Introduction Social media communication has become a popular way for firms to engage with customers. Research shows that firms/brands engagement with fans or customers on social media is effective to improve brand equity (Kim & Ko, 2012), drive sales (eMarketer, 2015), and enhance both transactional and relational customer behaviour (Kumar, Bezawada, Rishika, Janakiraman, & Kannan, 2016). Given the influences of social media communication, how brands effectively engage with fans or followers on social media is an important question for marketers. Most research on this topic is from the applied psychology and consumer behaviour literature, whose theories and content are dominantly tested in laboratory setting. Very few research (e.g., Lee, Hosanagar, & Nair, 2017) applied real behaviour data of field settings to study this issue. Additionally, existent research primarily focuses on social media like Facebook and Twitter in developed counties. To our knowledge, no research examines global brands social media communications in developing country, like China. Due to the policy constraints, people in mainland China have no access to foreign social media platforms. There is a local social media platform in China, named Weibo. Weibo is a NASDAQ-listed company and has nearly 100 million active users monthly. Many brands, both global and local ones, have created Weibo accounts and keep engaging with their fans. For example, there are 1,452 luxury brands and 3,707 beverage brands or firms on Weibo (Weibo Data Centre, 2017). This paper focuses on global brands’ communication practices on Weibo. The purpose of this paper is to analyse how brand posts that global brands put on their social media page are correlated with fan engagement. By using real data form Weibo, we answer the following questions: (1) what attributes of brand posts on Weibo do affect fan engagement with global brands? (2) how do these attributes affect fans engagement behaviour (i.e. liking, sharing, and commenting) differently? Conceptual Framework We decompose the attributes of global brand social media posts into five aspects, which are proposed to affect fans engagement. The first two aspects, interactivity and vividness, are derived from computer-mediated-communication research (Frotin & Dholakia, 2005; Hoffman & Novak, 1996). The next two, informative and emotional attributes, are developed from the literature of advertising (Hong, Muderrisoglu, & Zinkhan, 1987; Geuens, Pelsmacker, & Faseur, 2011). The last one is localization-related attributes, which is from global marketing communication literature (Kanso & Nelson, 2002) to capture the special characteristics of global brand posts on a local social media platform. We argue that all these five aspects of global brand posts affect fans engagement on brand page. Fans engagement is conceptualized as fans behavioural response to brand posts, which will influence attitude and behaviour of other fans. There are three kinds of fans responses on brands social media page, i.e., liking, sharing and commenting. We do control the time and date of post issued, the text length of post, whether having celebrity in post, whether related to a remarkable event, the number of followers of brand on Weibo, and product category. Interactivity 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” (Liu & Shrum 2002, p.54). Interactivity requires two-way interaction between not only customers and companies, but customers themselves (Hoffman & Novak, 1996). Brand posts on Weibo differ in the degree of interactivity. Some posts only have text, picture, or video to deliver messages of brands, which has no possibility to interact with fans. Some posts include a link that fans can click to get more information, which enhances the interactivity of communication. Other posts have questions, which stimulate interaction with fans and followers. There are posts inviting people to indicate their like or comment on social media, which are considered as high interactivity as well. Advertising research has found the positive correlation between interactivity of ads and consumers’ attitude (Coyle & Thorson, 2001). De Vries, Gensler and Leeflang (2012) found interactivity of brand post on brand fan pages partially positively related to brand popularity (measured by number of likes and comments). Empirical research (Lee et al., 2017) using Facebook data showed that having links negatively associated with customer engagement, and having questions increased comments but reduced likings. For the inconsistent findings in the literature, we re-examine this relationship by focusing on global brands on a not well-examined Chinese social media, Weibo. Vividness Vividness refers to the format richness of the message (Daft & Lengel, 1986; Fortin & Dholakia, 2005). As for brand posts, vividness is reflected by the number of sensory dimensions and senses presented (text, colours, pictures, and videos, etc.). The degree of vividness influences what and how multiple senses are stimulated (Coyle & Thorson, 2001). For example, a picture post will activate more senses than a text post because the former has colour that stimulates sight more vividly than the later. Vividness is related to but differs from interactivity. Interactivity focuses on the characteristic of two-way interaction of the communication, while vividness stands for the multiple senses stimulated by the communication. Some advertising research found that a vivid web advertisement results in higher attention and more clicks (Lohtia, Donthu, & Hershberger, 2003). As a result, we differentiate vividness of each post and predict that more vivid post associated with higher customer engagement. A vivid post attracts more likes, shares and comments. Localization-related attributes Localization-related attributes are symbols that posts have reflecting the characteristics of local culture and people. The debate of globalization (standardization) versus localization has lasted several decades. Even though a global standardized marketing strategy saves money and gains scale economy, the localization approach (Keegan, 1969) is supported by the reality that each market has unique tastes. In the international advertising literature, some scholars found that many multinational firms “plan globally and act locally” (Blackwell, Ajami, & Stephan, 1991). “Global-local dilemma” exists when global luxury brands internationalize into the Chinese market (Liu et al., 2016). The marketing communication research found that it is not advised for global brands to use the same appeals and symbols in advertising across different countries (Kanso & Nelson, 2002). In other words, the combination of global advertising theme and local communication expertise can result in enhanced effectiveness. We propose, on Weibo, people are more actively engaged to like, share, or comment global brand posts when Chinese elements, such as Chinse handwriting, Chinese festivals, and Chinese celebrity spokespeople, are present. Informative attributes Besides interactivity and vividness, which capture the format characteristics of global brand posts, content-related attributes are associated with customer engagement as well. Informative content is an important side of content attributes. One important function of brand posts is to deliver messages to customers. Research shows that on social networks people tend to have positive attitudes towards informative ads (Taylor, Lewin, & Strutton, 2011). Global brand posts with specific information should result in higher customer engagement than less informative posts. Additionally, global brand posts on social media may have different types of marketing information, such as product, price, promotion and placerelated ones respectively. The posts with varying degree of informative content may change customer engagement as well. Emotional attributes Emotional attributes are another side of the content characteristics of brand posts besides informative ones. Advertising research shows that using emotion appeal in ads is an effective way to gain people attention and generate actions (Holbrook & Batra, 1987). Emotional connections between customers and brands are considered more stable than cognitive association (Heath, Brandt, & Nairn, 2006). Some scholars find that emotional appeal on banner advertising result in positive effects on click-through rates in both B2B and B2C contexts (Lohtia et al., 2003). Empirical research (Lee et al., 2017) on Facebook shows that perceived emotion in brand posts strongly boost users’ likes and comments. Similarly, we propose that emotional attributes of global brand posts on Weibo are correlated with fans engagement. We conceptualize emotional attributes with three elements, emotional tone, emotional icon and emotional core. Emotional core reflects the type of emotions, such as humour, happiness, and love, etc. Emotional tone stands for the strength of emotion, i.e. the emotion is weak or strong. Emotional icon refers to whether the content of posts has emotional symbols, which can take form of icons or net slangs. Research Design Operationalization Dependent variables Customer engagement is operationalized as three variables, the number of likes, the number of shares, and the number of comments of each global brand post. Independent variables Vividness. Vividness is operationalized as four categories standing for different vividness degree, text only, (text and) static picture, (text, static and) animated picture, (text, picture and) video. Interactivity. Interactivity is operationalized as five 0-1 variables, having link, having question, having invitation/incentive to like, having invitation/incentive to share, having invitation/incentive to comment. Localization-related attributes. This part is operationalized as four 0-1 variables, having Chinese culture image, having Chinese culture colour, having Chinese festival, and having Chinese celebrity spokespeople (we include celebrity as a control variable). Informative attributes. This part is operationalized as seven 0-1 variables, whether a post having information of (1) brand name, (2) promotion/trial, (3) price, (4) segmentation, (5) product lunch time, (6) purchase distribution, and (7) public relation event. Emotional attributes. This part is operationalized as three variables, (1) emotional icon, a 0- 1 variable, having emotional symbols or not, (2) emotional core, a categorical variable, different type of emotions identified by surveyed respondents, and (3) emotion tone, a scalerating variable from 1 to 3, standing for none, weak and strong emotion. Control Variables. There are five control variables, (1) time of posts, including date and hour; (2) length of posts, i.e., the number of Chinese characters; (3) celebrity, whether there is a celebrity in a post; (4) event, whether a post is related to a remarkable event; (5) the number of fans of brand, and (6) product category (3 dummy variables to differentiate four categories). Data We chose 6 global brands across five product categories, specifically, beverage (Coca-cola and Starbucks), cosmetics (Olay and L’Oréal), and sports (Nike and Adidas). All these global brands created Weibo account before 2012 and have cumulated a large number of followers. We select the posts from Sept. 1, 2016 to Feb. 28, 2017, within 6 months. This time duration is long enough to get analytical data. This period covers main Chinese traditional festivals, such as Mid-Autumn Festival, National Day, and Chinse New Year, which results in more variances in localization-related variables. Data were collected through two stages. The first stage was to download raw data from Weibo’s brand pages with Internet worm program directly. Each brand has its page which contains all posts it issued and the number of people’s likes, shares, and comments for each post. The second stage is to code content and get the final dataset. Most of 0-1 variables, such as interactivity, localization-related attributes, and informative attributes were coded by two research assistants. Variables of vividness were also coded by them. Variables of emotional attributes were coded by a survey to ask 500 Weibo users. This survey-based coded method is well applied in published research (e.g., Kumar et al., 2016; Lee et al., 2017). Model The three dependent variables, the numbers of likes (y1), shares (y2), and comments (y3), are count data with a Poisson distribution. As a result, the basic model is as following: 𝑦ij = 𝛼 + 𝑒𝑥𝑝(Σ3p=1 𝛽1p 𝑣𝑖𝑣𝑖𝑑pj + Σ5q=1 𝛽2q 𝑖𝑛𝑡𝑒𝑟qj + Σ4r=1 𝛽lr 𝑙𝑜𝑐𝑎𝑙rj + Σ7s=1 𝛽4s 𝑖𝑛𝑓𝑜sj + Σ3p=1 𝛽5t 𝑒𝑚𝑜𝑡tj + 𝛽6 𝑑𝑎𝑡𝑒j + 𝛽7 ℎ𝑜𝑢𝑟j + 𝛽8 𝑙𝑒𝑛j + 𝛽9 𝑐𝑒𝑙𝑒j + 𝛽10 𝑒𝑣𝑒𝑛𝑡j + 𝛽11 #𝑓𝑜𝑙𝑙𝑜𝑤j + Σ4u=1 𝛽12u 𝑝𝑟𝑜𝑐𝑎𝑡uj) + 𝜀ij (1) Expected Results This research is among the first initiative to examine social media communication in China. We are still working on the data analysis so far. We intend to identify the influence of global posts on fans engagement, which are from not only the content of posts (informative and emotional attributes), but also the design of posts (vividness and interactivity), and especially the localization considerations. We expect to find that the content foci differ fans engagement. Posts of sales promotion should be more effective to enhance fans engagement than those of product demonstration. We can identify what form of posts stimulates fans participation more effectively. We will know whether a video post is more effective than a picture post. Most interestingly, we will know how the posts combined with Chinese cultural elements on social media are responded. For example, we could compare the difference between posts having foreign and local celebrity people. We will know how different customer engagement behaviour influence by the same characteristic of posts. We can identify the most influential factors to each of three customer responses (liking, sharing, and commenting). Comparing with the research using data from Facebook or other social media outlets, we can obtain implications guiding global brands to implement social media strategy across countries.

With the tremendous development of social media in recent years, luxury brands have welcomed social media with open arms as a means to engage with their customers. How luxury brands can make good use of social media marketing strategy is a top priority for both academicians and partitions but has not been well investigated. Many significant research gaps are present in this area warranting further explorations. For example, there is little research on the effectiveness of celebrity endorsement on social media (except for Chung and Cho’ (2017) study), even less for luxury brands. In addition, previous research on social media marketing mainly focuses on western-based platforms such as Facebook and Twitter. Given the mentioned research gaps, this study builds on social media marketing and celebrity endorsement literature and aims to investigate the key factors influencing the effectiveness of celebrity-based social media marketing efforts for luxury brands. In particular, China market and Sina Weibo are selected as our research setting. We propose the message characteristics, celebrity characteristics, celebrity-brand relationship and brand characteristics have impact on customer engagement in celebrity-based social media marketing activities. We combined two data sources to generate the key variables. First, we derived posts on the brand pages of 12 luxury brands (e.g., Cartier, Tiffany, BVLGARI and PIAGET) that were published on Sina Weibo from September 2016 to March 2017 on the middle of December 2017. Then celebrity-related posts were all picked out. The final sample includes more than five hundred celebrity-related brand posts. For each post, it contains three kinds of engagement measures (i.e., the number of Likes, Retweets and Comments) observed on Sina Weibo. They are dependent variables in our model. We built up regression models to test the hypotheses. To generate those variables that cannot be derived directly from the posts, a set of independent coders was recruited to quantify those variables. The exploratory results show that posts related to celebrities (e.g., brand ambassador/spokesman) take a significant part of posts on Sina Weibo brand pages.