Proposing to your fans: Which brand post characteristics drive consumer engagement activities on social media brand pages?

Electronic Commerce Research and Applications 26 (2017) 23–34

Contents lists available at ScienceDirect

Electronic Commerce Research and Applications journal homepage: www.elsevier.com/locate/ecra

Proposing to your fans: Which brand post characteristics drive consumer engagement activities on social media brand pages? Carsten D. Schultz University of Hagen, Germany

a r t i c l e

i n f o

Article history: Received 8 December 2016 Received in revised form 15 September 2017 Accepted 15 September 2017 Available online 21 September 2017 Keywords: Brand page Brand posts Comment Engagement Like Popularity Share Social media Social networking site

a b s t r a c t Directly engaging consumers with brand messages (posts) is one advantage of social networking sites. Using consumer engagement as a theoretical framework, the current study analyzes consumer engagement activities with brand posts, taking into account post characteristics, such as vividness, interactivity, content, and publication timing, while also controlling for post length, number of fans, and industry differences. The study identifies differences across different consumer engagement activities and industries. As such, vivid post characteristics yields mixed results, whereas post interactivity has a mainly positive effect on social interactions. If content categories address only some portion of the target audience, they negatively affect post interaction, compared to the baseline category. In terms of post publication, time at the top of the brand page increases the number of interactions, whereas weekday versus weekend has no effect on consumer engagement behavior. The findings challenge research and practice alike to account for these important differences. Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction New media have introduced significantly new ways to disseminate information, especially through new social exchanges characterized by collaboration, community, conversation, and sharing (Hennig-Thurau et al., 2010; Trusov et al., 2009). For example, social networking sites provide online platforms for networks of individual and organizational users to share and consume various types of information (Schultz, 2016). Companies use so-called brand pages of social networking sites to engage with consumers who also create some portion of the page content (Pöyry et al., 2013). Consumers who engage with brands on social networking sites have stronger relationships with those brands than consumers who do not interact via social media (Hudson et al., 2016). In addition people can create value for and with brands through social media, as reflected in the notion of consumer engagement value, such as by directly engaging in transactions, initiating recommendations, affecting purchase decisions, and providing data for market research (Kumar et al., 2010). Whereas most previous investigations of social networking sites center on modeling (online) social networks (e.g., Ansari et al., 2011; Trusov et al., 2010) or detailing the opportunities E-mail address: [email protected] https://doi.org/10.1016/j.elerap.2017.09.005 1567-4223/Ó 2017 Elsevier B.V. All rights reserved.

and risks associated with them (e.g., Champoux et al., 2012; Munnukka and Järvi, 2014; Pfeffer et al., 2014), an open question remains regarding what drives users to interact with brands on such sites. Functional, hedonic, and social values (de Vries and Carlson, 2014; Jahn and Kunz, 2012; Pöyry et al., 2013; Yang and Lin, 2014) as well as brand and social interactions and selfpresentation motives (de Vries and Carlson, 2014; Jahn and Kunz, 2012) lead people to use brand pages on social networking sites. But the interactions also depend on the characteristics of brand posts, which can drive various user behaviors, such as liking, commenting, or sharing (de Vries et al., 2012; Luarn et al., 2015; Pletikosa Cvijikj and Michahelles, 2013). Therefore, the current study seeks to extend the literature by investigating brand post characteristics, namely vividness, interactivity, and content types, as well as timing, length, number of fans, and degree of involvement in an industry. In particular, the study identifies various content types of brand posts and combines these themes with vivid and interactive post characteristics, in order to understand which elements activate and enhance engagement with these brand messages and lead to post interactions. The effect of these elements on the corresponding consumer engagement activities are analyzed by means of a regression approach following de Vries et al. (2012) and Sabate et al. (2014).

24

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

Previous research demonstrates that such social media engagement and interaction result, for example, in commitment and loyalty (e.g., Wirtz et al., 2013; Zheng et al., 2015), as well as purchase intention (Beukeboom et al., 2015; Hutter et al., 2013; Ng, 2013) and sales value (Kumar et al., 2016, 2017; Pöyry et al., 2013). In contrast to a simple (e.g., Rishika et al., 2013) or aggregated (e.g., Kumar et al., 2016) social participation value, the present study addresses consumer activities, such as liking, commenting, and sharing, referring to varying degrees of consumer engagement on a social networking site (Tsai and Men, 2013). These engagement behaviors follow traditional user differences in activity – from connection to consuming to contributing (Sashi, 2012; Tsai and Men, 2013). Beyond incorporating content types, the study considers hashtags as another interactive measure. Further dominant characteristics are post vividness and post interactions that both affect consumer engagement with brand posts on a social networking site (de Vries et al., 2012; Luarn et al., 2015; Pletikosa Cvijikj and Michahelles, 2013). As such, this study addresses the question of which characteristics and content drive liking, commenting, and sharing behavior with respect to brand posts on social networking sites. These findings help explain a degree of consumer interactions with brand posts and allow social media managers to plan their social media strategies accordingly, in order to obtain such positive outcomes as commitment, loyalty, purchase intention, and sales value. The next section briefly discusses the theoretical framework and prior literature on brand post interactions. These insights lead on to the hypotheses and research framework. After presenting the data and method, this article presents and discusses the empirical results. Finally, the article concludes with some limitations and directions for further research.

2. Literature review and hypotheses development 2.1. Theoretical framework The present study uses consumer engagement as a theoretical framework. Following a behavioral focus, consumer engagement is defined as ‘‘behavioral manifestations that have a brand or firm focus, beyond purchase, resulting from motivational drivers (Van Doorn et al., 2010, p. 254).” Based on the behavioral dimension, Kumar et al. (2010) propose the construct of consumer engagement value including the consumer lifetime, referral, influencer, and knowledge value (Kumar et al., 2010). Brodie et al. (2011) subsequently conceptualize five aspects of consumer engagement. Firstly, consumer engagement encompasses the psychological state that occurs by virtue of interactive, cocreative customer experience with a focal agent/object (e.g., a brand). Secondly, consumer engagement occurs within a dynamic, iterative relationship. Thirdly, no relational occurrences are isolated; but they are interdependent. Fourthly, consumer engagement is multidimensional reflecting various cognitive, emotional, and behavioral dimensions. Lastly, consumer engagement differs across individuals and situations. One aspect to note is that even though consumer engagement is individually different and may vary in certain situations, brand managers are interested in the predominant reactions of their target groups. Consumer engagement has drawn increasing research interest and has consequently been addressed in social media as well (e.g., Brodie et al., 2013; Kabadayi and Price, 2014; Zheng et al., 2015). ‘‘[S]ocial media has provided firms – both large and small – with a new tool for customer engagement (Rishika et al., 2013, p. 114).” This study is particularly related to the outcome of consumer engagement that results in brand post interactions (Pletikosa Cvijikj and Michahelles, 2013; Wirtz et al., 2013). Based on different

levels of consumer engagement (Schultz, 2016; Tsai and Men, 2013), brand post interactions also represent various behavioral activities. Consequently, treating brand post interactions uniformly, as in measures like post interaction rate,1 neglects the behavioral differences and subsequently their varying outcomes. Consumer engagement not only affects brand interactions, but also community and brand commitment, satisfaction, and loyalty (Wirtz et al., 2013). As potential outcomes, Kristofferson et al. (2014) caution liking publicly, such as on a social networking site, which may not lead to engagement in subsequent relevant behavior, as demonstrated by their experimental studies concerning a charitable organization. However, engaged consumers also exhibit loyalty, satisfaction, empowerment, connection, trust, and commitment (Brodie et al., 2013). Economic, entertainment, and social benefits mediate consumer loyalty and satisfaction induced by consumer engagement (Gummerus et al., 2012). Similarly, perceived benefits and costs lead to engagement with a brand page, that in turn enhances online commitment and consequently, brand loyalty (Zheng et al., 2015). While positive brand evaluations may precede consumer engagement with brand pages and brand messages, Beukeboom et al. (2015) indicate that engagement with a brand page in a social networking site also positively affects brand evaluations and purchase intention. Consumer engagement in social media may not only lead to purchase intention, but also affect profitability (Kumar et al., 2010, 2016; Rishika et al., 2013). However, Kaptein et al. (2016) caution that increased engagement leads to increased spending for inactive members only, whereas already engaged members decrease spending. Beyond, for example, sales value (Kumar et al., 2016; Pöyry et al., 2013), engaged consumers create value as co-creators in business processes, such as for market research, product development, and recommendations (de Vries and Carlson, 2014; Kabadayi and Price, 2014). Consumer engagement in a brand community results in brand post interactions (Pletikosa Cvijikj and Michahelles, 2013; Wirtz et al., 2013), which precedes different value dimensions (Kabadayi and Price, 2014; Kumar et al., 2010; Sashi, 2012). Accordingly the present study seeks to understand how brand post characteristics increase engagement in a social networking site measured by brand interactions. The present study draws in particular on different levels of consumer activities, such as liking, commenting, and sharing (Luarn et al., 2015; Pletikosa Cvijikj and Michahelles, 2013; Schultz, 2016). In this way, the study demonstrates that these behaviors differ across various brand post characteristics.

2.2. Brand post vividness and interactivity Beyond content type categories, post vividness and post interactions are two dominant characteristics affecting consumer engagement with brand posts on a social networking site (de Vries et al., 2012; Luarn et al., 2015; Pletikosa Cvijikj and Michahelles, 2013). Before the next section addresses previous research results on brand post interactions, we briefly introduce both concepts. Vividness and interactivity both refer to media characteristics of brand posts. Vividness reflects the degree to which information addresses various senses. Different media types, such as text, images, and videos, thus represent different levels of media richness (de Vries et al., 2012; Luarn et al., 2015). Similarly, as pictures augment purely textual information, videos explicitly affect audio-visual processes. As vivid media activate users, they resonate with 1 Post interaction rate is, for example calculated as the sum of interactions (likes, comments, and shares) divided by post reach (number of fans).

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

engaged consumers and lead to brand interactions (de Vries et al., 2012; Luarn et al., 2015; Pletikosa Cvijikj and Michahelles, 2013). The present study specifically adopts the four level (no, low, moderate, high) scale of de Vries et al. (2012). Interactivity refers to the degree to which brand posts animate users to respond. Liu and Shrum (2002, p. 54) define interactivity 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.” Different interactive brand posts elements include links to a different website, votings, calls to action, contests, and questions. One extension of previous research is the inclusion of hashtags as another interactive element. Hashtags are a way for brands and consumers to specifically tag content and allow other users on the social networking site to find similar tagged content. Practitioners express mixed feelings about the effect of hashtags in Facebook brand posts. In line with previous research, the present study analyzes levels of no, low, moderate, and high interactivity (de Vries et al., 2012), since interactive post characteristics activate users to engage with brand messages by liking, commenting, and sharing behavior. As such, the present study adopts a black box approach, considering vividness and interactivity characteristics as inputs, and brand interactions as outputs. The consumer as the black box is not investigated further. However, it is important to note that epistemic, social, and hedonic values may drive such consumer engagement outcomes (Yang and Lin, 2014). Consequently, brand post content may resonate with engaged users, in order to serve different value elements (de Vries and Carlson, 2014; Jahn and Kunz, 2012; Yang and Lin, 2014). 2.3. Brand post interactions The ability to engage consumers directly through brand messages (i.e., posts) is a benefit of social networking sites (de Vries et al., 2012; Gomez-Arias and Genin, 2009) and a key dimension of brand social media strategy (Schultz, 2016). Thus, the characteristics of brand posts probably affect consumer interactions and help create consumer engagement (e.g., de Vries et al., 2012; Luarn et al., 2015; Sabate et al., 2014), which in turn may yield purchase intentions (Beukeboom et al., 2015; Pöyry et al., 2013) and business value (Kabadayi and Price, 2014; Kumar et al., 2017). Pletikosa Cvijikj and Michahelles (2011) analyze the post type, content category, and day of the week. Both post type and category are significant for the duration of the interaction as well as the likes and comments ratios – defined as the number of likes and comments adjusted by the number of fans. They also find that posts on Tuesdays and Thursdays prompt different numbers of comments. In a subsequent study, vivid posts with entertaining or informative content are shown to increase the likes ratio, whereas interactivity, remuneration, and post timing decrease it (Pletikosa Cvijikj and Michahelles, 2013). The authors also find that interactive posts decrease the comments ratio, but the ratio is positively affected by any content at all. The shares ratio increases with entertaining posts that contain pictures or videos, but decreases during peak hours and for status posts. Users interact longer with entertaining, informative, and vivid content posted during peak hours (Pletikosa Cvijikj and Michahelles, 2013). In their influential study, de Vries et al. (2012) show that vivid and interactive posts, as well as the share of positive comments, influence the number of likes. By contrast, the number of comments is affected by interactive brand posts, such as questions, and the share of positive and negative comments. Based on 164 posts from five Spanish travel agencies, both image and video are shown to influence the number of likes, but neither links nor timing do so (Sabate et al., 2014). In contrast, the number of comments

25

is affected by images and timing. Across 10 brand pages, Luarn et al. (2015) find that a moderate level of vividness generates the highest level of engagement. User engagement also increases with higher interactivity. If brand posts offer remuneration, information, entertainment, and sociability, users are more likely to like the posts. However, social content creates more comments than entertainment, information, and remuneration. Sharing is evoked predominantly by entertaining and informative posts, rather than remuneration or social ones. Posts including images and a moderate amount of text receive the highest number of likes, adjusted by the number of fans (Trefzger et al., 2016). The authors conclude that brand posts should allow fast processing. Table 1 summarizes the related research on post characteristics affecting consumer engagement behavior. 2.4. Hypothesis development This study seeks to explain brand post interactions, in the form of likes, comments, and shares, on the basis of post vividness, interactivity, and content, as well as post timing (i.e., length of time the post is positioned at the top of the page, and weekdays versus weekend). This study also controls for post length and number of fans of the corresponding brand page. The analysis extends to the post content category, includes hashtags as an interactivity aspect, and takes place in the domain of apparel and food retailing. Brand posts can be categorized as status updates, photos, albums, links, events, or videos. The different media types appeal to user senses to varying degrees, which indicates their vividness (de Vries et al., 2012; Luarn et al., 2015). However, the ideal degree of media vividness remains a topic of debate, and might depend on different interaction measures (Pletikosa Cvijikj and Michahelles, 2011). For example, moderate (link) and high (video) vividness appear to invoke more engagement than no (status) and low (picture, album) vividness (Luarn et al., 2015). Yet, a moderate level outperforms a high level of vividness, which is surprising, in that links in a post direct users away from the brand page and toward other content. In contrast, links may have a negative effect on the number of comments (Sabate et al., 2014). Other studies indicate positive effects of low (photo) and high (video) vividness on the number of likes (de Vries et al., 2012; Sabate et al., 2014; Trefzger et al., 2016), and the number of comments appear to be affected positively by low vividness (photo) posts (Sabate et al., 2014). The difference between picture and video posts might reflect the varying amounts of time that users need for processing (Sabate et al., 2014; Trefzger et al., 2016), although the quality of the media might also affect post interactions. Even though the results differ, the relevance of post vividness is generally supported. Therefore, this study predicts: H1. The degree of brand post vividness has a positive effect on post interaction. Interactivity characteristics animate users to respond to brand posts. The level of post interactivity has been studied in previous research and yielded in mixed findings (de Vries et al., 2012; Luarn et al., 2015; Pletikosa Cvijikj and Michahelles, 2013). For example, Pletikosa Cvijikj and Michahelles (2013) show that low interaction posts (status and photo) create more likes and comments than highly interactive ones (link and video). The use of questions, which constitutes a highly interactive element, leads to more likes but fewer comments (de Vries et al., 2012). With the same operationalization, Luarn et al. (2015) find that higher interactivity generally increases the probability of liking, commenting, and sharing. Higher interactivity increases the probability of user engagement, such as liking, commenting and sharing, (de Vries et al., 2012; Luarn et al., 2015), though interaction time

26

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

Table 1 Overview of literature on brand post characteristics affecting consumer interactions. Study

Number of Posts

Brand Pages

Industry

Post Type

Content Type

Post Timing

Additional Variable

Dependent Variable

Method

Pletikosa Cvijikj and Michahelles (2011) de Vries et al. (2012)

120

1

Consumer goods

Media

Category

Day of the week

–

Likes ratio, comments ratio, duration

Kruskal-Wallis non-parametric ANOVA

355

11

Diverse

Vividness, interactivity

Entertainment, information

OLS regression

5035

100

Food & beverage

Vividness, interactivity

Entertainment, information, remuneration

Industry, post length, valence of comments Manufacturer vs. retailer1

Likes, comments

Pletikosa Cvijikj and Michahelles (2013) Sabate et al. (2014)

Top of page, weekday Peak hours, weekday

Negative binomial regression

164

5

Travel agency

Media

–

Fan number, post length

Luarn et al. (2015)

1030

10

Diverse

Vividness, interactivity

–

Likes, comments, shares

ANOVA

Trefzger et al. (2016) This study

560

3

Automotive

Media

Entertainment, information, remuneration, social –

Business hours, weekday –

Likes ratio, comments ratio, shares ratio, duration Likes, comments

–

Post length

Likes ratio

ANOVA

792

13

Apparel & food retailing

Vividness, interactivity

Top of page, weekday

Industry, fan number, post length

Likes, comments, shares

OLS regression

category

OLS regression

Notes: ANOVA = analysis of variance, OLS=ordinary least squares. 1 Secondary analysis.

may also induce a negative effect (Pletikosa Cvijikj and Michahelles, 2013). In turn, modeling various interactive elements, this study proposes:

may only appeal to part of the target audience (de Vries and Carlson, 2014; Jahn and Kunz, 2012; Pöyry et al., 2013; Yang and Lin, 2014). Therefore, this study proposes:

H2. The degree of brand post interactivity has a positive effect on post interaction.

H3. Brand post content types which address the majority (minority) of the target audience have a positive (negative) effect on post interaction in comparison to product posts.

The content of brand posts has been captured primarily by assigning posts to selected categories of general content types, such as information versus entertainment posts (de Vries et al., 2012), possibly supplemented by remuneration posts (Pletikosa Cvijikj and Michahelles, 2013) and social posts (Luarn et al., 2015). The findings are mixed. De Vries et al. (2012) indicate that neither informative or entertaining posts affect the number of likes or comments. In contrast, the likes and comments ratio depend on entertaining, informative, and remunerating posts; entertaining and informative posts also influence the shares ratio (Pletikosa Cvijikj and Michahelles, 2013). Based on their ANOVA results, Luarn et al. (2015) identify varying effects of content types, from high to low: likes (remuneration, information, entertainment, social), comments (social, entertainment, information, remuneration), and shares (entertainment, information, remuneration, social). Another approach distinguishes between brand post topics. For a travel agency, posts are related to travel tips, promotional offers, and other marketing campaigns (Pöyry et al., 2013). Similarly, posts are categorized as related to events, statements, charity, and products in a wine setting (Rishika et al., 2013). Pletikosa Cvijikj and Michahelles (2011) cite announcements, information, questions, questioners, promotions, and statements. These topics are provided by a social media manager. The categories significantly affect the number of likes and comments. The present study therefore adopts the second approach, together with content analysis. The reference group is the product category that presents information on specific products. As this category presents the expected baseline in a retail setting, the study proposes that other post categories are increasingly able to activate the audience and in turn enhance engagement and interaction behavior. However, we also note that certain information, such as events and promotions,

The timing of brand posts is relevant for user interactions with the content, so that prior research has addressed various aspects of post timing. For example, Pletikosa Cvijikj and Michahelles (2013) differentiate between low and peak-user activities, and determine that peak-user activity time negatively affects engagement, as measured by likes and shares. However, user activity and post timing are not independent. Sabate et al. (2014) instead distinguish between business and non-business hours and find that posting during business hours has a positive effect only on the number of comments. For global brands and their pages, spanning different time zones, this distinction may not be applicable. The duration at the top of the page depends on the posting strategy and can be scheduled accordingly (de Vries et al., 2012). According to findings involving other online instruments, the longer the exposure of a brand post at the top of the page, the more likes and comments it receives (de Vries et al., 2012). Accordingly: H4a. The longer a brand post is positioned at the top of the page, the higher the number of post interactions. Another aspect of post timing is the day of the week. Previous research contrasts weekdays with weekends (de Vries et al., 2012; Pletikosa Cvijikj and Michahelles, 2013; Sabate et al., 2014). User activities on Facebook might be higher on weekdays (Golder et al., 2007), but de Vries et al. (2012) and Sabate et al. (2014) find no significant effect, and Pletikosa Cvijikj and Michahelles (2013) even identify a negative effect of weekdays on the number of likes, but a positive effect on the number of comments and no effect on the number of shares. As social media accessibility increases for example through mobile devices, this present study expects no weekday versus weekend effect.

27

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

Vividness

H1

Interactivity

H2

Content type

H3

Top position

H4a

Weekday

H4b

Likes

Industry

Comments

Post length

Shares

Fan number

Fig. 1. Conceptual framework.

Table 2 Data overview.

*

Industry

Brand

Fans*

Posts

Likes

Comments

Shares

Apparel retailer

C&A Ernsting’s family Esprit H&M Primark Zara

5,426,106 161,156 1,200,880 19,078,157 2,791,462 21,443,252

42 64 33 193 71 12

6317 5700 16,954 528,118 90,740 70,153

124 3766 289 4049 4468 1284

290 593 583 7718 2404 1615

Food retailer

Aldi Süd Edeka Kaufland Lidl Penny Real Rewe

110,691 281,355 287,277 1,762,177 59,990 419,717 168,598

41 35 42 107 29 66 63

6510 6784 12,058 30,525 5684 10,378 3502

376 1494 5303 7925 593 2107 487

153 952 1335 3513 412 1515 572

As at May 31, 2014.

H4b. Post published on weekdays lead to neither more nor less post interactions than post published on weekends. Furthermore, de Vries et al. (2012) find users to comment less in the accessories and cosmetics product category, compared to the food category, and no differences for liking behavior across the six product categories analyzed. Additionally, some differences exist in a food and beverage setting between manufacturers and retailers (Pletikosa Cvijikj and Michahelles, 2013). In commenting behavior, posting time significantly increases engagement for retail brands. For manufacturers, content type enhances sharing behavior. In contrast, the present study addresses two different retail industries: apparel and food. In general, apparel products are assumed to be more involving for the majority of target groups than products of food retailers. To account for potential differences, the present study controls for the industry. Brand posts from food retailers are used as the reference group. Finally, in line with previous research, this study controls for post length and number of fans. Post length affects the amount of information and user ability to process it (de Vries et al., 2012; Sabate et al., 2014; Trefzger et al., 2016). Previous results indicate either no significant effect (de Vries et al., 2012), a positive effect on the number of likes (Sabate et al., 2014), or a benefit from a moderate level of text that supports fast processing (Trefzger et al., 2016). More fans of a brand page should increase the likelihood of user engagement, liking, commenting, and sharing (Sabate et al., 2014). Fig. 1 displays the conceptual framework for this study.

well-established, active on a social networking site: C&A, Ernsting’s family, Esprit, H&M, Primark, and Zara. This industry was chosen in line with previous research; both brands and consumers are highly engaged and likely to participate in social interactions (Goh et al., 2013; Park and Cho, 2012; Schultz, 2016). The apparel brands represent an explorative sample of a variety of retail types and sizes. For example, Esprit, H&M, and Zara are globally represented brands; C&A and Primark are located primarily in Europe, and Ernsting’s family is a retailer in Austria and Germany. In comparison, the food retailing industry is generally less involving from a consumer perspective – exceptions revolve, for example, around individual food incompatibility or food scandals. Consequently, this study assumes food retailing to be less engaging than apparel retailing. The seven food retail brands represent all German food retailers active on the social networking site, based on an 80% market share in 2014 (TradeDimensions, 2015). The food retail brands are Aldi Süd, Edeka, Kaufland, Lidl, Penny, Real, and Rewe. The brand pages are those that appear on Facebook, the leading global social networking site, which all retail brands use. Facebook reported 1860 million active users per month for the fourth quarter of 2016 and revenue of $27,638 million for 2016, a 54.2% increase from $17,928 million in 2015 (Facebook, 2017). For all retail brands, all brand posts and the resulting user interactions on Facebook were recorded from April 14 to May 31, 2014. The 798 brand posts that were retrieved, prompted a total of 793,423 likes, 32,265 comments, and 21,655 shares. Table 2 contains a data overview.

3. Methodology 3.2. Operationalization of variables 3.1. Data collection and sample The study draws on data from two retail industries, namely apparel and food retailing. The six apparel retail brands are

The study operationalizes four levels of vividness according to the types of posts (de Vries et al., 2012; Luarn et al., 2015; Pletikosa Cvijikj and Michahelles, 2013). Status posts offer no

28

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

Table 3 Topic categories.

*

Topic Category

N

Description

Example

Charity

12

Information on a charity event

Competition

98

Posts presenting a competition

Content

39

Informative posts not directly related to a product, such as recipes and nutrition facts

Download the ultimate summer hit ‘‘Heart of Glass” by Gisele Bündchen and Bob Sinclar – all royalty will go to charity. #COMPETITION: Which Marvel character would you most like to be and why – Thor, Captain America, The Hulk, Iron Man or Spiderman. . .? Tell us below for a chance to #WIN one of 25 Marvel goodie bags. Our noodle-rucola salad with strawberries not only tastes delicious, but is also

Coverage

74

Holiday

36

Reports on events directly related to neither products nor promotions, such as photo shoots or global sports events. Posts referring to a holiday, season, or weekend

HR

13

Information about recruiting events

Product Promotion

400 104

Statement

22

Posts related to specific products Promotional activities, such as discounts, sales, or store openings. Posts stating an opinion on a topic, such as family values and sustainability.

prepared lightning fast. You can find the recipe here.* We LOVE all your Dream Team posts! Who’ll you wear yours with for the World Cup? We wish you and your loved ones a beautiful and Happy Easter! How will you spend the holidays? Hello everyone, this weekend we are at the job fair Münsterland and will gladly inform all interested about career opportunities. We look forward to interesting conversations!:-) * Here are our best ‘‘New Arrivals” of the week! Which style do you prefer? Relive the opening of our first Australian Flagship store in Melbourne, featuring the band Haim as DJ’s for the #HMAustralia launch! Hands down, we’ve got the cutest fans!!

Translated message.

vividness, containing only short text messages. Pictorial posts, such as photos and albums, represent a low level of vividness. Events and links reflect a moderate level of vividness, providing additional information after a click. Finally, video posts provide high vividness. As noted, interactivity might be operationalized in terms of media posts (Pletikosa Cvijikj and Michahelles, 2013) or the post characteristics (de Vries et al., 2012; Luarn et al., 2015). This study adopts the latter approach and records the following post characteristics: links, hashtags, calls to action, competitions, questions, and votings. In turn, it is possible to define four levels of interactivity (de Vries et al., 2012). The base-line of no interactivity level implies that none of the characteristics is present. A link to a website or hashtag represents a low level of interactivity. A moderate level of interactivity involves a call to action, such as visiting a certain website, commenting on an event or product, checking out an announced promotion, or joining a contest. Finally, questions and vote options represent a high level of interactivity. To capture the content of a brand post, the brand posts are assigned to categories associated with charity, competition, content, coverage, holiday, human resources, products, promotions, or statements. Table 3 provides an overview of these topic categories. For the time period a post is positioned at the top of the brand page, the measure assesses the span in seconds, until the next post appears. For apparel retail posts, the time until the next posting could not be calculated for the last 6 posts, so that the empirical analysis is based on 792 of 798 brand posts. The day of the week and number of fans both correspond to the publication date. Post length can be measured as the number of characters (Sabate et al., 2014), words (de Vries et al., 2012), or text lines (Trefzger et al., 2016); the current study uses the number of characters. Finally, user interactions are measured by the number of likes, comments, and shares on a brand post. 3.3. Method Previous studies use analysis of variance or regressions to analyze post interactions (Table 1). This study takes a regression approach to determining the degree to which a post characteristic affects post interactions. The dependent variables yi are the number of likes, comments, and shares, representing count data with

a Poisson distribution. In line with previous research (de Vries et al., 2012; Pletikosa Cvijikj and Michahelles, 2013; Sabate et al., 2014), the study relies on the natural logarithms of the dependent ~i ¼ lnðyi þ 1Þ. The model for the brand variables, calculated as y posts j is:

~i;j ¼ b0;i þ y

3 6 4 X X X b1;i;f v iv idi;f ;j þ b2;i;g iai;g;j þ b3;i;h topici;h;j f ¼1

g¼1

h¼1

þ b4;i positioni;j þ b5;i weekdi;j þ b6;i industryi;j þ b7;i lengthi;j þ b8;i fansi;j þ ei

ð1Þ

where j, is data on brand post j; i, refers to either the like, comment, or share model; ~i , refers to the logarithm of the three dependent variables, the y number of likes, comments, and shares per brand post; vividi,f, dummy variable for post vividness which includes picture, event, and video, with no vividness as the reference category; iai,g, dummy variable for post interactivity which includes link, hashtag, call to act, question, contests, and voting, with no interaction as the reference category; topici,h, dummy variable for post content type which includes charity, competition, content, coverage, holiday, HR, promotion, and statement, with product as the reference category; positioni, duration in seconds for which the brand post is positioned at the top of the page; weekdi, dummy variable if the brand post is published on a weekday; industryi, dummy variable if the post is from an involved industry; lengthi, post length of the brand post measured by the number of characters; fansi, number of fans on the publication date of the brand post; and ei, error term for the like, comment, and share model. All three models assume a uniform distribution across all brands and its fans. Even though stronger brand relationships of social media users are, for example, found across three national groups (France, UK, USA), the authors also recommend managers

29

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34 Table 4 Standardized estimation results for post interactions. ln Like

ln Comment

ln Share

Vividness

No Low Moderate High

(reference) Picture Event Video

– 0.103** 0.353*** 0.005

– 0.151*** 0.215*** 0.171***

– 0.138** 0.071 0.110***

Interactivity

No Low

(reference) Link Hashtag Call to act Contest Question Voting

– 0.042 0.102*** 0.076** 0.139*** 0.004 0.126***

– 0.120*** 0.196*** 0.152*** 0.256*** 0.146*** 0.241***

– 0.008 0.043 0.117*** 0.192*** 0.054 0.167***

Moderate High Content type

Product (reference) Charity Competition Content Coverage Holiday Human Resources Promotion Statement

– 0.133*** 0.234*** 0.014 0.025 0.011 0.104*** 0.117*** 0.023

– 0.070** 0.127** 0.019 0.009 0.033 0.089*** 0.035 0.113***

– 0.015 0.093* 0.057 0.038 0.078** 0.068** 0.089** 0.046

Timing

Position Weekday

0.094*** 0.012

0.154*** 0.009

0.177*** 0.017

Control

Post Length Fans

0.083** 0.312***

0.036 0.027

0.057 0.197***

4.374*** 29.531*** 0.446 0.431

2.738*** 16.557*** 0.311 0.292

0.054 7.280*** 0.166 0.143

Unstandardized Constant F-Value R2 Adjusted R2 ***

p < 0.01, **p < 0.05, *p < 0.1.

to consider the degree of brand characteristics and uncertainty avoidance motivations of users (Hudson et al., 2016). In the present models, such differences are not explicitly included, but are partly accounted for via the industry and the variance of fans.

4. Results In the present data, brands and the correspondingly industry (industry) are correlated with the number of fans (fans). Consequently, the industry dummy is excluded from the analysis, as the number of fans potentially explains more variance in all thee models. The standardized estimation results for the like, comment, and share model are presented in Table 4. The variance inflation index for all included independent variables is below the threshold value of 5. The highest variance inflation index is 3.5 for the inclusion of a picture. Thus, there is no indication of high multicollinearity in the results presented. The adjusted R2 is 0.431 (R2 0.446) in the like model. The (ln) number of likes is significantly positively affected by picture, hashtag, call to act, contest, and voting. In comparison to the product category, no category has a positive effect on likes. Time at the top of the page and fan number positively enhance the liking behavior. Events and post length, as well as the content categories of charity, competition, human resource, and promotion have significantly negative effects. The comment model explains 31.1% (R2) of the comment variance, with an adjusted R2 of 0.292. The categories hashtag, voting, call to act, contest, question, voting, and statement, as well as time at the top of the page, significantly increase the (ln) number of comments. All vivid characteristics (pictures, events, and videos), links as interactive characteristic, and the categories charity, competition, and human resources have negative impacts. For the share model, the explained variance R2 is 0.166 (adjusted R2 0.143). The categories picture, video, call to act,

contest, voting, holiday and promotion, as well as time at the top of the page and number of fans, have significant positive effects on the (ln) number of shares. Competition and the human resources category negatively affect sharing behavior. In turn, post vividness indicates significant effects on likes, comments, and shares. All post vivid characteristics have negative effects on comments. Additionally, events negatively impact on the number of likes. Consequently, Hypothesis 1 is rejected. However, the retail brands predominantly use pictures in their brand posts (n = 723, 90.6%), events (n = 15, 1.9%) and videos (n = 35, 4.4%) are used infrequently. Thus, this finding requires cautious interpretation, even if it confirms some prior results (de Vries et al., 2012; Sabate et al., 2014). Other than linking to a website, post interactivity characteristics enhance interactions, generally supporting Hypothesis 2. Links lead users away from brand posts, although they do not increase interaction behavior, and they also do not appear to decrease the likelihood of interacting. As an extension to previous research, hashtags provide further insights. They are similar to links, in that they provide a means to access additional information (posts with the same hashtag), but they also differ from links, because users remain on the social networking site. In the present study, a total of 207 posts include at least one hashtag that positively affect the number of likes and comments. The study also extends previous insights into content types, yielding several new and related categories. The reference category presents and informs about specific products without promotional elements. In comparison to the product category, both content and coverage categories had no effect on either interaction metric. The majority of compared content categories negatively impact on brand interaction behavior. As charity, competition, and human resources only appeal to some of the target audience, these results partially support Hypothesis 3. However, the positive effects – statements on comments as well as holiday and promotion on shares – represent individual results that do not support

30

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

Table 5 Standardized estimation results for post interactions for apparel and food retailers. ln Like

ln Share

Apparel

Food

Apparel

Food

Apparel

Food

Vividness

No Low Moderate High

(reference) Picture Event Video

– 0.274* 0.340*** 0.131

– 0.176** 0.029 0.022

– 0.135 0.157 0.100

– 0.164** 0.037 0.238***

– 0.334* 0.008 0.274**

– 0.158** 0.004 0.147**

Interactivity

No Low

(reference) Link Hashtag Call to act Contest Question Voting

– 0.063 0.146*** 0.090** 0.089** 0.024 0.132***

– 0.212*** 0.138 0.042 0.089 0.059 0.016

– 0.019 0.166*** 0.179*** 0.270*** 0.117** 0.262***

– 0.221*** 0.083 0.111** 0.077 0.200*** 0.141***

– 0.042 0.140*** 0.150*** 0.179*** 0.069 0.213***

– 0.005 0.198** 0.098* 0.084 0.055 0.044

Moderate High Content type

Product (reference) Charity Competition Content Coverage Holiday Human Resources Promotion Statement

– 0.146*** 0.241*** – 0.041 0.065* 0.102*** 0.130*** 0.047

– 0.037 0.019 0.011 0.097* 0.175*** 0.080 0.017 0.070

– 0.108** 0.223*** – 0.022 0.116*** 0.051 0.007 0.025

– 0.065 0.189** 0.015 0.024 0.060 0.098** 0.155*** 0.228***

– 0.025 0.196*** – 0.056 0.050 0.052 0.028 0.008

– 0.106** 0.201** 0.094 0.063 0.313*** 0.054 0.263*** 0.140**

Timing

Position Weekday

0.020 0.026

0.191*** 0.033

0.129*** 0.014

0.192*** 0.048

0.137*** 0.009

0.087* 0.013

Control

Post Length Fans

0.220*** 0.051

0.001 0.249***

0.160*** 0.041

0.036 0.267***

0.231*** 0.104**

0.072 0.201**

9.307** 22.044*** 0.532 0.508

3.951*** 6.121*** 0.263 0.220

4.329 11.537*** 0.373 0.341

2.348*** 10.925*** 0.389 0.353

9.491** 7.941*** 0.290 0.254

0.512 4.432*** 0.205 0.159

Unstandardized Constant F-Value R2 Adjusted R2 ***

ln Comment

p < 0.01, **p < 0.05, *p < 0.1.

Hypothesis 3. Specifically, promotion posts create a negative effect on likes, but a positive one on shares. This finding indicates that promotional content may not activate users to like, but to actively recommend sales and discounts. Regarding the time a brand post is positioned at the top of the brand page, the results indicate significant positive effects on liking, commenting and sharing behavior, thus confirming Hypothesis 4a. On average, a brand post is positioned at the top of the page for 18.81 h (SD = 24.83). Naturally, the question remains as to how much time different content needs, in order to enable consumers to engage with brand posts. Posting on weekdays versus weekends does not affect brand post interactions, as indicated by Hypothesis 4b. In total, brands posted 86 messages on weekends versus 712 posts during the week. Consequently, the data are skewed toward weekday posts. Moreover, post length negatively affects only the number of likes. The average post is 196,4 characters long (SD = 151.65). Previous research finds a medium number to drive the highest number of likes (Trefzger et al., 2016). Fast processing capability and time may explain this result. In turn, more engaging activities, such as comments and likes, may be less prone to this effect. Alternatively, increasingly engaged consumers are willing to invest more in brand posts and in the brand community. The number of fans only has a significant positive effect on liking and sharing behavior, but no effect on the number of comments. These findings in part are in line with previous results (Sabate et al., 2014). By applying a social interaction framework (Schultz, 2016), different social interaction strategies are identified, especially for the apparel retail brands. The three interaction measures are significantly highly correlated and may influence one another. Additional research should

also address this interrelationship, in particular the temporal diffusion and order effects of these interrelationships. Furthermore, the data is split according to the apparel or food retail industry, so as to account for potential differences. Table 5 reports the standardized estimation results for both industries.2 The industry-specific analysis shows several differences across post characteristics. Whereas differences between significant and non-significant results can be attributed to structural differences, three coefficients are inconsistent. Competitions have negative (positive) effects on comments and shares in apparel (food) retailing. An inspection of the corresponding posts reveals that apparel brands utilize more posts on contest coverage, which receives fewer brand post interactions. Furthermore, the holiday category limits or enhances liking behavior in apparel or food retailing respectively. The 17 and 19 posts in apparel and food retailing refer respectively to similar topics: Mother’s Day, Easter, and weekends. Food retail brands received more interactions for these topics than apparel retail brands. 5. Discussion and implications 5.1. Discussion Analyzing the characteristics of brand posts increases our understanding of consumer interactions and attitudes. Accord2 The three vividness characteristics for apparel retail posts have high variance inflation factors (picture 16.65, event 9.48, video 9.84), due to the small reference group. An alternative calculation, with picture as a reference category, produces results similar to those presented in the main text, which has the no vividness reference category.

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

ingly, the present study yields important insights on consumer behavior in social media. Other studies indicate positive effects of low (photo) and high (video) vividness on the number of likes (de Vries et al., 2012; Sabate et al., 2014; Trefzger et al., 2016), and the number of comments appear positively affected by low vividness (photo) posts (Sabate et al., 2014). The present study only finds positive effects of picture and video on likes and shares, as well as negative effects on comments. The use of pictures is prevalent in retail postings, so consumers may expect some minimal amount of post vividness. In turn, brands may need to adjust their social media strategy, monitoring future developments in the media that are available for use in brand posts. Media content and consumer expectations seem to evolve continuously, so that brands must adjust their creative strategies to meet these expectations, and address their target audiences in a new, creative way. Brands should also carefully evaluate whether more expensive videos really add value, relative to comparable quality pictures. In contrast, events are rarely used. Events may only address some part of the target social audience and as such, may not be deemed relevant for a geographically dispersed audience. One potential avenue for brands is to enable digital events to be similarly spatially independent. Interactive characteristics drive likes, comments, and shares, a finding in line with Luarn et al. (2015). Implementing these elements in brand posts supports a high-interaction social media strategy (Schultz, 2016). These elements also contribute to consumer engagement and enable brands to interact with target audiences. Hashtags, calls to action, contests, questions, and votes can activate target groups. An interesting challenge arises for retailers though, if their products evoke actions from different groups, which might be perceived differently by other audiences. In such cases, the brand must take particular care in managing its single brand page. Such interactions also enable brands to monitor whether internal and external perceptions of the brand align with one another. Consumer interactions vary across content types. In comparison to product category, the majority of content types lead to fewer brand interactions. An early study only identified categories as relevant (Pletikosa Cvijikj and Michahelles, 2011). This study further adds to the stream on brand post interactions, enhancing our understanding of this finding. Because some content categories, such as charity, competition, holiday, and human resources appeal only to part of the target audience (in comparison to product posts), consumers engage less with such brand posts. For example, posts related to competition negatively affect interaction measures, but posts about the contest itself receive many likes, comments, and shares. Consumers seem to react differently to contests than they do to additional posts related to the contests. Thus, an initial contest post may lead to short-term consumer activation, and additional coverage might draw significantly less interest. Brand managers should consider these effects when they conduct contests through their brand pages on social networking sites. Furthermore, product promotions (e.g., discounts, sales) by apparel retailers negatively affect liking behavior, but positively enhance commenting and sharing for food retail brands. Product promotions, which predominantly convey the transactional character, leave consumers less inclined to interact – except by directly conveying such offers to friends. As such, brands may actively select various content categories so as to influence target group behavior. Whereas consumers engage with product posts directly, brands may extend their reach to less active and potential new user, by selecting content type which predominantly activates sharing behavior. In terms of the length of time a post is positioned at the top of the page, brands must plan their posting schedule strategically.

31

Brand posts should appear at the top of the page for a reasonable amount of time, in order to receive the necessary exposure, but they also need to be replaced regularly to keep the page fresh. Following negative consumer interactions (e.g., Champoux et al., 2012; Munnukka and Järvi, 2014; Pfeffer et al., 2014), new brand posts may be an alternative to deleting content, which generally upsets consumers. The newer messages can attract attention and divert the focus from negative interactions. However, this strategic behavior might be suitable only in specific social networking sites; brand managers need to attend to other social media that may host a relevant discussion. The study indicates no difference between posts published on weekdays versus weekends. This result is in line with de Vries et al. (2012) and Sabate et al. (2014), but contrasts with Pletikosa Cvijikj and Michahelles (2013). However, posts are skewed toward weekday posts. Finally, post length has significant negative effects on interaction measures in apparel retailing. Continued research should analyze why consumers use social media and how much attention they devote to their various preferences and purposes. Brand posts should focus on a single topic and provide measures of activation, so as to facilitate consumer engagement and interactions on the brand page. That is, brand posts should be designed according to the purpose of the post and the brand’s overall social media strategy. Acknowledging the danger of engagement (Kaptein et al., 2016), brands should consider whether engagement claims additional resources otherwise used for business relevant tasks. 5.2. Theoretical implications Based on survey approaches, previous research identifies predominantly intentional and brand-related outcomes of consumer social engagement (e.g., Beukeboom et al., 2015; Gummerus et al., 2012; Zheng et al., 2015). Furthermore, actual social participation values (e.g., Kumar et al., 2016; Rishika et al., 2013) induce consumer spending, cross-buying, and in turn profitability. Also, consumer engagement and brand interactions may not only be positive, as engagement may also claim too much non-salesrelated attention (Kaptein et al., 2016). In addition, relating results of the present study to previous literature on brand post interactions (see Table 1), the findings indicate differences across the three interaction measures. In consequence, aggregating social interactions may not reflect such differences across different consumer engagement levels (Schultz, 2016; Tsai and Men, 2013). Given that liking, commenting, and sharing behavior differ from one another, the theory needs to account for these differences in analyzing social media behavior. In this way, the present study supports the notion of a multidimensional, as well as individual and situation-specific consumer engagement construct (Brodie et al., 2011). The results for brand post interactions as behavioral outputs of consumer engagement (Van Doorn et al., 2010; Kumar et al., 2010) further underline the complexity of consumer engagement and call for corresponding measurement that reflects its complexity. Research also needs to relate these various interaction measures, indicating different levels of consumer engagement to the various consumer engagement values (Kumar et al., 2010). In what ways do likes, comments, and shares refer to different lifetime, referral, influencer, and knowledge values? Existing studies provide examples of content categories (e.g., Pöyry et al., 2013; Rishika et al., 2013), but rarely address brand post content. One previous study utilizes an externally provided categorization that includes interactive elements, and the categorization is in part interdependent (Pletikosa Cvijikj and Michahelles, 2011). Naturally, brand posts may include more than one topic and thus pursue multiple objectives; however, this study finds that brands utilize social posts in such a way that their

32

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

followers are able to quickly capture the primary content message. Retail brands predominantly use product posts that relate to specific product-related elements, such as product introduction and presentation. These posts present the baseline type for the retail function. Correspondingly, as other content addresses only segments of the target audience, these messages receive fewer interactions from the engaged audience. For brands, utilizing such dedicated content seems relevant for addressing different motives (Jahn and Kunz, 2012; Pöyry et al., 2013). As such, brands may reach out to different segments of the target audience and enable multiple thematic engagement – balancing such activity, so that the focus is on the primary target group defined within the social media strategy. In this way, no one single content, but rather the interrelated focus may drive consumer engagement (Fang et al., 2016). Research is thus challenged to reflect content beyond entertainment, information, remuneration, and social categories (de Vries et al., 2012; Pletikosa Cvijikj and Michahelles, 2013; Luarn et al., 2015) towards capturing all aspects of a content topic. Topic dimensions and tonality, as well as additional approaches from linguistics, may contribute to achieving this objective. Above and beyond the individual post topic, research should consider the total number of topics and how these topics are composed within the broader context of a brand’s social media strategy. In this way, the interplay of brand messages drives consumer engagement, so that the theory needs to address this interrelatedness in future studies. The present study further reveals the complexity of social media. Beyond various user characteristics (see e.g., de Vries and Carlson, 2014; Jahn and Kunz, 2012), environmental variables such as market and target group characteristics, affect consumer engagement behavior with brand messages. As such, not only differences across consumer activity levels in liking, commenting, and sharing are relevant, but market characteristics as well. The study identifies such differences between the apparel and food retail industries. Social strategies consequently need to account for market segmentation and target group operationalization, which correspondingly affects related research. The degree of involvement in an industry and its products highlights the situation and user-specific dimensions of consumer engagement. Although the industry differences in this study are not particularly pronounced (see Table 5), they further support this element of consumer engagement. These findings not only question single interaction measures, but also draw attention to the concept of consumer engagement and thus the relevance of consumer experience management and the customer journey (Lemon and Verhoef, 2016). Considering consumer engagement beyond purchasing behavior, further research needs to address the relevance of interaction measures, also considering effects of consumer cocreation and thus its social influence on one’s own social group on (and beyond) the social networking site. 5.3. Managerial implications As mentioned, likes, comments and shares refer to different consumer activities and can be addressed via various elements of brand posts. Consequently, treating these interactions homogenously, such as in measures like post interaction rate, does not take these behavioral differences into consideration. Consumer activities reflect behavior from connecting to a brand page and consuming its brand messages, to contributing actively on the brand page and posts, as well as on the social networking site (Sashi, 2012; Tsai and Men, 2013). Even though Kristofferson et al. (2014) caution that liking may not lead to consequent behavior, research suggests that social media positively affect, for example, commitment and loyalty (e.g., Wirtz et al., 2013; Zheng et al., 2015), as well as purchase intention (Beukeboom et al., 2015; Hutter et al., 2013;

Ng, 2013) and sales value (Kumar et al., 2016, 2017; Pöyry et al., 2013). Brands have to consistently align social media objectives with different consumer engagement behaviors. For example, sharing brand posts with friends probably has a stronger effect than receiving updates on one’s own news feed about comments by a friend. An important associated challenge is determining how much money should be directed at primary engaged and interactive community members who contribute to the community, as at secondary members who provide value according to other objectives (Kaptein et al., 2016; Pöyry et al., 2013). Whereas a monotonic strategy of brand postings does not seem to create an engaging brand page, brands may utilize different post characteristics (post vividness, interactivity, content, and publication timing) to achieve the necessary balance and mix – thereby addressing different individual and market needs as well. In order to limit the effect of informational (e.g., Pöyry et al., 2013) and attentional (Kaptein et al., 2016) overload, brands can actively use publication timing and select appropriate content that only addresses some of the target audience. In contrast to gaining broad exposure, brands have to provide the corresponding timeframe and address contributing activities, such as comments and shares, in order to potentially reach ‘fans of fans’ as well. A challenge for brands is that social networking sites may adjust their algorithmic distribution of content – thus affecting the content that is transferred to feeds of friends. For example, Facebook has implemented changes to filter out unpaid promotional content transferred from brands, posted and posting as status updates (Kumar et al., 2016). Brands thus need to pay attention to such changes. In turn, engaging consumers in more active contribution behavior may shift the relevance of different measures and the corresponding activity level. As social media also provide comparative value (Schultz, 2016), brands can utilize and manage the available data to address corresponding social media decisions. Such observable data as used in this study enable brands to apply corresponding methods and to make informed decisions. Brands may thus not only rely on historic data but can use data from comparable brands in an industry and pursue the intended social media interaction strategy.

6. Limitations and future research This study used data from two retail industries which use a social networking site to engage and interact with their consumers. The demographic details of the fans of the brand pages in this study were not available. Despite some evidence of demographic similarities between the whole and the purely social target audience (Pletikosa Cvijikj and Michahelles, 2011), the findings are relevant only for retailers whose targets are congruent in terms of both demographics and behaviors. Further research needs to validate whether that are groups congruent in demographics behave similarly, as this study cannot confirm whether the social media population included in these data actually represent the overall target group. Similarly, no data were available pertaining to the relationships between brand page fans. For example, do fans and the friends of these fans behave similarly? The outlined questions relate to the self-selection bias inherent in social networking sites. Although the results may be limited to the target social group, managers may be especially interested in the potential of their social fan base. Likes, comments, and shares are important measures of brand post popularity and engagement, but an open question remains regarding how these different levels of consumer activity might result in relevant outcomes for business, such as greater consumer loyalty and more revenue. Whereas Kumar et al. (2016) aggregate all interaction measures into one (brand post receptivity) and find receptivity to significantly explain consumer spending,

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

cross-buying, and profitability, the present study indicates that some differences exists across liking, commenting, and sharing behavior. Following a customer engagement value framework (Kumar et al., 2010), further research might consider different value dimensions to provide insights into their contributions to the overall value of social networking sites, thus yielding further insights into the question of whether social media provide sufficient monetary value (e.g., Dhar and Chang, 2009; Onishi and Manchanda, 2012; Yu et al., 2013). For example, how are the three interaction metrics similar or different in terms of their connections to similar business measures or altogether distinct ones? Additionally, is engagement and interaction behavior always positive or does it sometimes consume other resources (Kaptein et al., 2016)? Future research can extend a single social network approach by including the interdependence of multiple social channels in the analysis. Social media managers have to decide which themes to address via which social media channel. Is spreading similar content via various channels a viable option? What type of content may drive what kind of consumer reactions via which social channels? Similarly, the data set in this study does not include additional activities, such as other offline advertisements or promotions. As Kumar et al. (2016) indicate, there are synergistic effects between brand messages and television advertising, as well as e-mail marketing. Kumar et al. (2017) find time-varying effects between social media and television advertising, product sampling, and product sampling. Accordingly, further research could usefully address the reciprocal interaction between offline and online activities and thus provide a more complete picture of drivers of brand interactions and their offline effects. This study measured the user interactions in retrospect; an interesting extension might be to investigate their temporal dispersion. This study also only briefly discussed the interaction between the dependent measures, so that further studies could analyze the effects of and relationships between likes, comments, and shares. Similarly, this study focused on single brand posts. However, brands should vary between different kinds of content to reduce any potential monotonic feel of their brand page. Further research might address the post portfolio by a given brand. What order, number, timing, or repeat patterns contribute towards an engaging social media strategy that also adds to the bottom line? Similar to interrelated characteristics affecting repurchase intention and perceived value (Fang et al., 2016), the mix of addressed values and motives create an engaging brand community. Finally, social networks generally function as bidirectional communication channels, so that further research might also investigate user posts. Beyond insightful survey data (Pöyry et al., 2013; Yang and Lin, 2014), it is worth considering what additional knowledge research and practice can gain from mixed method approaches to the social media contribution from users (e.g., Jahn and Kunz, 2012). As such, to what extant do engaged users create value in business processes? Acknowledgements An earlier version of the study was presented at the 18th International Conference on Electronic Commerce in Suwon, Republic of Korea, August 17-19, 2016. The author acknowledges the helpful discussions and comments from the research session and thanks the anonymous reviewers for their valuable comments and suggestions. Thanks too go to Brian Bloch for his linguistic editing. References Ansari, A., Koenigsberg, O., Stahl, F., 2011. Modeling multiple relationships in social networks. J. Mark. Res. 48, 713–728.

33

Beukeboom, C.J., Kerkhof, P., de Vries, M., 2015. Does a virtual like cause actual liking? how following a Brand’s Facebook updates enhances brand evaluations and purchase intention. J. Interact. Marketing 32, 26–36. Brodie, R.J., Hollebeek, L.D., Juric´, B., Ilic´, A., 2011. Customer engagement: conceptual domain, fundamental propositions, and implications for research. J. Serv. Res. 14, 252–271. Brodie, R.J., Ilic, A., Juric, B., Hollebeek, L., 2013. Consumer engagement in a virtual brand community: an exploratory analysis. J. Bus. Res. 66, 105–114. Champoux, V., Durgee, J., McGlynn, L., 2012. Corporate Facebook pages: when ‘‘fans” attack. J. Bus. Strategy 33, 22–30. de Vries, N.J., Carlson, J., 2014. Examining the drivers and brand performance implications of customer engagement with brands in the social media environment. J. Brand Manage. 21, 495–515. de Vries, L., Gensler, S., Leeflang, P.S.H., 2012. Popularity of brand posts on brand fan pages: an investigation of the effects of social media marketing. J. Interact. Marketing 26, 83–91. Dhar, V., Chang, E.A., 2009. Does chatter matter? The impact of user-generated content on music sales. J. Interact. Marketing 23, 300–307. Facebook, 2017. Facebook Q4 2016 Results. Retrieved April 30, 2016 http://investor. fb.com/results.cfm. Fang, J., George, B., Shao, Y., Wen, C., 2016. Affective and cognitive factors influencing repeat buying in e-commerce. Electron. Commer. Res. Appl. 19, 44–55. Goh, K.-Y., Heng, C.-S., Lin, Z., 2013. Social media brand community and consumer behavior: quantifying the relative impact of user- and marketer-generated content. Inf. Syst. Res. 24, 88–107. Golder, S.A., Wilkinson, D.M., Huberman, B.A., 2007. Rhythms of Social Interaction: Messaging Within a Massive Online Network. In: Steinfield, C., Pentland, B.T., Ackerman, M., Contractor, N. (Eds.), Communities and Technologies 2007: Proceedings of the Third Communities and Technologies Conference, Michigan State University 2007. Springer, London, London, pp. 41–66. Gomez-Arias, J.T., Genin, L., 2009. Beyond monetization: creating value through online social networks. Int. J. Electron. Bus. Manage. 7, 79–85. Gummerus, J., Liljander, V., Weman, E., Pihlström, M., 2012. Customer engagement in a Facebook brand community. Manage. Res. Rev. 35, 857–877. Hennig-Thurau, T., Malthouse, E.C., Friege, C., Gensler, S., Lobschat, L., Rangaswamy, A., Skiera, B., 2010. The impact of new media on customer relationships. J. Serv. Res. 13, 311–330. Hudson, S., Huang, L., Roth, M.S., Madden, T.J., 2016. The influence of social media interactions on consumer–brand relationships: a three-country study of brand perceptions and marketing behaviors. Int. J. Res. Mark. 33, 27–41. Hutter, K., Hautz, J., Dennhardt, S., Füller, J., 2013. The impact of user interactions in social media on brand awareness and purchase intention: the case of MINI on Facebook. J. Product Brand Manage. 22, 342–351. Jahn, B., Kunz, W., 2012. How to transform consumers into fans of your brand. J. Serv. Manage. 23, 344–361. Kabadayi, S., Price, K., 2014. Consumer – brand engagement on Facebook: liking and commenting behaviors. J. Res. Interact. Marketing 8, 203–223. Kaptein, M., Parvinen, P., Pöyry, E., 2016. The danger of engagement: behavioral observations of online community activity and service spending in the online gaming context. Int. J. Electron. Commerce 20, 50–75. Kristofferson, K., White, K., Peloza, J., 2014. The nature of slacktivism: how the social observability of an initial act of token support affects subsequent prosocial action. J. Consum. Res. 40, 1149–1166. Kumar, V., Aksoy, L., Donkers, B., Venkatesan, R., Wiesel, T., Tillmanns, S., 2010. Undervalued or overvalued customers: capturing total customer engagement value. J. Serv. Res. 13, 297–310. Kumar, A., Bezawada, R., Rishika, R., Janakiraman, R., Kannan, P.K., 2016. From social to sale: the effects of firm-generated content in social media on customer behavior. J. Marketing 80, 7–25. Kumar, V., Choi, J.B., Greene, M., 2017. Synergistic effects of social media and traditional marketing on brand sales: capturing the time-varying effects. J. Acad. Mark. Sci. 45, 268–288. Lemon, K.N., Verhoef, P.C., 2016. Understanding customer experience throughout the customer journey. J. Marketing 80, 69–96. 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. J. Advertising 31, 53–64. Luarn, P., Lin, Y.-F., Chiu, Y.-P., 2015. Influence of Facebook brand-page posts on online engagement. Online Inf. Rev. 39, 505–519. Munnukka, J., Järvi, P., 2014. Perceived risks and risk management of social media in an organizational context. Electron. Markets 24, 219–229. Ng, C.S.-P., 2013. Intention to purchase on social commerce websites across cultures: a cross-regional study. Inf. Manage. 50, 609–620. Onishi, H., Manchanda, P., 2012. Marketing activity, blogging and sales. Int. J. Res. Mark. 29 (3), 221–234. Park, H., Cho, H., 2012. Social network online communities: information sources for apparel shopping. J. Consum. Marketing 29, 400–411. Pfeffer, J., Zorbach, T., Carley, K.M., 2014. Understanding online firestorms: negative word-of-mouth dynamics in social media networks. J. Marketing Commun. 20, 117–128. Pletikosa Cvijikj, I., Michahelles, F., 2011. A Case Study of the Effects of Moderator Posts within a Facebook Brand Page. In: Datta, A., Shulman, S., Zheng, B., Lin, S.D., Sun, A., Lim, E.-P. (Eds.), Social Informatics: Third International Conference, SocInfo 2011, Singapore, October 6–8, 2011. Proceedings. Springer, Berlin Heidelberg, Berlin, Heidelberg, pp. 161–170.

34

C.D. Schultz / Electronic Commerce Research and Applications 26 (2017) 23–34

Pletikosa Cvijikj, I., Michahelles, F., 2013. Online engagement factors on Facebook brand pages. Soc. Network Anal. Mining 3, 843–861. Pöyry, E., Parvinen, P., Malmivaara, T., 2013. Can we get from liking to buying? Behavioral differences in hedonic and utilitarian Facebook usage. Electron. Commer. Res. Appl. 12, 224–235. Rishika, R., Kumar, A., Janakiraman, R., Bezawada, R., 2013. The effect of customers’ social media participation on customer visit frequency and profitability: an empirical investigation. Inf. Syst. Res. 24, 108–127. Sabate, F., Berbegal-Mirabent, J., Cañabate, A., Lebherz, P.R., 2014. Factors influencing popularity of branded content in Facebook fan pages. Eur. Manage. J. 32, 1001–1011. Sashi, C.M., 2012. Customer engagement, buyer-seller relationships, and social media. Manag. Decis. 50, 253–272. Schultz, C.D., 2016. Insights from consumer interactions on a social networking site: findings from six apparel retail brands. Electron. Markets 26, 203–217. TradeDimensions, 2015. Die Top 30 des Lebensmittelhandels 2014 nach GesamtUmsätzen. Frankfurt am Main. Trefzger, T., Baccarella, C., Voigt, K.-I., 2016. Antecedents of Brand Post Popularity in Facebook: The Influence of Images, Videos, and Text, 15th International Marketing Trends Conference, Venice, Italy.

Trusov, M., Bodapati, A.V., Bucklin, R.E., 2010. Determining influential users in internet social networks. J. Mark. Res. 47, 643–658. Trusov, M., Bucklin, R.E., Pauwels, K., 2009. Effects of word-of-mouth versus traditional marketing: findings from an internet social networking site. J. Marketing 73, 90–102. Tsai, W.-H.S., Men, L.R., 2013. Motivations and antecedents of consumer engagement with brand pages on social networking sites. J. Interact. Adv. 13, 76–87. Van Doorn, J., Lemon, K.N., Mittal, V., Nass, S., Pick, D., Pirner, P., Verhoef, P.C., 2010. Customer engagement behavior: theoretical foundations and research directions. J. Serv. Res. 13, 253–266. Wirtz, J., den Ambtman, A., Bloemer, J., Horváth, C., Ramaseshan, B., van de Klundert, J., Canli, Z.G., Kandampully, J., 2013. Managing brands and customer engagement in online brand communities. J. Serv. Manage. 24, 223–244. Yang, H.-L., Lin, C.-L., 2014. Why do people stick to Facebook web site? A value theory-based view. Inf. Technol. People 27, 21–37. Yu, Y., Duan, W., Cao, Q., 2013. The impact of social and conventional media on firm equity value: a sentiment analysis approach. Decis. Support Syst. 55, 919–926. Zheng, X., Cheung, C.M.K., Lee, M.K.O., Liang, L., 2015. Building brand loyalty through user engagement in online brand communities in social networking sites. Inf. Technol. People 28, 90–106.