perfectly aligned with online BP, BP and BAP may have different influences on brand communities, such as the number of replies, retweets, and favorites. Previous study that has conducted systematic literature reviews of consumer perceptions regarding digital brand personality reveal that there are inconsistencies and knowledge gaps in the dimensions, antecedents, and outcomes of digital brand personality (traits) (Ghorbani et al., 2022). These literature reviews and empirical results suggest that brand account personality, an endogenous characteristic of brand posts, influences consumer communication outcomes. However, to the best of our knowledge, there have been no empirical studies on the relationship between the BAP and consumer engagement. Therefore, this study examined whether brand account personality influences consumer engagement as an important design-related characteristic of brand posts.

Furthermore, regarding Twitter, little research has been conducted on the design and content of brand posting strategies in the context of Japan, even though the country has a rich user base. This study is based on a business perspective and emphasizes these factors from the view of a brand. We aim to fill the research gaps in previous studies by empirically demonstrating the importance of the design and content of Twitter brand posts in encouraging customer engagement (de Vries et al., 2012; Menon et al., 2019). Drawing on existing research findings on the content of brand posts and the concept of brand account personality, this study investigates the stimuli of content and design (vividness, interactivity, and brand account personality) that shape one’s engagement toward brand accounts on Twitter.

This paper comprises five sections, including an introduction. The first section discusses Twitter’s popularity in Japan and the fact that Japanese companies use it as a powerful social media tool to connect brands with consumers. The second section includes a literature review of social media studies, a description of the conceptual framework, and hypothesis development. The methodology is explained in the third section, in which the samples and methodological procedures are presented, followed by the results in the fourth section. Finally, discussion, implications, limitations, and suggestions for future research are presented in the fifth section.

1-1.  Research Background

The Japanese social media market is very different from that of the US, where Facebook has become the most widely used social media platform. Japan is known as a “the Twitter nation” (Akimoto, 2011). According to Statista (2021), Japan had more than 58.2 million Twitter users and ranked second globally as of October 2021. The Ministry of Internal Affairs and Communications (MIC) reports that the Twitter penetration rate in Japan reached 42.3%, which was higher than that of Facebook (31.9%) in 2020 (MIC, 2021). The penetration rate was especially high among the younger generations—67.6% for those in their teens and 79.8% for those in their 20s. On Twitter, users decide which tweet to react to (Wood & Burkhalter, 2013); therefore, they play the role of information gatekeepers (Manzanaro et al., 2018). For this reason, tweets must be well designed to evoke the attention of many users and garner engagement on Twitter.

Japanese companies view Twitter and Facebook as significant media for communicating with consumers and actively utilizing them. According to a survey on social media use by Japanese companies, 56.2% of companies in Japan utilized Twitter as of 2015, aiming to build daily relationships with consumers on Twitter (NTTCom Online Marketing Solutions Corporation & Looops Communications Inc., 2015).

According to Allied Architects (2019), who conducted an awareness survey of Twitter users on the use of Twitter by corporations (n = 1,100), consumers follow companies’ accounts on Twitter to gather valuable information on a particular brand, such as new information or coupons (52.5%), because they like the brand (24.7%) or join campaigns (22%). Architects (2019) show in another study (n = 685) that over 70% of followers think that companies’ official accounts have a positive impact, such as becoming more informed (46.4%), becoming more loyal (37.6%), increasing the number of visits to shops (30.1%), being more inclined to recommend the brand to others (16.6%), and posting more information about the company on social media (12.8%). Thus, Twitter has become one of the main communication tools used by Japanese companies.

2.  LITERATURE REVIEW, CONCEPTUAL FRAMEWORK, AND HYPOTHESES

2-1.  Literature Review

For more than a decade year, the growing popularity of social media has encouraged academics to conduct research on this topic. According to Tafesse (2015), studies on social media can be broadly classified based on two perspectives: audience and business.

Research from an audience perspective investigates users’ motivations and consequences of brand engagement on social media (Tafesse, 2015). For instance, Muntinga et al. (2011) investigated consumers’ online brand-related activities and corporate outcomes. Jang et al. (2013) studied how corporations’ use of social media affects consumer behavior, such as purchase intentions.

The present study is based on a business perspective, which is classified into research on the adoption of social media and research on content published on social media (Tafesse, 2015).

Studies on social media adoption investigate various factors and outcomes regarding the adoption of social media by corporations. These factors include the size of an organization, product category, marketing strategy, and international orientation, which are internal factors, as well as the existence of competitors and demographic changes. The second stream is research on social media content strategies, which is the interest of the current study. In the context of social media platforms, we define content strategy as one that increases brand post engagement by considering the elements to be included in them (Tafesse, 2015).

Table 1 presents the studies that focus on content strategy, referring to Deng et al.’s (2023) review format. These studies were selected based on the criterion that the study was conducted empirically at the brand post level to determine the factors that influence engagement (i.e., the number of likes, shares, and comments on brand posts). It summarizes the authors, social media platforms investigated, samples used, stimuli, and responses within the conceptual frameworks. Stimuli refer to independent variables in brand posts such as photos, videos, hashtags, or links. Responses are the dependent variables in brand posts that represent the engagement metrics for each brand post, such as the number of likes, comments (replies), or shares (retweets). The score in the response category refers to the aggregate number of engagement metrics.

Table 1Summary of previous studies regarding content strategy

#	Authors (published year)	Platform	Samples	Stimuli			Responses
				Content	Design		Likes	Comments	Share	Score
1	Chen et al. (2015)	Facebook	882 posts by 147 global brands	■			■	■	■
2	Cvijikj and Michahelles (2013)	Facebook	5,035 brand posts by 100 international food/beverage companies	■	■		■	■	■
3	de Vries et al. (2012)	Facebook	355 brand posts from 11 international brands	■	■		■	■
4	Dolan et al. (2019)	Facebook	2,236 posts by 12 Australian wine brands over a 12-month period	■			■	■	■
5	Khan et al. (2016)	Facebook	1,922 posts from 15 different facebook pages of 5 fast food brands across 3 countries	■	■		■	■	■
6	Labrecque et al. (2020)	Facebook	15,788 posts from 90 brands	■			■	■	■
7	Lee et al. (2018)	Facebook	106,316 posts by U.S. 782 companies across 6 genres	■			■	■
8	Lei et al. (2017)	Facebook	600 posts by 6 integrated resorts in Macau	■	■		■	■	■
9	Luarn et al. (2015)	Facebook	1,030 posts by 10 popular brands across various industries	■	■		■	■	■
10	Moran et al. (2019)	Facebook	757 posts from an Irish radio station		■		■	■	■
11	Sabate et al. (2014)	Facebook	164 posts from 5 Spanish travel agencies		■		■	■
12	Schultz (2017)	Facebook	792 posts by 13 brands in apparel or food industry	■	■		■	■	■
13	Tafesse (2015)	Facebook	191 posts by 5 top selling automotive brands in the UK	■	■		■		■
14	Tafesse (2016)	Facebook	4,190 posts from 85 brands across 3 industries	■			■		■
15	Araujo and Kollat (2018)	Twitter	281,291 tweets by 15 global food companies	■			■		■
16	Araujo et al. (2015)	Twitter	19,343 tweets by 65 global brands	■	■				■
17	Cruz and Lee (2014)	Twitter	19,462 tweets from 23 global brands and 1,257,151 tweets that mention or reply to them		■		■	■	■
18	Davis et al. (2019)	Twitter	24,960 tweets by 96 brands across 19 industries	■	■		■		■
19	Leek et al. (2019)	Twitter	838 tweets by 4 companies across 4 European countries	■			■	■	■
20	Matosas López (2018)	Twitter	41,392 tweets from 45 brands in Spanish food industry	■	■				■
21	McShane et al. (2021)	Twitter	41,141 tweets from 15 celebrity acccounts and 13 corporate accounts	■			■		■
22	McShane et al. (2019)	Twitter	119,050 tweets by 48 B to B brands across 8 industries	■	■		■		■
23	Mizukoshi (2019)	Twitter	146 survey responses for 2 Japanese brands’ account		■
24	Vargo (2016)	Twitter	7,447 tweets by 17 brands	■			■		■
25	Zanini et al. (2019)	Twitter	3,205 tweets from Sao Paulo FC	■			■		■
26	Chu et al. (2022)	Sina Weibo	254,389 posts by 67 automotive brands	■	■		■	■	■
27	Chae (2020)	Facebook, Twitter	3,252 branded posts by 11 global brands	■	■		■	■	■
28	Menon et al. (2019)	Facebook, Twitter	242 posts on Facebook and 143 tweets on Twitter by a Nordic Airline	■	■		■	■	■
29	Pezzuti et al. (2021)	Facebook, Twitter	7,382 posts on Facebook and 8,226 tweets on Twitter by 18 brands across various industries	■			■	■	■
30	Devereux et al. (2020)	Facebook, Twitter, Instagram	2.607 posts from 109 small retail firms in Australia	■	■		■	■
31	Kusumasondjaja (2018)	Facebook, Twitter, Instagram	10,752 posts from 43 leading Indonecian brands	■			■	■
32	Ashley and Tuten (2015)	Facebook, Myspace, Twitter	One week of 28 brands’ postings	■						■
33	Aichner (2019)	Facebook, Youtube, Instagram, Twitter	20,954 postings by 78 European FCs	■			■	■	■	■

The results of each study in Table 1 are summarized in Table 2 for design-related stimuli and Table 3 for content-related stimuli, in which both the stimuli and responses are combined. The results of statistical testing for each combination are presented as numbers in square brackets that correspond to the number of studies in Table 1.

Table 2Effects of design-related stimuli on responses

Stimuli		Responses	Statistical analysis results
			Significant (+)	Significant (−)	Non Significait
Vividness	Media presence	Likes	[10][22]
		Shares	[10][22]
		Comments	[10]
	Photo	Likes	[2][5][8][11][12][26][28]		[3][13]
		Shares	[2][5][12][26][28]		[8][13][20]
		Comments	[2][5][11][26]	[12]	[3][8][28]
	Event	Likes	[5]	[12]
		Shares	[5]		[12]
		Comments	[5]	[12]
	Video	Likes	[2][3][5][8][11][26]		[12][13][28]
		Shares	[2][5][12][13][26][28]		[8]
		Comments	[5][26]	[12][28]	[2][3][8][11]
Interactivity	Interactivity	Likes		[13]
		Shares		[13]
		Comments
	Link	Likes	[5]	[28]	[3][8][11][12]
		Shares	[5][16]		[8][12][16][20][28]
		Comments	[5]	[11][12]	[3][8][28]
	Hashtag	Likes	[12][27]	[22][26]
		Shares	[26][27]	[22]	[12][16][20]
		Comments	[12][26]		[27]
	Voting	Likes	[12]		[3]
		Shares	[12]
		Comments	[12]		[3]
	Call to act	Likes	[5][8][10][12]		[3][28]
		Shares	[5][10][12]		[8][28]
		Comments	[5][8][10][12]		[3][28]
	Contest	Likes	[3][8][12]
		Shares	[8][12]
		Comments	[8][12]		[3]
	Question/Quiz	Likes	[5]	[3]	[8][12][28]
		Shares	[5]		[8][12][28]
		Comments	[3][5][12][28]		[8]
Brand Account Personality	Brand personality	Likes	[17][26]	[17]
		Shares	[17][26]	[17]
		Comments	[26]		[17]
	Linguistic feature	Likes	[1][6][15][18][21][22][27][29]	[6][18]	[6][18]
		Shares	[1][6][15][18][21][27][29]	[18][22]	[6][18][20]
		Comments	[1][6][29]		[27]
Other characteristics	Message length	Likes		[7][27]	[4][18]
		Shares	[18]	[27]	[4]
		Comments		[7]	[4][27]
	Number of followers	Likes	[11][12][13][24][26][28]
		Shares	[12][13][24][26][28]
		Comments	[11][26]		[12][28]
	Position	Likes	[3][5][12]
		Shares	[12]		[5]
		Comments	[3][5][12]
	Date	Likes	[4]	[2][22]	[11][12][13][30]
		Shares		[22]	[2][4][12][13]
		Comments	[2]		[4][11][12][30]

Table 3Effects of content-related stimuli on responses

Stimuli		Responses	Statistical analysis results
			Significant (+)	Significant (−)	Significant	Non Significait
Post theme		Score	[33]	[32]
		Likes	[5][8][13][14][15][24][27][30]	[12][27]	[9][25][31]	[8][12][14][31][33]
		Shares	[12][13][14][15][16][24][27]	[12][13][27]	[9]	[5][8][12][14][16]
		Comments	[8][27]	[12][27]	[9][31]	[5][8][12][30][31][33]
Theme type	Informative	Likes	[2][4][5][7]			[3][13][28]
		Shares	[4]			[2][13][28]
		Comments	[2][5][7]			[3][4][28]
	Entertaining	Likes	[2][4][5][13][28]			[3]
		Shares	[2][28]			[4][5][13]
		Comments	[2][5][28]			[4]
	Social	Likes				[28]
		Shares				[28]
		Comments				[28]
	Remunerative	Likes	[4][28]	[2]
		Shares	[4]			[2]
		Comments	[2][28]			[4]
	Promotional	Likes	[28]
		Shares	[28]
		Comments	[28]

Four categories of design-related stimuli were observed: vividness, interactivity, brand account personality, and others. The findings for design-related stimuli are mixed. Some researchers have found a positive effect of videos on the level of vividness (containing images, photos or videos) are associated with more comments and user sharing (e.g., Chua & Banerjee, 2015; Chu et al., 2022), while others have not (e.g., Schultz, 2017). Similarly, the presence of a link as a level of interactivity has both positive (e.g., Khan et al., 2016) and negative results (e.g., Menon et al., 2019) with regard to brand post engagement. Although Chu et al. (2022) found a positive impact on garnering engagement, the stimuli should be considered a brand account personality combined with linguistic features, which we elaborate in the following section. Other characteristics include stimuli such as the message length of brand posts (e.g., Davis et al., 2019) or the day of the brand post (e.g., Cvijikj & Michahelles, 2013).

Regarding content-related stimuli, two categories were identified: theme and post-theme. Generally, in previous studies, samples are coded into five themes, namely, informative, entertaining, social, remunerative, or promotional. Positive effects on brand post engagement have been observed for informative (Cvijikj & Michahelles, 2013; Khan et al., 2016; Lee et al., 2018) and entertainment (Cvijikj & Michahelles, 2013; Tafesse, 2015; Lee et al., 2018; Menon et al., 2019) posts. However, Cvijikj and Michahelles (2013) and Menon et al. (2019) found mixed results for social, remunerative, and promotional posts. Post-theme summarizes studies that used their own theme type and those that could not be classified into the five themes (e.g., Araujo & Kollat, 2018; Tafesse, 2016). In the following section, we develop the hypotheses based on the results of studies in Tables 2 and 3.

2-2.  Hypothesis Development

We develop our hypotheses based on the literature review, and they include hypotheses relating to vividness, interactivity, and brand account personalities for design-related stimuli, and informative, entertaining, promotional, and remunerative content for content-related stimuli, as shown in Figure 1.

Figure 1 Conceptual framework

Note: “Social,” the stimuli employed by Menon et al. (2019), was not hypothesized, because the operationalization includes a question, which is covered under a level in interactivity.

 Vividness

Vividness is defined as “the representational richness of a mediated environment as defined by its formal features; that is, how an environment presents information to the senses” (Steuer, 1992, p. 81). Steuer states that vividness comprises two main variables: breadth and depth. Breadth refers to the number of sensory systems that can be stimulated by a medium and depth is the level of resolution within each sensory system (Steuer, 1992). His study aimed to find the determinants of telepresence, and “vividness” was one of the variables that explained it. While vividness is also referred to as media richness, brand posts on social media will always have some level of vividness - be it null, low, moderate or high (de Vries et al., 2012; Chua & Banerjee, 2015; Soares et al., 2019). A video is considered more vivid than a picture because the former stimulates both visual and aural senses, while the latter stimulates only the sense of sight (de Vries et al., 2012).

Previous researchers have understood vividness as applicable to the analysis of corporations’ online messages, such as websites or brand posts on social media, and have incorporated it into their conceptual frameworks (de Vries et al., 2012; Tafesse, 2015; Menon et al., 2019).

Except for McShane et al. (2019) and Moran et al. (2019), who measured the presence of media as a level of vividness, vividness has been classified into two main levels when included in the conceptual framework: low for photos or pictures and high for videos or events (Chu et al., 2022; Cvijikj & Michahelles, 2013; Khan et al., 2016). While some studies show a negative effect of the level of vividness on brand post engagement (e.g., Schultz, 2017), most studies generally report positive effects (e.g., Chu et al., 2022; Cvijikj & Michahelles, 2013; Khan et al., 2016). Chu et al. (2022) found that higher vividness leads to higher engagement. Based on the dominant conclusions in the literature and theoretical background, the higher the level of vividness in a brand post, the more engagement is achieved. Thus:

Hypothesis 1. 

Tweets with a higher level of vividness achieve higher brand post engagement.

 Interactivity

According to Liu and Shrum (2002), interactivity is defined as “the degree to which two or more communication parties can act on each other, on the communication medium, and on the messages and the degree to which such influences are synchronized” (p. 54). Before the advent of social media platforms, brands and consumers were connected through one-way communication, such as via offline advertisement, so they could not “act on each other.” However, in the era of social media, interactive stimuli encourage users to react to brand posts (Schultz, 2017), and thus are considered in content strategy studies (e.g., de Vries et al., 2012; Schultz, 2017; Menon et al., 2019). Tafesse (2015) operationalized interactivity by tallying the number of stimuli (e.g., links, hashtags, questions), while previous studies classified interactivity into three levels: low for link, hashtag, and/or voting; medium for call to act, and/or contest; and high for question/quiz (e.g., de Vries et al., 2012; Khan et al., 2016; Schultz, 2017; Chae, 2020).

The effect of interactivity on brand post engagement differs by stimulus type. For instance, “link” is found to have positive (e.g., Khan et al., 2016), negative (e.g., Sabate et al., 2014), and no relationships (e.g., de Vries et al., 2012; Lei et al., 2017), while the “call to act” has a mostly positive effect (e.g., Khan et al., 2016; Schultz, 2017; Moran et al., 2019). An overview of the literature suggests that interactivity generally has a positive effect on brand post engagement. For instance, Khan et al. (2016) and Luarn et al. (2015) conclude that higher interactivity can lead brand post to garner higher brand post engagement. Thus:

Hypothesis 2. 

Tweets with higher levels of interactivity lead to higher brand post engagement.

 Brand Account Personality

Brand personification strategies are popular subjects of study and are usually classified into two main categories based on the occurrence of situation personification: graphic and textual content (Chen et al., 2015). The graphic content strategy itself entails anthropomorphism, zoomorphism, and teramorphism (Brown, 2011). Textual content refers to the use of language that brands use, such as use of personal pronouns (Chen et al., 2015). Kim, Kwon, and Kim (2017) state that marketers should recognize that their behavior on social media can humanize their brands, especially when consumers recognize human-like characteristics in brands on social media. According to Maehle et al. (2011), the anthropomorphic theory explains this phenomenon: First, anthropomorphizing can turn non-human objects into humans and familiar objects; second, people feel comfortable and reassured when they interact with anthropomorphic objects; and third, anthropomorphizing reduces the complex and ambiguous uncertainty of objects by providing the anthropomorphized object with human characteristics.

The findings on brand personification strategies in advertisements and social media platforms also support the anthropomorphic theory. For instance, the presence of a brand personality has been found to induce brand post engagement, such as the number of likes (e.g., Cruz & Lee, 2014; Chen et al., 2015; Lee et al., 2018). Hart and Royne (2017) show that anthropomorphism can improve consumers’ attitudes toward a brand and advertisement, and even enhance purchase intentions. Kim et al. (2020) show that consumers develop favorable attitudes toward an anthropomorphized brand, engaging positively with the brand’s posts, compared with a non-anthropomorphized brand.

However, these studies have overlooked the crucial elements of brand account personality. According to Mizukoshi (2019), when a brand forms an account on social media, the account also has a personality, called a “brand account personality.” Mizukoshi claims that brand personality and brand account personality are significantly different and that the degree of alignment between the two influences engagement in the brand community, such as the number of replies, retweets, and favorites. We define brand account personality as “a set of human characteristics associated with brand accounts”, referring to Aaker (1997). The presence of a brand account personality can be a stimulus for crafting a brand post. According to the anthropomorphic theory, the effect of the presence of a brand account personality should be similar to that of a brand personality. Thus:

Hypothesis 3. 

Tweets with a brand account personality increase brand post engagement.

 Informative Content

Consumers visit stores to check for new products, news, events, or campaigns about their favorite brands, which constitute, in other words, brand-related information. Such information motivates consumers to visit online brand communities (Muntinga et al., 2011). Informative content has been observed to positively affect brand post engagement (e.g., Cvijikj & Michahelles, 2013; Khan et al., 2016; Dolan et al., 2019). Thus:

Hypothesis 4. 

Informative tweets lead to higher brand post engagement.

 Entertaining Content

If consumers enjoy a brand page, they are likely to visit it again (Ashley & Tuten, 2015). Muntinga et al. (2011) show that entertaining content in a brand post encourages consumers to consume (e.g., watch videos), contribute (e.g., comment), and create brand-related content (e.g., write product reviews). Generally, entertainment content has a positive relationship with brand posts engagement (e.g., Cvijikj & Michahelles, 2013; Khan et al., 2016; Menon et al., 2019). Thus:

Hypothesis 5. 

Entertaining tweets lead to higher brand post engagement.

 Promotional Content

The empirical investigations of promotional content are scarce (Lee et al., 2018). Among the few studies, Lee et al. (2018) found that promotional content, which they call “direct informative content” such as mentioning prices and deals, reduces brand post engagement on social media when provided in brand posts alone. Menon et al. (2019) empirically investigated brand posts on Facebook and Twitter and concluded that promotional content could increase Twitter engagement; however, this was not fully supported in the case of Facebook. Thus:

Hypothesis 6. 

Promotional tweets lead to lower brand post engagement.

 Remunerative Content

Remunerative content offers economic benefits and attracts attention in the form of contests, coupons, and other offers (Menon et al., 2019). Luarn et al. (2015) show that remuneration is highly effective in facilitating liking but not for commenting and sharing. Indeed, there may even be a positive relationship between remunerative content and brand post engagement (e.g., Dolan et al., 2019; Menon et al., 2019). Thus:

Hypothesis 7. 

Remunerative tweets lead to higher brand post engagement.

3.  METHODOLOGY

3-1.  Sampling and Dataset

SHARP’s Twitter account (@SHARP_JP) is selected as the sample for three reasons. First, SHARP was found to have a brand account personality that was significantly different from its brand personality (Mizukoshi, 2019). In addition, SHARP’s Twitter account (@SHARP_JP) boasts of over 800,000 followers as of January 2022 and tweets almost every day, indicating that it generates a rich set of tweets for analysis. Finally, the company started using Twitter in 2011, and Mr. Takahiro Yamamoto has since been the sole account handler (Bunshun Online, 2018). He is well known as the account handler of @SHARP_JP, and his engagement with followers is second among the prevailing corporate Twitter accounts (Wakejima, 2018). Mr. Yamamoto has received marketing-related awards for his tweets, such as the Saji Keizo Award by the Osaka Advertisement Association, which is given to active creators in the Kansai area (AdverTimes, 2017), and the JDC Award by Siemple DCRI, which is awarded to companies that demonstrate outstanding communication on the Internet (DCRI, 2021).

The main reason we focused our analysis on a single electrical manufacturer in Japan is that there are few empirical studies and little experiential knowledge on brand account personality. In addition to the challenges of measuring common metrics of brand account personality across diverse brands, we believe that analysis of highly differentiated brand account personalities using cross-section data may dilute the effect on customer engagement. Since brand account personality is a novel stimulus in the context of social media marketing, we decided to analyze the account personality of SHARP, and attempt to elucidate its effect on customer engagement.

We empirically analyzed 500 tweets published by SHARP’s official Twitter account from May to August 2021. The account handler of SHARP’s official Twitter account, Mr. Takahiro Yamamoto, kindly provided us with tweet data in the form of CSV files retrieved from the Twitter Activity Dashboard, which stores past tweet activities.

Although the total number of original tweets during the study period amounted to 1776, 1276 tweets were excluded because they were not appropriate for analysis in terms of the conceptual framework employed in this study. The excluded tweets included SHARP replies to themselves, replies to other Twitter users, and tweets pertaining to SHARP retweets. They were not independent and were closely related to other tweets, so unexpected factors outside the conceptual framework could influence the results of the empirical investigation if they were included in the dataset. To be more specific, the excluded tweets include those by SHARP that are replies to the account’s own tweets, that mention a user through the @ function (i.e., SHARP’s reply to another user), and tweets that include “>RT” and “>>RT” at the end of sentences (i.e., the tweet is a retweet of a previously published tweet). After data cleaning, exactly 500 tweet samples were used as the dataset for the empirical analysis.

We checked our assumption and conceptual framework by carefully examining the samples. This examination revealed that the sample did not include remuneration tweets. Therefore, the “remuneration” variable was omitted, and Hypothesis 7 was withdrawn. The updated conceptual framework is illustrated in Figure 2.

Figure 2 Updated conceptual framework

3-2.  Operationalization of Variables

We determined the operationalization of variables by following the approach in the extant literature (e.g., de Vries et al., 2012; Kwon & Sung, 2011; Menon et al., 2019; Schultz, 2017), as described below. Table 4 summarizes the operationalization of vividness and interactivity.

Table 4Operationalization of vividness and interactivity

Code	Vividness	Interactivity
None	(baseline)	(baseline)
Low	Picture (photo or image)	Link to a website Hashtags
Medium		Call to act (encourage brand followers to do something)
High	Video (mainly videos from YouTube)	Question Quiz

 Design Variables

The design variables considered were vividness, interactivity, and brand accounting personality. Vividness and interactivity are categorical variables and brand account personality is a binary variable.

Vividness: Vividness has three categories (none, low, and high). If a tweet has photo(s) and/or image(s), it is coded as “low.” When featured images/feature-graphic links are present in tweets, they are not deemed to be photos or images, because they correspond to links in tweets. If featured images are also identified as photos or images, the categorization of vividness is useless. In the case of video, mainly YouTube videos, coding is “high.” A GIF can be considered as a type of video, so it is coded as “high” as well. If a tweet had neither of these, it was coded as “none,” which was the baseline. Events were not found in the sample as stimuli; therefore, they were not considered.

Interactivity: Interactivity has four categories (none, low, medium, and high). A tweet with link(s) or hashtag(s) is classified as “low.” Following previous studies (de Vries et al., 2012; Menon et al., 2019), links to a company’s website were excluded. Most links in SHARP’s tweets were directed at either the account handler’s blog or an article about its products and/or services on review websites. When SHARP quotes someone or its past tweet in its tweet, the quoted tweet is considered a link and classified under “low” as well. A tweet is categorized under “medium” when it calls to act; in other words, it encourages followers to do something. Examples include “please check this website,” “tell me how to use this function on Twitter,” and so forth. “High” tweets are those with questions or quizzes for its followers, such as “do you know this product?” A tweet was coded as “none” as the baseline if it had no interactivity. Contests and voting were not operationalized for the sample that did not have a tweet with the stimuli.

Brand account personality: The criteria for coding brand account personality were developed based on Kwon and Sung (2011). The scheme by Kwon and Sung was designed for English; therefore, it was modified to the Japanese version because all tweets by SHARP were in Japanese. If a tweet meets any of the three conditions in Appendix 1, it is coded as 1, indicating that it has a brand account personality. When a tweet, for instance, say “I had a nice ramen today!!!”, the tweet is coded as 1 because it includes a personal pronoun “I” and a nonverbal cue “!!!”. Otherwise, 0 was coded as the baseline value. The reason why we do not directly put each item by Kwon and Sung (2011) into the model is that we define each as the components of BAP and study BAP as one variable. BAP was overlooked in the stream of content strategy studies, so the significance as the overall concept is needed to be assessed first.

 Content Variables

Each tweet was classified according to one of the three content variables. These variables were treated as binary data, with the coder assigning a value of “1” to indicate the presence of a characteristic. If it is absent, the coder codes it as “0.” If a tweet contained more than one category, it was coded in this manner.

Informative: A tweet was categorized as “Informative” when a brand post included information on the company itself, its product, brand, or services. Examples of SHARP’s informational posts include introductions of new products/services, past products/services, and articles mentioning SHARP’s products/services.

Entertaining: Entertainment posts include content that attracts the followers’ attention. Examples of SHARP’s entertainment posts include humorous tweets, introductions of products/services from other companies, and accounts for handler posts.

Promotional: Promotion posts include various offers and campaigns. Examples of SHARP promotional posts include discounts, crowdfunding, and current campaigns.

 Control Variables

We controlled for three variables: day of the week, message length, and number of followers. As timing affects engagement in tweets (Cvijikj & Michahelles, 2013; McShane et al., 2019), we controlled for the day of the week. Weekends were treated as a baseline and coded as “0” and weekdays coded as “1.” Message length was also controlled for, as the number of words was found to have an impact on the engagement that brand posts garner (Chae,‍ ‍2020; Davis et al., 2019; Lee, Hosanagar, & Nair, 2018). Therefore, we controlled for the length of the message, measured by the number of characters in each tweet. Next, we obtained the number of characters using the LEN function in EXCEL. Finally, the number of followers when each tweet was published was added as a control variable because Tafesse (2015) shows a positive relationship with brand post engagement. Data on the number of followers were obtained using Comnico Marketing Suite, a cloud service that provides a variety of free data on social media, including Twitter.

3-3.  Coding Procedure

The content of the tweets was analyzed based on a coding scheme that included a description of the operationalization of variables in the previous section. All independent variables in the conceptual framework were coded. Vividness and interactivity were manually coded because the categorization of these variables based on the criteria was clear.

Brand account personality and content variables were manually coded by two individuals, including the author and a research assistant, because the judgment about brand account personality and the categorization of tweets into three content types are complex, and inter-coder agreement rates must be calculated to ensure the reliability of the coding. The research assistant was a native Japanese trained to code the samples using a Japanese version of the coding scheme. The procedure for coding the independent variables was determined following the steps suggested by Lombard et al. (2005).

Based on their suggestions, we first chose Cohen’s kappa as an index of inter-coder reliability because this coding meets the three conditions for calculating inter-coder reliability: sample tweets are independent; the categories of these variables are independent, mutually exclusive, and exhaustive; and the coders operate independently (Cohen, 1960). The statistical software R was used to calculate the kappa statistics. The kappa value for each variable was interpreted based on benchmarks suggested by Landis and Koch (1977).

The research assistant was given instructions based on the coding scheme, and the authors answered any questions. Subsequently, the reliability of the classification was formally assessed in a pilot test using some samples, which generated acceptable kappa statistics to proceed to the next step. The reliability of the categorization was then formally calculated using 10% of the total sample; in other words, 50 tweets were randomly selected from the entire sample. The kappa statistics were 0.734 for brand account personality, 0.839 for informative content, 0.864 for entertainment content, and 0.878 for promotional content. As Landis and Koch (1977) suggest, all kappa statistics were above the substantial level (0.61–0.80) and three variables were almost perfect.

Finally, the full sample was coded by both the research assistant and the author, and any discrepancy in coding was solved through negotiation. Therefore, we moved on to the coding of the full dataset of 500 tweet samples.

3-4.  Model Specification

The dependent variables (number of likes, retweets, and replies) in the conceptual framework are count variables that follow a Poisson distribution (Coxe et al., 2009; Tafesse, 2015). Therefore, Poisson regression was selected for the R analysis. We took the natural logarithm of the three dependent variables, in line with previous studies (de Vries et al., 2012; Cvijikj & Michahelles, 2013; Sabate et al., 2014; Schultz, 2017). The model is shown in Figure 3, and a detailed explanation of each variable is provided in Table 5.

Figure 3 Poisson regression model

Table 5Descriptions of variables

Variable	Definition	Base category
log(yij)	log(y1j), log(y2j), and log(y3j) are log of likes, log of retweets, and log of replies per tweet j, respectively	–
αi	α1, α2, and α3, are constant terms	–
βf,g,b,o,e,p,d,l,n	parameters to be estimated for design and content variables	–
vivfj	dummy variable indicating whether tweet j has which level of vividness (f)	no vividness
intgj	dummy variable indicating whether tweet j has which level of interactivity (g)	no interactivity
bapj	dummy variable indicating whether brand account personality is present or not on tweet j	absent
infoj	dummy variable indicating whether tweet j is informative	no information
entj	dummy variable indicating whether tweet j is entertaining	no information
promoj	dummy variable indicating whether tweet j is promotional	no information
dayj	dummy variable indicating if tweet j is published on weekdays or weekends	weekdays
lengj	count variable that shows the character count of tweet j	–
numj	count variable indicating the number of followers when tweet j was published	–
εij	ε1j, ε2j, and ε3j are error terms for log(y1j), log(y2j), and log(y3j), respectively	–

4.  RESULTS

4-1.  Descriptive Statistics and Frequency Statistics

The descriptive statistics of the numerical variables are summarized in Table 6. On average, SHARP’s tweets received 2013.98 likes (SD = 4217.17), 408.55 retweets (SD = 1175.92), and 25.49 replies (SD = 58.44) per tweet. The average message length was 59.92 characters (SD = 36.69). The average number of followers between the periods was 824,927 (SD = 1,208.3). Concerning “day of the week,” about 94% of tweets were published on weekdays.

Table 6Descriptive statistics of numerical variables

Variable	Mean	SD	Minimum	Maximum
Likes	2013.98	4217.17	146	45632
Retweets	408.55	1175.92	6	15113
Replies	25.49	58.44	0	714
Message length	59.92	36.69	3	140
Number of followers	824927	1208.34	822534	826955

Table 7 summarizes the frequency statistics of the design and content variables. Among the 500 tweets published by SHARP between May and August 2021, 28.6% exhibited vividness and 42.8% exhibited interactivity. To be more specific, 27.2% of tweets contained a “Picture” and 1.4% of tweets have a “Video” under the vividness category. Regarding interactivity, 36.2% of tweets were linked to a website, 4.8% called for action, and 1.8% included a question. Brand account personality was present in 59.6% of the tweets. Of the 500 tweets, 47.2% were informative, 66.6% entertaining, and 7.6% promotional. The sum of the percentages was not 100, because some tweets fell into more than one content category. Of the 500 tweets, 107 exhibited two types of content; the breakdown per category combination is summarized in Table 8. There were no tweets in more than three categories in the dataset.

Table 7Frequency statistics of design and content variables

		Frequencies	Percentages
Vividness	No	357	71.4%
	Picture	136	27.2%
	Video	7	1.4%
	SUM	500	100.0%
Interactivity	No	275	55.0%
	Link to website	192	38.4%
	Call to act	24	4.8%
	Question	9	1.8%
	SUM	500	100.0%
Brand personality	Present	298	59.6%
	Absent	202	40.4%
	SUM	500	100.0%
Content variables	Informative	236	47.2%
	Entertaining	333	66.6%
	Promotional	38	7.6%
	SUM	607	121.4%

Note: The sum of content variables is not 500 because 107 tweets had two contents.

Table 8Breakdown per category combination

Combination	Frequency	Percentage
IE	88	82.2%
IP	7	6.5%
EP	12	11.2%
ALL	107	100%

Note: I-Informational; E-Entertaining; P-Promotional

4-2.  Results of the Poisson Regression Model

The Poisson regression model was tested following Tafesse (2015), who state that the assumptions of homoskedasticity and multicollinearity be verified. The standardized residuals of the three dependent variables were evenly distributed in our test. The correlation and variance inflation factor (VIF) for the independent variables were calculated, as summarized in Tables 9 and 10. The numbers indicate no multicollinearity in the sample tweets (all VIFs < 10).

Table 9The correlation of independent variables

	1	2	3	4	5	6
1. Vividness	–
2. Interactivity	−0.2382	–
3. BAP	−0.1120	−0.0977	–
4. Informative	0.2644	−0.0418	−0.5442	–
5. Entertaining	−0.0774	−0.1123	0.6355	−0.5876	–
6. Promotional	−0.0062	0.2322	−0.1484	−0.1653	−0.2130	–

Table 10VIF for independent variables

Variable	VIF	1/VIF
Vividness	1.37	0.73
Interactivity	1.46	0.69
BAP	1.89	0.53
Informative	2.36	0.42
Entertaining	2.18	0.46
Promotional	1.40	0.71
Weekdays	1.02	0.98
Followers	1.03	0.98
Message length	1.38	0.72
Mean VIF	1.56

Table 11 presents the results of the Poisson regression analysis. The model was significant overall for likes (F (12, 487) = 22.76, p < 0.001, R2 = 0.359, adj. R2 = 0.344), for retweets (F (12, 487) = 9.78, p < 0.001, R2 = 0.194, adj. R2 = 0.174), and for replies (F (12, 473) =13.29, p < 0.001, R2 = 0.252, adj. R2 = 0.233).

Table 11Poisson regression results

All posts			Log Likes β	Log Retweets β	Log Replies β
Vividness	No	(baseline)	–	–	–
	Low	Picture	0.104**	0.222***	0.076
	High	Video	–0.467****	–0.454**	–0.563***
Interactivity	No	(baseline)	–	–	–
	Low	Link website/Hashtag	–0.201****	–0.035	–0.287****
	Medium	Call to act	0.290****	0.588****	0.574****
	High	Question	–0.180	–0.066	0.400**
BAP	Absent	(baseline)	–	–	–
	Present	BAP	0.228****	0.204***	0.115*
Information		No information (baseline)	–	–	–
		Information	–0.101**	0.044	0.033
Entertainment		No entertainment (baseline)	–	–	–
		Entertainment	0.185****	0.231***	0.110
Promotion		No promotion (baseline)	–	–	–
		Promotion	–0.066	–0.008	–0.079
Control variables		Weekdays	–0.108	–0.122	–0.182**
		Followers	0.001	0.001	0.001
		Message length	0.001	0.002**	–0.001*
Constant		N	499	499	485
		F-value	22.76****	9.78****	13.29****
		R2	0.359	0.194	0.252
		Adj. R2	0.344	0.174	0.233

Note: We report standardized coefficients. N is not 500 because some observations were deleted due to missingness after taking the natural logarithm.

Signif. codes: ****p < 0.001; ***p < 0.01; **p < 0.05; *p < 0.10

 4-2-1. Design Variables

In the test on design variables, first, a low level of vividness (i.e., “picture”) was significantly and positively related to the number of likes (βpic = 0.104, p < 0.05) and retweets (βpic = 0.222, p < 0.001). However, there was no significant relationship between low levels of vividness and replies. The high level of vividness (i.e., “video”) was significantly and negatively related to the number of likes (βvideo = −0.467, p < 0.01), retweets (βvideo = −0.454, p < 0.05), and replies (βvideo = −0.563, p < 0.01), contrary to Hypothesis 1.

Second, a low level of interactivity (i.e., “link to website”) was significantly and negatively related to the number of likes (βlink = −0.201, p < 0.001) and replies (βlink = −0.287, p < 0.001), but not related to the number of retweets. The medium level of interactivity (i.e., “call-to-act) was significantly and positively related to the number of likes (βact = 0.290, p < 0.001), retweets (βact = 0.588, p < 0.001), and replies (βact = 0.574, p < 0.001). The high level of interactivity (i.e., “question”) was significantly and positively related to the number of replies (βquestion = 0.400, p < 0.05), but not related to the number of likes and retweets. Overall, the results partially support Hypothesis 2.

The presence of brand account personality was significantly and positively related to the number of likes (βBAP = 0.228, p < 0.001), retweets (βBAP = 0.204, p < 0.01), and replies (βBAP = 0.115, p < 0.10), supporting Hypothesis 3.

 4-2-2. Content Variables

Regarding our content variables, contrary to Hypothesis 4, information content was negatively related to the number of likes (βinfo = −0.101, p < 0.05). No relationship was found between the presence of information and the numbers of retweets and replies. Entertaining content was found to have a significant and positive relationship with the number of likes (βent = 0.185, p < 0.001) and retweets (βent = 0.231, p < 0.01), partially supporting Hypothesis 5. Promotional content was not related to the number of likes, retweets, or replies, which leaves Hypothesis 6 unsupported.

5.  DISCUSSIONS, IMPLICATIONS, AND FUTURE RESEARCH

5-1.  Discussions

This study examined the impact of a brand’s Twitter post characteristics on consumer engagement. The results show that different stimuli have different effects on brand post engagement.

First, Hypothesis 1 is not supported, resulting in contradictory results. This hypothesis states that tweets with a higher level of vividness achieve higher brand post engagement. The presence of a picture was found to increase the number of retweets, whereas that of a video negatively affected the number of likes, retweets, and replies. This can be explained by Luarn et al.’s findings (2015), where a brand post with videos received fewer likes compared with those including photos, suggesting that this is because of the time consumers need to process vivid information in a brand post. In short, photos are easier to react to than videos. In addition, as Sabate et al. (2014) notes, commenting requires additional effort compared to linking, which might be the reason why the presence of video negatively affected the number of replies most. Finally, the use of pictures is common in electrical manufacturer postings, so followers may have expected low level of vividness, as Schultz (2017) also explains.

In regard to interactivity in Hypothesis 2—that is, tweets with higher levels of interactivity lead to higher brand post engagement—we obtained mixed results, and the hypothesis was not supported. Linking to websites or hashtags was found to decrease the number of likes and replies, whereas calling to act increased all brand post engagement metrics—in line with previous studies (e.g., Khan et al., 2016; Schultz, 2017; Moran et al., 2019). The presence of a link leads brand posts to achieve lower engagement, possibly because clicking on it takes consumers to another page that does not interest them (Menon et al., 2019) or simply increases their risk of not moving back to the original posts (Sabate et al., 2014).

The presence of brand account personality increases the number of likes, retweets, and replies, as shown in previous studies (e.g., Cruz & Lee, 2014; Chen et al., 2015; Lee et al., 2018) in case of brand personality. This result supports Hypothesis 3 that tweets with a brand account personality increase brand post engagement. Thus, we confirmed the assumption that the effect of brand account personality is similar to that of brand personality.

Hypothesis 4 states that informational content has a negative effect on the number of likes, which contradicts the established finding of a positive effect of informational content (e.g., Cvijikj & Michahelles, 2013; Khan et al., 2016; Dolan et al., 2019). This may be due to the uniqueness of SHARP’s brand account personality. Consumers know the account is unique and has a personality, and thus expect the account to behave accordingly. As a result, they might engage less when presented with business-like information.

Partially supporting Hypothesis 5, that is, entertaining tweets lead to higher brand post engagement, entertainment content was found to positively affect the number of likes and retweets, as shown in previous studies (e.g., Cvijikj & Michahelles, 2013; Menon et al., 2019).

Finally, promotional content was found to have no effect on brand post engagement, a finding which rejects Hypothesis 6 and contradicts Menon et al.’s (2019) findings as well. This may have been due to the small sample size. Over 200 samples were classified as entertainment or informational content, but only 38 tweets were categorized as promotional content, and thus could not generate results.

5-2.  Theoretical Implications

This study adds to the body of knowledge on the design and content of brand accounts and consumer engagement in social media. Although many previous studies have investigated the outcomes of brand posts such as the number of likes, shares, and replies, scholars seldom consider brand account personality, which reflects brand personality, as a design element of brand posts. In the social media environment, brand account personality is regarded as different from brand personality (Cruz & Lee, 2014; Mizukoshi, 2019; Chu et al., 2022); hence, brand account personality as a design element is a key determinant in increasing brand post engagement. By combining existing research findings on the content of brand posts with the concept of brand account personality, we advance the present state of knowledge by shedding light on the mechanisms that underpin the social media engagement of followers who post brands.

The results revealed that vividness, interactivity, and entertainment content are significant determinants of a brand post’s popularity, as suggested by previous studies. Although consistent results could not be obtained for vividness and interactivity, the reason could be explained by a concept called “intrusiveness”. Intrusiveness is related to the degree of perceived irritation or annoyance triggered by the methods used in marketing practices. For example, when paid advertising interrupts consumer’s media content consumption, it becomes intrusive (Noguti & Waller, 2020). The impact of social media ad clutter is often beyond the control of marketers (e.g., Youn & Kim, 2019). The result for vividness and interactivity might have been affected by an intrusive element in the brand tweet, which could not be covered by the model in this study. Moreover, future research might need to consider taking vividness and interactivity as not continuous variables but breaking them down into some elements such as “video” or “link” when incorporating in the model because the result shows different effect of each and indicates the need to investigate these individually.

Brand account personality was also found to be a crucial element in increasing brand post engagement. Overall, this study contributes to the literature by testing the variables employed in previous studies and providing empirical evidence of brand account personality in the conceptual framework (e.g., de Vries et al., 2012; Menon et al., 2019).

By studying the relationship of anthropomorphizing a brand and brand account personality (personal pronouns, nonverbal cues, and information on human representatives) on engagement on Twitter, the knowledge gained can facilitate a better understanding of engagement behavior on social media in the context of Japan. Our findings also further empirical research in social media in the area of the anthropomorphization of brands—a neglected topic. Anthropomorphizing brand posts and bestowing personality on a brand account make the brand account and posts seem more human and, thus, more familiar. Followers gain comfort and reassurance when they interact with brand posts. The results of this study are consistent with the literature (e.g., Maehle et al., 2011), providing an explanation of the influencing mechanism of the design and content of brand posts on consumer engagement on social media. We thus clarify the predictive relationship among the constructs in the context of social media marketing.

Finally, the findings of the current study largely depend on the account handlers of SHARP’s Twitter accounts. The account is operated by the handler alone, so it cannot be denied that his personality traits or the way he talks were also key determinants of the engagement SHARP received on Twitter. Future research should emphasize the traits of account handlers on social media when investigating brand post engagement.

5-3.  Managerial Implications

 How to Increase the Number of Likes

Marketers can increase the number of likes for a brand post on Twitter through a call to action, exhibiting brand account personality, or entertaining content in their posts. Calling to act is the major factor that maximizes the number of likes, followed by brand account personality and entertaining content. Videos (mainly linked to YouTube videos) and links to websites or hashtags had a negative effect on the number of likes; therefore, the inclusion of these elements should be carefully considered.

 How to Increase the Number of Retweets

Marketers should focus on including pictures, a call to action, brand account personality, and entertainment content, as these characteristics increase the number of retweets on a brand tweet. A call to action had the highest impact, followed by entertaining content, pictures, and brand account personality. However, videos had a negative impact on the number of retweets; therefore, their use should be carefully considered. Although links to websites, questions, information, and promotions are visible in brand tweets, they had no effect on the number of retweets.

 How to Increase the Number of Replies

A call to action, asking questions, and brand account personality positively affected the number of replies. Both video and website links negatively affected the number of replies, and there was no relationship between the number of replies and presence of pictures, entertainment, information, or promotions. Although entertainment had a positive effect on the number of likes and retweets, the same was not true for replies, probably because replying requires more effort to engage than liking and retweeting, both of which can be done by tapping the screen.

5-4.  Limitations and Directions for Future Research

This study had two limitations. The first limitation is about the design factors in our model. Regarding vividness and interactivity, the hypothesis is not fully supported, and a new approach should be explored such as variable addition or operationalization of variables, considering intrusiveness. Although brand account personality was found to have positive effect on the brand post engagement, it can be accessed in detail by breaking down the concept into some parts, such as personal pronouns or the inclusion of nonverbal cues.

In addition, we only investigated company-specific factors (i.e., the design and content of brand tweets). Consumer-specific factors, such as Twitter users’ demographics and national characteristics, may have contributed to the engagement of brand posts, but they were not considered in this study. Younger users may prefer entertaining posts to informative posts and vice versa for older audiences. Future studies could mitigate this limitation by incorporating consumer-side factors into their conceptual frameworks.

The second limitation concerns the generalizability of the findings in the current study, which can be divided into two issues. First, it focuses on electronics manufacturers in Japan. By enriching the variation in the sample of corporate accounts, industrial and/or corporate differences in the characteristics that influence brand post engagement could be considered in future research.

The second issue was the period during which the sample tweets were collected and the sample size. This study investigated only tweets spanning a three-month period. It would be optimal to collect samples over the course of a year to strengthen our conclusions. In addition, samples should be fully garnered to cover as many variables as possible. In the present study, the number of samples for video, the level of vividness, was 7, and the number of samples for Question, one of the elements of interactivity, was low at 9. These may be the reasons why the hypothesis on vividness/interactivity was not fully supported. Future research should use larger and more diverse samples and analyze them in terms of a variety of aspects to add new knowledge.

Future studies should investigate multiple corporate accounts within the same industry and of similar sizes to check for possible differences stemming from the account handler’s specific characteristics. Therefore, an investigation of meticulously manipulated corporate accounts handled by imaginary account handlers with different characteristics in a laboratory setting may be more feasible. It would also be possible to investigate industrial differences in the effect of brand account personality on brand post engagement.

 ACKNOWLEDGEMENTS

Our special thanks are due to Mr. Takahiro Yamamoto, the account handler of SHARP’s Twitter account, for generously sharing the past tweet datasets. We are indebted to Takamasa Kajimoto for his assistance in coding the samples, and our family for continuous support. We would thank the referee and the IJMD editors for their helpful comments.

Appendix 1　The criteria for coding brand account personality

1. The information of human representatives is present (pictures, names, contact information, etc.).

The presence of the information above clearly indicates that a tweet is created by a person, and brand personality is present.

2. Uses personal pronouns (I, we, you, they, etc.)

If a tweet includes personal pronouns, it is categorized as a tweet with brand personality. However, in Japanese, one can omit subjects when readers understand what the omitted subject indicates (Sunagawa, 1990). Subject omission often occurs in SHARP’s tweets, and coders need to imagine the subject of the tweets. Mr. Yamamoto revealed that the subject of tweets by the account can be either the company itself or himself (based on interviews with Mr. Takahiro Yamamoto on September 20, 2021). Therefore, coders judge which option is the subject of every tweet. When they understand that the subject is “SHARP,” the tweet is coded 0, meaning brand personality is absent, under the brand personality column. When they understand that the subject is “the handler,” they assume that the personal pronoun, I, is present but omitted from the tweet, the tweet is coded 1, that is meaning brand personality is present.

3. Include nonverbal cues (abbreviations, emoticons, and repeated punctuation)

SHARP’s examples of abbreviations include “ripu” (abbreviation of “ripurai,” meaning “reply”), “kurafan” (abbreviation of crowdfunding). Emoticons include “”, “”, “”, for instance. Repeated punctuation refers to “!!!” or “???,” for example.

REFERENCES

•  Aaker,  J. (1997). Dimensions of Brand Personality. Journal of Marketing Research, 34(3), 347–356.
• AdverTimes (2017). SHARP SNS no naka no hito ni saji keizo syou, Osaka koukoku kyoukai syou ha kobayasi seiyaku, tokubetusyou ni hirapa- [The Saji Keizo Award for Sharp SNS account handler; the Osaka Advertising Association Award goes to Kobayashi Pharmaceutical, and a special award goes to Hirapar.]. AdverTimes. Retrieved 5 January 2022, from https://www.advertimes.com/20170315/article246303/
•  Aichner,  T. (2019). Football clubs’ social media use and user engagement. Marketing Intelligence; Planning, 37(3), 242–257.
•  Akimoto,  A. (2011). Japan, the Twitter nation. Retrieved 5 January 2022, from https://www.japantimes.co.jp/life/2011/05/18/digital/japan-the-twitter-nation/
• Allied Architects (2019). 2019 nendo “Twitter kigyou kousiki account” no riyou zittai wo tyousa [Survey on the use of “official Twitter corporate accounts” in 2019]. Allied Architects. Retrieved 10 November 2020, from https://www.aainc.co.jp/news-release/2019/01886.html
•  Araujo,  T., &  Kollat,  J. (2018). Communicating effectively about CSR on Twitter. Internet Research, 28(2), 419–431.
•  Araujo,  T.,  Neijens,  P., &  Vliegenthart,  R. (2015). What motivates consumers to re-tweet brand content? Journal of Advertising Research, 55(3), 284–295.
•  Ashley,  C., &  Tuten,  T. (2015). Creative Strategies in Social Media Marketing: An Exploratory Study of Branded Social Content and Consumer Engagement. Psychology & Marketing, 32(1), 15–27.
•  Brown,  S. (2011). It’s Alive Inside! A Note on the Prevalence of Personification. Irish Marketing Review, 21(1 & 2), 3–11.
• Bunshun Online (2018). SHARP kousiki Twitter, naze konnnani follower ga iruno? [Sharp’s official Twitter, why does it have so many followers?]. Bunshun Online. Retrieved 5 January 2022, from https://bunshun.jp/articles/-/6002
•  Chae,  M.-J. (2020). The effects of message tone and formats of CSR messages on engagement in social media. The Journal of Asian Finance, Economics and Business, 7(10), 501–511.
•  Chen,  K.,  Lin,  J.,  Choi,  J., &  Hahm,  J. (2015). Would You Be My Friend? An Examination of Global Marketers’ Brand Personification Strategies in Social Media. Journal of Interactive Advertising, 15(2), 97–110.
•  Chu,  X.,  Liu,  Y.,  Chen,  X.,  Ding,  Z., &  Tao,  S. (2022). What motivates consumer to engage in microblogs? The roles of Brand Post Characteristics and brand prestige. Electronic Commerce Research, 22(2), 451–483.
•  Chua,  A. Y. K., &  Banerjee,  S. (2015). How businesses draw attention on Facebook through incentives, vividness and interactivity. IAENG International Journal of Computer Science, 42(3), 275–281.
•  Cohen,  J. (1960). A Coefficient of Agreement for Nominal Scales. Educational And Psychological Measurement, 20(1), 37–46.
•  Coxe,  S.,  West,  S. G., &  Aiken,  L. S. (2009). The Analysis of Count Data: A Gentle Introduction to Poisson Regression and Its Alternatives. Journal of Personality Assessment, 91, 121–136.
•  Cruz,  R. A., &  Lee,  H. J. (2014). The brand personality effect: Communicating Brand Personality on twitter and its influence on Online Community engagement. Journal of Intelligence and Information Systems, 20(1), 67–101.
•  Cvijikj,  I., &  Michahelles,  F. (2013). Online engagement factors on Facebook brand pages. Social Network Analysis And Mining, 3(4), 843–861.
• DCRI (2021). “Daiikkai JDC Award” “Engagement syou zyusyou interview” SHARP kabusiki gaisya sama [“The 1st JDC Awards” “Engagement Award Winner Interview” SHARP Corporation]. Siemple digital crisis research institute. Retrieved 5 January 2022, from https://dcri-digitalcrisis.com/interview/jdc1st_sharp/
•  Davis,  S. W.,  Horváth,  C.,  Gretry,  A., &  Belei,  N. (2019). Say what? How the interplay of tweet readability and brand hedonism affects consumer engagement. Journal of Business Research, 100, 150–164.
•  de Vries,  L.,  Gensler,  S., &  Leeflang,  P. (2012). Popularity of Brand Posts on Brand Fan Pages: An Investigation of the Effects of Social Media Marketing. Journal of Interactive Marketing, 26(2), 83–91.
•  Deng,  Q.,  Hine,  M. J.,  Ji,  S., &  Wang,  Y. (2023). What makes brand social media posts engaging? An integrative framework and future research agenda. Journal of Internet Commerce, 22(1), 1–39.
•  Devereux,  E.,  Grimmer,  L., &  Grimmer,  M. (2019). Consumer engagement on social media: Evidence from small retailers. Journal of Consumer Behaviour, 19(2), 151–159.
•  Dolan,  R.,  Conduit,  J.,  Frethey-Bentham,  C.,  Fahy,  J., &  Goodman,  S. (2019). Social media engagement behavior. European Journal of Marketing, 53(10), 2213–2243.
•  Ghorbani,  M.,  Karampela,  M., &  Tonner,  A. (2022). Consumers’ brand personality perceptions in a digital world: A systematic literature review and research agenda. International Journal of Consumer Studies, 46(5), 1960–1991.
•  Hart,  P., &  Royne,  M. (2017). Being Human: How Anthropomorphic Presentations Can Enhance Advertising Effectiveness. Journal of Current Issues & Research In Advertising, 38(2), 129–145.
•  Jang,  Y.,  Chang,  S., &  Chen,  P. (2013). Exploring social networking sites for facilitating multi-channel retailing. Multimedia Tools And Applications, 74(1), 159–178.
•  Khan,  I.,  Dongping,  H., &  Wahab,  A. (2016). Does culture matter in effectiveness of social media marketing strategy? An investigation of Brand Fan Pages. Aslib Journal of Information Management, 68(6), 694–715.
•  Kim,  J.,  Kwon,  E., &  Kim,  B. (2017). Personality structure of brands on social networking sites and its effects on brand affect and trust: evidence of brand anthropomorphization. Asian Journal of Communication, 28(1), 93–113.
•  Kim,  T.,  Sung,  Y., &  Moon,  J. (2020). Effects of brand anthropomorphism on consumer-brand relationships on social networking site fan pages: The mediating role of social presence. Telematics And Informatics, 51, 101406.
•  Kusumasondjaja,  S. (2018). The roles of message appeals and orientation on social media brand communication effectiveness. Asia Pacific Journal of Marketing and Logistics, 30(4), 1135–1158.
•  Kwon,  E., &  Sung,  Y. (2011). Follow Me! Global Marketers’ Twitter Use. Journal of Interactive Advertising, 12(1), 4–16.
•  Labrecque,  L. I.,  Swani,  K., &  Stephen,  A. T. (2020). The impact of pronoun choices on consumer engagement actions: Exploring Top Global Brands’ Social Media Communications. Psychology; Marketing, 37(6), 796–814.
•  Landis,  J., &  Koch,  G. (1977). The Measurement of Observer Agreement for Categorical Data. Biometrics, 33(1), 159–174.
•  Lee,  D.,  Hosanagar,  K., &  Nair,  H. (2018). Advertising Content and Consumer Engagement on Social Media: Evidence from Facebook. Management Science, 64(11), 5105–5131.
•  Leek,  S.,  Houghton,  D., &  Canning,  L. (2019). Twitter and behavioral engagement in the healthcare sector: An examination of product and service companies. Industrial Marketing Management, 81, 115–129.
•  Lei,  S. S.,  Pratt,  S., &  Wang,  D. (2017). Factors influencing customer engagement with branded content in the social network sites of Integrated Resorts. Asia Pacific Journal of Tourism Research, 22(3), 316–328.
•  Liu,  Y., &  Shrum,  L. J. (2002). What is interactivity and is it always such a good thing? Implications of definition, person, and situation for the influence of interactivity on advertising effectiveness. Journal of Advertising, 31(4), 53–64.
•  Lombard,  M.,  Snyder-Duch,  J., &  Bracken,  C. C. (2005). Practical Resources for Assessing and Reporting Intercoder Reliability in Content Analysis Research Projects. Retrieved 12 January‍ ‍2022, from https://www.researchgate.net/publication/242785900_Practical_Resources_for_Assessing_and_Reporting_Intercoder_Reliability_in_Content_Analysis_Research_Projects/citations
•  Luarn,  P.,  Lin,  Y., &  Chiu,  Y. (2015). Influence of Facebook brand-page posts on online engagement. Online Information Review, 39(4), 505–519.
• MIC (2021). Reiwa 2 nendo zyouhou tsuusinn media no riyoujikan to zyouhou koudou ni kannsuru tyousa [2020 Survey on Information and Communication Media Usage Time and Information Behavior]. Soumu.go.jp. Retrieved 5 January 2022, from https://www.soumu.go.jp/main_content/000765258.pdf
•  Maehle,  N.,  Otnes,  C., &  Supphellen,  M. (2011). Consumers’ perceptions of the dimensions of brand personality. Journal of Consumer Behaviour, 10(5), 290–303.
•  Manzanaro,  L.,  Valor,  C., &  Paredes-Gázquez,  J. (2018). Retweet if you please! Do news factors explain engagement? Journal of Marketing Communications, 24(4), 375–392.
•  Matosas López,  L. (2018). Variables of twitter’s brand activity that influence audience spreading behaviour of branded content. ESIC Market Economics and Business Journal, 49(3), 455–499.
•  McShane,  L.,  Pancer,  E., &  Poole,  M. (2019). The influence of B to B social media message features on Brand Engagement: A fluency perspective. Journal of Business-to-Business Marketing, 26(1), 1–18.
•  McShane,  L.,  Pancer,  E.,  Poole,  M., &  Deng,  Q. (2021). Emoji, playfulness, and brand engagement on Twitter. Journal of Interactive Marketing, 53, 96–110.
•  Menon,  R.,  Sigurdsson,  V.,  Larsen,  N.,  Fagerstrøm,  A.,  Sørensen,  H.,  Marteinsdottir,  H., &  Foxall,  G. (2019). How to grow brand post engagement on Facebook and Twitter for airlines? An empirical investigation of design and content factors. Journal of Air Transport Management, 79, 101678.
•  Mizukoshi,  K. (2019). The influence of social media use on brand personality and brand account personality. Journal of Business Administration and Economics, 1(1), 1–16.
•  Moran,  G.,  Muzellec,  L., &  Johnson,  D. (2019). Message content features and social media engagement: Evidence from the Media Industry. Journal of Product & Brand Management, 29(5), 533–545.
•  Muntinga,  D.,  Moorman,  M., &  Smit,  E. (2011). Introducing COBRAs. International Journal of Advertising, 30(1), 13–46.
•  Noguti,  V., &  Waller,  D. S. (2020). ‘Motivations to use social media: Effects on the perceived informativeness, entertainment, and intrusiveness of paid mobile advertising’. Journal of Marketing Management, 36(15–16), 1527–1555.
• NTTCom Online Marketing Solutions Corporation., & Looops Communications Inc. (2015). Dai 7kai kigyo ni okeru social media katsuyo ni kansuru chosa [7th Survey on the Use of Social Media in Companies]. NTTCom Research. Retrieved January 5, 2022, from https://research.nttcoms.com/database/data/001978/
•  Pezzuti,  T.,  Leonhardt,  J. M., &  Warren,  C. (2021). Certainty in language increases consumer engagement on social media. Journal of Interactive Marketing, 53, 32–46.
•  Sabate,  F.,  Berbegal-Mirabent,  J.,  Cañabate,  A., &  Lebherz,  P. (2014). Factors influencing popularity of branded content in Facebook fan pages. European Management Journal, 32(6), 1001–1011.
•  Schultz,  C. (2017). Proposing to your fans: Which brand post characteristics drive consumer engagement activities on social media brand pages? Electronic Commerce Research And Applications, 26, 23–34.
•  Soares,  J. C.,  Sarquis,  A. B.,  Cohen,  E. D., &  Soares,  T. C. (2019). Social media marketing communication: effect of interactivity and vividness on user engagement. Brazilian Journal of Marketing, 18(4), 244–268.
• Statista (2021). Twitter: most users by country | Statista. Retrieved 29 November 2021, from https://www.statista.com/statistics/242606/number-of-active-twitter-users-in-selected-countries/
•  Steuer,  J. (1992). Defining Virtual Reality: Dimensions Determining Telepresence. Journal of Communication, 42(4), 73–93.
•  Sunagawa,  Y. (1990). Ellipsis and non-ellipsis: A case study of Japanese thematization. Bungeigengokenkyu, 18, 15–34.
•  Tafesse,  W. (2015). Content strategies and audience response on Facebook brand pages. Marketing Intelligence & Planning, 33(6), 927–943.
•  Tafesse,  W. (2016). An experiential model of consumer engagement in Social Media. Journal of Product & Brand Management, 25(5), 424–434.
•  Vargo,  C. J. (2016). Toward a tweet typology: Contributory consumer engagement with brand messages by content type. Journal of Interactive Advertising, 16(2), 157–168.
•  Wakejima,  S. (2018). Nakano hito ga kataru, “SHARP san” ga ukeire rareru riyuu: kousiki Twitter account no gokui [The account handler talks about why “SHARP san” is accepted: The secret to managing an official Twitter account]. DIGIDAY. Retrieved 5 January 2022, from https://digiday.jp/brands/dbl2018_sharp/
•  Wood,  N., &  Burkhalter,  J. (2013). Tweet this, not that: A comparison between brand promotions in microblogging environments using celebrity and company-generated tweets. Journal of Marketing Communications, 20(1–2), 129–146.
•  Youn,  S., &  Kim,  S. (2019). Newsfeed native advertising on Facebook: young millennials’ knowledge, pet peeves, reactance and ad avoidance. International Journal of Advertising, 38(5), 651–683.
•  Zanini,  M. T.,  Carbone de Moraes,  F.,  Lima,  V.,  Migueles,  C.,  Lourenco,  C., &  Reis Irigaray,  H. A. (2019). Soccer and twitter: Virtual Brand Community engagement practices. Marketing Intelligence & Planning, 37(7), 791–805.