A Picture Is Worth a Thousand Words: Segmenting

INTMAR-00171; No. of pages: 17; 4C:

Available online at www.sciencedirect.com

ScienceDirect Journal of Interactive Marketing 32 (2015) 53 – 69 www.elsevier.com/locate/intmar

A Picture Is Worth a Thousand Words: Segmenting Consumers by Facebook Profile Images Iris Vilnai-Yavetz ⁎ & Sigal Tifferet Department of Business Administration, School of Economics and Business Administration, Ruppin Academic Center, Emek Hefer, 4025000, Israel

Abstract Conventional segmentation efforts usually focus on verbal or behavioral data while ignoring visual cues, which play a significant role in impression management. Drawing on theoretical work regarding motivations for impression management (need to belong and need for selfpromotion), we propose that Facebook users differ from each other in the composition of visual elements they portray in their Facebook profile photos (PPs), and thus can be segmented based on this composition. In this exploratory study we present a methodological proof of concept for the visual segmentation of Facebook users. Using a randomly selected international sample of 500 Facebook accounts, we analyze data implicit in PPs and identify visual cues relevant to virtual impression management. Using these cues we segment users into types, and relate the types to demographics, Facebook usage, and brand engagement as reflected in the Facebook profile. At the theoretical level, the findings suggest that the current accepted motivations for Facebook impression management (need to belong and need for self-promotion) should be expanded to include a third motivation, need for self-expression. At the practical level, the findings demonstrate the utility of visual segmentation, which can later be implemented using computerized systems. © 2015 Direct Marketing Educational Foundation, Inc., dba Marketing EDGE. All rights reserved. Keywords: Facebook; Impression management; Photo analysis; Cluster analysis; Segmentation; Social presence

Introduction Facebook is one of the two most popular Web sites in the world, second only to Google (Alexa Internet, Inc. 2013), with a reported 1.2 billion active users monthly and 725 million daily (Facebook 2013). Facebook is also the number one mobile app used in the United States, after passing the popular Google Maps in October 2012 (comScore 2013). As a result, Facebook is rapidly emerging as a platform for advertising and marketing: indeed, the Facebook business model is based on advertising income (Mourdoukoutas 2013). But while advertising indiscriminately to 725 million people is bound to produce a certain number of matches (i.e., the right ad being seen by the right person at the right time), it is not highly efficient. Marketers therefore have a strong interest in finding ⁎ Corresponding author. E-mail addresses: [email protected] (I. Vilnai-Yavetz), [email protected] (S. Tifferet).

ways to segment Facebook users so as to target ads for goods and services more effectively. One of the key reasons people use Facebook is for purposes of impression management (Nadkarni and Hofmann 2012). Facebook users often employ verbal and nonverbal presentation of preferred brands for this purpose (Chen, Fay, and Wang 2011; Hollenbeck and Kaikati 2012; Labrecque, Markos, and Milne 2011; Smith, Fischer, and Yongjian 2012), offering marketers a convenient way to match users with products and services they are likely to find appealing. Another rich source of information is the personal data that Facebook users post in their profiles, including both demographic data and information on the person's interests and activities. But Facebook profiles also contain implicit data, such as photographs, which are mostly overlooked. Facebook users upload 350 million new images every day (Henschen 2013), providing data that marketing researchers can use to understand impression management — and to design ways of targeting online ads more effectively.

http://dx.doi.org/10.1016/j.intmar.2015.05.002 1094-9968/© 2015 Direct Marketing Educational Foundation, Inc., dba Marketing EDGE. All rights reserved.

54

I. Vilnai-Yavetz, S. Tifferet / Journal of Interactive Marketing 32 (2015) 53–69

In this study we have three aims. First, we suggest a method for segmenting social networking site (SNS) users based on visual cues instead of self-reports, and we illustrate the method on a random international sample of Facebook users. Second, we identify relevant visual cues present in virtual impression management. Third, we use these cues in order to identify different user types, and to relate them to theoretical motivations for using SNSs. We do this by analyzing visual data depicted in Facebook profile pictures (PPs) and identifying the impression management tactics they reflect. We first isolate segments of Facebook users based on the way they use PPs to create first impressions. We then examine whether the identified segments differ by their Facebook usage patterns, by their demographics, and by their brand engagement and preferred product categories as reflected in their Facebook pages. Theoretical Background Photos as Online Impression Management Tools The brain processes visual data 60,000 times faster than it does text (Parkinson 2012). It is unsurprising, then, that advances in computing power and broadband capacity have been accompanied by a tremendous increase in the visual content uploaded to social networks (Lee 2014). Today, two-thirds of the content on social media consists of images (Citrix report 2014). Nearly half of all Internet users have reposted a photo or video they have found online, and more than half have posted a photo or video that they have personally created (Duggan 2013). Tweets with images receive 150% more retweets than those with text only (Cooper 2013). Indeed, surveys suggest that including images is the most effective means of optimizing social media content (i.e., increasing the number of shares, “likes”, and followers that result; Aragon 2014). We suggest that profile pictures serve two interrelated purposes in SNS. First, according to social presence theory (Short, Williams, and Christie 1976), communication media differ in the level of social presence they enable, where social presence is defined as the degree to which the interaction partners have a sense of personal human contact. Social presence is highest in face-to-face communication, and lowest in wholly text-based communications, including the forms of electronic communication normally employed by firms engaging in e-commerce (Gefen and Straub 2004). One way to increase social presence in electronic communications is to add images such as photographs. Xu (2014), in a study of trust and credibility in the context of online consumer reviews, found that participants showed greater affective trust in reviewers who included a profile photo, and under some conditions photos were associated with greater perceived credibility of online reviews. Xu (2014) explained these findings on the grounds that feelings of uncertainty in interpersonal relationships make people uneasy (Berger and Calabrese 1975), and that profile photos reduce the feeling of unpleasantness which that uncertainty generates. Accordingly, we suggest that Facebook users similarly seek to increase their social presence and thereby reduce the uncertainty inherent in online interpersonal relationships by representing themselves via a PP.

The second purpose served by PPs in social media, which follows directly from the first, is impression management. In impression management, people seek to construct an image of themselves based on their ideas about how others will interpret that image (Leary and Kowalski 1990). First impressions are formed quickly, in a matter of moments, and as a result must be based on limited information, such as external appearance (Burgoon, Guerrero, and Floyd 2010). The centrality of first impressions in evaluations leads people to employ various impression management tactics, including verbal and nonverbal cues (Moore, Hickson, and Stacks 2010). Verbal tactics include, for instance, information filtering or “pleasing the audience” (Baumeister 1989); nonverbal tactics include facial expressions, body language, and physical appearance (Fletcher 1989). On the Internet, both social presence theory and impression management underlie the use of images and other visual cues, such as symbols, by individuals or entities that want to convey a particular message. For example, Vilnai-Yavetz and Tifferet (2013) found that top-ranked universities include more images in their Web sites than lower-rated academic institutions; and Vilnai-Yavetz and Tifferet (2009) found that prospective students relate the presence of images on a university site to better service. Koernig (2003) suggested that online images can be analyzed systematically for the same communication tactics that have been demonstrated in printed ads (Berry and Clark 1986). For instance, the qualities associated with an object can be physically represented using cues such as colors, logos, or symbols, while pictures, videos, etc. can encourage visitors to the site to visualize the object. Facebook Images as Impression Management Tools Facebook users can express themselves through explicit declarations regarding their interests or favorite books, films, or music (Pempek, Yermolayeva, and Calvert 2009). Yet viewers of Facebook profiles rely less on these explicit statements, and more on implicit cues such as those found in posted images (Zhao, Grasmuck, and Martin 2008). For example, when participants in one study evaluated the personality of Facebook users whom they did not already know, they based their impressions primarily on the users' PPs (Ivcevic and Ambady 2012). The appeal of these images can even raise the response rate to friendship requests (Tifferet, Gaziel, and Baram 2012; Wang et al. 2010). Lately, in addition to the PP, Facebook has enabled users to add a cover photo as part of its new Timeline format (Smith 2012). In the user's profile, the cover photo is a large, banner-style image that dominates the page, while the PP is smaller and less prominent. However, the PP is still the image that appears in Facebook friend requests and news feeds, and it therefore continues to be the basis of others' first impressions. Thus we focus in the current analysis on the PP. With Xu's (2014) findings in mind, we suggest that social network users use PPs to increase their social presence and reduce uncertainty in interaction partners by presenting a particular image of themselves, an image chosen to convey a particular impression about the user's identity. Specifically, users select profile photos whose visual elements reflect (for


example) a given emotional expression, dress preference, or preferred outdoor environment with the expectation that viewers of the picture will feel they “know” the individual better because of these elements. The use of these impression management techniques thus reduces uncertainty in the viewer (Berger and Calabrese 1975) and increases the sense of psychological social presence (Gefen and Straub 2004) more than would the mere presence of a photo (compare, for instance, the information conveyed by passport photos). Previous studies suggest two motives for Facebook use: self-promotion and belonging (Nadkarni and Hofmann 2012; Underwood, Kerlin, and Farrington-Flint 2011). We propose that these two motives also direct the specific composition of visual elements that users choose for the purpose of non-verbal impression management in their Facebook profile pictures. We also posit that Facebook users differ from each other in their specific motivations for Facebook impression management and in their specific composition of visual elements, and further, that they can be segmented based on this composition. In the current study we analyze impression management in Facebook PPs based on a set of conceptual constructs representing different aspects of impression management. The four constructs chosen are expressions of emotion, status, activeness, and total look. Below we elaborate on each construct and our theoretical rationale behind this choice, based on the two motives for Facebook use suggested in previous studies — namely, self-promotion and belonging (Nadkarni and Hofmann 2012; Underwood, Kerlin, and Farrington-Flint 2011). Expressions of Emotion Emotional expressions are a time-honored means of impression management (Rafaeli and Sutton 1987). While people who are in the grip of an emotion may have difficulty hiding or masking their true feelings, individuals are also practiced at conveying particular emotional states whether these are accurate representations of their true feelings or not (e.g., when a salesperson smiles at a customer). Emotions can be expressed in three main ways: verbal (what one says), nonverbal (e.g., eye contact, smiling), and paraverbal (e.g., tone of voice, speaking loudly) (Oksenberg, Coleman, and Cannell 1986). Positive emotions are easily expressed through nonverbal means, and particularly by engaging the muscles of both the mouth and the eyes (Ekman, Davidson, and Friesen 1990). People may also tend to divert their eyes when experiencing negative emotions such as sadness, or fear (Adams and Kleck 2005). Feelings of belonging intrinsically elicit positive emotions (Baumeister and Leary 1995). Therefore, we suggest that expressions of positive emotions in Facebook photos, such as smiling and looking directly at the camera, reflect a need to belong. To put it differently, Facebook users who are driven at least partially by a need to belong are expected to choose PPs that show such positive emotional expressions. Status In today's society, impression management is often bound up with people's aspirations for a given social status —

55

something that, in turn, may paradoxically reflect a need for both belonging (i.e., membership in a social group) and self-promotion. A key way in which people signal both their social group (need to belong) and social status (need for self-promotion) is through their clothing and the objects (accessories) they associate themselves with (Johnson et al. 2008). With respect to objects, these may be status symbols such as cars (Belk 2004) or fashion accessories (Han, Nunes, and Drèze 2010). With respect to clothing, a formal style of dress is often associated with high status (Nellisen and Meijers 2011). For instance, in one study, restaurant clients in business suits were served earlier in comparison to clients in casual dress (Stead and Zinkhan 1986). In another, participants associated formal business attire with authority and competence (Cardon and Okoro 2009). Andersen (2008) suggests that individuals in more formal clothing typically receive higher levels of compliance compared to people who are informally dressed. The notion of status has to be seen in context. Hirschman (2003), for example, points to “rugged individualism” as a “core American cultural value” expressed through the symbolism of “men, dogs, guns, and cars.” Within that frame of reference, status may be conveyed not by formal and expensive clothing and accessories, but by the opposite (see the Method section below). Specifically, we argue that the need for self-promotion may be reflected in the degree to which profile photos portray their subjects in clothing or with objects that are associated with high status within their frame of reference. However, the need to belong may be reflected in the degree of fit between the type of dress portrayed and the normative dress in the wearer's social surrounding (Bellezza, Gino, and Keinan 2014). As the current study was not designed to analyze human networks, we consider the clothing and objects pictured in PPs mainly in their role as potential symbols of social status (need for self-promotion), but not as a symbols of belonging to a social group. Activeness Modern society glorifies activeness and adventurousness (Underwood, Kerlin, and Farrington-Flint 2011). These traits are most associated with pursuits categorized as extreme sports, such as snowboarding, windsurfing, or hang gliding. But activeness can be reflected by involvement in any type of outdoor activity. Here again we can benefit from Hirschman's (2003) notion of “rugged individualism,” where the world of “men, dogs, guns, and cars” finds expression in earthy outdoor activities such as hunting. We suggest that PPs which portray their subjects as active, outdoorsy, and adventurous reflect the need for self-promotion. Total Look The fourth and final construct examined here is what we call “total look,” by which we mean the total look of the photo, not the subject. That is, our concern is not the subject's overall appearance (hair, clothing, accessories, etc.), but rather the general design choices made by the user — for instance, the degree to which the image is artistically processed versus naturalistic, or compositional questions such as whether the subject is shown alone or with other people (the specific criteria examined will be detailed in the Method section). The total

56


look construct may inform questions such as whether the individual is aiming to present some ideal self (for instance, via a heavily edited photo) rather than his or her actual self. The issue of ideal versus actual selves on Facebook has been the subject of recent research (Back et al. 2010; Hollenbeck and Kaikati 2012). The total look of a profile image may also suggest whether the account holder is driven more by the need to belong (e.g., users whose PPs show them as part of a group) or by the need for self-promotion (e.g., users whose PPs show them doing something unique and nonconformist [Bellezza, Gino, and Keinan 2014].) Segmenting Users of Social Networking Sites Identifying different groups of consumers has long been recognized as an important marketing tool. Segmentation systems have been developed based on consumers' motivations and demographics (e.g., Farrag, El Sayed, and Belk 2010), preferences (e.g., Jiang and Balasubramanian 2014), attitudes and beliefs (e.g., Strizhakova, Coulter, and Price 2012), activities (e.g., Gilboa 2009), shopping behavior (e.g., Wu and Chou 2011), service usage behavior (e.g., Bose and Chen 2010), and lifestyles (e.g., Ye, Li, and Gu 2011). However, a review of the SNS literature in general, and the SNS impression management literature in particular, yields – to the best of our knowledge – only four SNS segmentation studies (Alarcon-del-Amo, Lorenzo-Romero, and Gomez-Borja 2011; Constantinidesa and Zinck Stagnoa 2011; DiMicco and Millen 2007; Underwood, Kerlin, and FarringtonFlint 2011). Moreover, all four of these studies have methodological weaknesses, the most important being the use of self-reports. For instance, Underwood, Kerlin, and Farrington-Flint (2011) employed cluster analysis (CA) to segment 113 Facebook users based on their self-reported scores on both self-promotion and communication-related behaviors. They identified three segments: high broadcasters (interested mainly in self-promotion), high communicators (interested mainly in relationship maintenance), and a high interaction segment (interested in intensive interaction). Alarcon-del-Amo, Lorenzo-Romero, and GomezBorja (2011) used an online survey to question SNS users about their SNS activities, experiences, and interaction patterns; they identified four segments: introverts, novel users, versatile users, and expert-communicators. Finally, Constantinidesa and Zinck Stagnoa (2011) identified market segments among future students in The Netherlands who reported their social media habits in a national survey. Unfortunately, while self-reports are a useful means of gathering data in the social sciences, the method has several pitfalls including the difficulty of remembering past behavior (Brewer 2000), social desirability bias (De Jong, Pieters, and Fox 2010), and the fact that some psychological behaviors may be unconscious (e.g., Dijksterhuis et al. 2005). In the last case, for example, a man might display a photo of himself with an expensive sports car in the background without any selfawareness of his concern with status. With the aim of collecting more objective data, DiMicco and Millen (2007) segmented Facebook users employed by IBM into three behavioral types based on data collected from their Facebook profiles: number of Facebook friends, interests, and

job information. However, their study was only an initial case study that employed preliminary exploration of only 68 Facebook profiles in a non-representative sample. In fact, many samples are culturally biased, especially in light of the fact that 80% of Facebook users are from outside the U.S. and Canada (Facebook 2013). Along with DiMicco and Millen (2007), other researchers have begun to highlight the ways in which activity on social networks may be superior to the demographic data and self-reported attitudes traditionally used to segment consumers. Hargittai (2007), while not offering a segmentation herself, suggested various usage parameters as worthy of future study. Bachrach et al. (2012) analyzed a large data set of 180,000 Facebook users for features such as the size or density of friendship networks, the number of images uploaded, or the number of group memberships. They did not segment the users for marketing purposes, but they showed that these parameters can be used successfully to predict certain personality traits. Hill, Provost, and Volinsky (2006) followed another direction. They investigated rates at which target customers adopted a new telecommunications service, comparing 21 customer segments identified by the telecommunications firm. The researchers found that, in general, network neighbors (i.e., people who had communicated directly over a social network with a current subscriber of the service) adopted the new service at rates considerably higher than non-network neighbor target customers. Hill, Provost, and Volinsky (2006) concluded that analyzing customers based only on traditional attributes may miss potential customers. These findings support our own argument that traditional segmentation methods are at best a crude instrument in identifying people likely to try a product or service, and help make the case for examining the value of user attributes that may be derived from visual impression management. We suggest that objective and quantitative yet innovative data are a necessary input for supporting high quality marketing decision-making in today's brave new world of social media. The current study aims to meet this objective by showing how Facebook PPs can be used to segment consumers based on their impression management tactics, thereby providing an innovative source of information compared to traditional socio-demographic and self-reported data. We consider this study an important exploratory initial stage. It presents a proof of concept that may serve as a conceptual and methodological basis for future “Big Data” analysis. Hence we start our study with the following three propositions. P1. Facebook users differ in the composition of visual elements (e.g., emotional expressions, style of dress, outdoor versus indoor environment) they use to present themselves in PPs. P2. It is possible to identify separate groups of Facebook users based on their composition of visual elements. P3. Human needs in general, and users' motives for Facebook impression management in particular, are the source of variance between users. Therefore, the visual element composition patterns shared by members in a group should relate to their key motive for Facebook impression management.


Method Scoring Tool and Data Collection In order to identify different segments of Facebook users, we built a scoring tool that included variables related to, or reflecting, the conceptual constructs described above. After constructing an initial version of the tool, we reviewed previous studies of Facebook photos in order to identify possible additional variables. McAndrew and Jeong (2012) assessed Facebook impression management variables such as whether a picture includes objects, making faces at the camera, and graphic editing of photos. Steele, Evans, and Green (2009) examined characteristics such as smiling, location, eye coverings (e.g., sunglasses), and number of people. Zhao, Grasmuck, and Martin (2008) discussed characteristics such as the presence versus absence of people and expressions of affection in the context of how users employ pictures in identity construction on Facebook. This process resulted in a 20-item checklist. Three independent coders then used the checklist to separately code Facebook profile images. After discussing discrepancies, the coders constructed a final checklist comprised of 17 items, of which 10 related directly to the

57

profile picture and 7 related to other information available from the profile (see Table 1). With the final checklist established, an experienced rater coded the profile pictures of 500 randomly selected Facebook users in the fall of 2012 as part of a wider study assessing Facebook images (as described in Tifferet and Vilnai-Yavetz 2014). Since the variables were based on objective measures (e.g., no smile, smile showing teeth, smile without teeth), there was no need for additional independent coders. However, to assure coding accuracy of a single coder with these data, we arranged for independent coders to analyze a subsample of the Facebook profiles, and then tested for inter-rater reliability. The inter-rater measures were satisfactory (see Table 1). This procedure is described in detail below. Research Sample We collected a sample of 550 randomly selected Facebook profiles using ImageCrashers software (http://www.imagecrashers. com/GRP.do). This software displays random profiles selected from Facebook using Facebook Graph API. Of the 550 profiles, we excluded 45 profiles that did not state the user's gender and another

Table 1 Study variables and ICC coefficients. No.

Name

Variable content

Accounts assessed

ICC(2,1)

95% CI

1 2 3

Subject ID Gender

All All All

– – 1.0

– – –

4

Relationship Status

All

.88

[.80, .93]

5 6 7 8

Year of birth Friends Likes Style

All All All with Timeline All

– 1.0 1.0 .68

– – – [.57, .77]

9

Image

With image

.91

[.87, .94]

10 11

Number of people Object

Human All

.94 .53

[.91, .96] [.39, .65]

12

Smile

Only one person

.71

[.55, .84]

13

Eye_Cover

Only one person

.80

[.71, .92]

14

Eye_Contact

Only one person

.83

[.72, .91]

15

Dress

Only one person

.81

[.68, .89]

16

Situation

Human

.76

[.67, .84]

17

Timeline

Serial number Facebook ID (URL) 0. Male 1. Female 0. Not committed (single, it's complicated, in an open relationship, widowed, separated, divorced) 1. Committed (in relationship, engaged, married) Year Number of friends Number of likes 0. No profile photo/blank 1. Original photo/image 2. Artistic design (black–white, drawing) 0. Non-human photo or image 1. Human profile photo Number of adults and kids 0. No Object 1. Object 0. No Smile 1. Smile without teeth 2. Smile with teeth 0. None or eyeglasses 1. Sunglasses 0. Not looking at camera 1. Looking at camera 0. Minimal (swimming suit, bare) 1. Sportswear (running shorts) 2. Weekday casual (T-shirt) 3. Smart casual or branded attire 4. Formal/semi-formal 0. Indoors 1. Outdoors 0. No 1. Yes

All

–

–

Note: CI: confidence interval; ICC: IntraClass Correlation coefficient.

58


five that represented a company and not a personal user. The study sample therefore contained 500 Facebook profiles, of which 198 (39.6%) belonged to females and 302 (60.4%) to males (see Appendix A for an example of a PP). To analyze the images, we first checked the full sample of 500 PPs for the style of image and the presence of people and objects, where objects were defined as any inanimate item (a vehicle, a musical instrument, etc.). We then checked all the images that included people (n = 420) for an outdoor setting and for number of people. Last, we checked those images showing just one person (child or adult; n = 321) for sunglasses, eye contact (i.e., whether or not the person was looking at the camera), smile intensity, and level of formality in dress. A flow chart of the analysis process is shown in Fig. 1. Inter-rater Reliability Three independent coders were sent links to the same set of 100 Facebook pages, sampled randomly from the original 500 Facebook links. Due to a gap of a few months between the full data coding and this inter-rater reliability test, eight links were unavailable. After receiving instructions for the coding, the three coders coded the first five profiles separately, and then discussed issues of interpretation with the other coders and with the researchers. Then they continued to code the rest of the sampled Facebook profiles independently. A two-way, random-effects, single measure, absolute IntraClass Correlation (ICC) coefficient (ICC(2,1) model) was used to assess inter-rater reliability. The aim of this test is specifically to see how accurate a single coder will be when coding the data (Landers 2011). Point estimates of the ICCs were interpreted based on Fleiss (1986) as follows: excellent (.75–1), modest (.4–.74), and poor (0–.39). In general, the coders showed excellent reliability for most variables (Table 1). For three variables – showing an object (ICC = .53), style of

Fig. 1. Sample structure.

the image (ICC = .68), and type of smile (ICC = .71) – the inter-rater reliability was modest.

Research Variables Impression Management Constructs The research variables were divided into four segmentation constructs representing different aspects of impression management in Facebook: total look, status, activeness, and expressions of emotion. Total look reflected the general design choices made by the account holder, and was defined by three items: the style of the image, the decision to feature people or not, and the number of people in the image. Style reflected the decision to present no image (0), an unprocessed image (i.e., a realistic photo = 1), or a processed image (i.e., one that was graphically altered = 2). Presence of people reflected the decision to include (1) or not include (0) people in the image. Number of people showed the number of adults and children appearing in the photo on a continuous scale of 0 and up. Status displays were tested by the presence of objects and by the formality of the individual's dress. Objects were recorded only if they were central in the photo (0 = no object, 1 = presence of an object). Most of the objects displayed in the PP were electronic gadgets (37%) or vehicles (24%), both of which can serve as status symbols (see, for instance, Belk 2004). In most cases the vehicles shown were luxury vehicles (e.g., sports cars or heavy motorcycles). However, the study sampled users from around the world, including many less-affluent countries. In these countries, the mere ownership of a vehicle or an electronic gadget may serve as a status symbol, regardless of the brand. Dress style or formality of dress was rated only in photos that featured as subject just one person (see research sample section). It was coded as 0 = minimal (swimming suit, bare torso), 1 = sportswear, 2 = weekday casual (jeans and T-shirt), 3 = smart casual or branded attire, or 4 = formal/semi-formal. We rated activeness using an outdoor setting as a proxy. We assessed the setting only for images that included at least one person and that showed the situational context — e.g., at home or in an office for indoor photos (coded 0), or engaging in sports or an identifiable outdoor activity for outdoor photos (coded 1). Photos that were discernibly taken either indoors or outdoors but with no situational context (i.e., photos where the background was unidentifiable) were not rated. We measured emotional expression only in photos with just one individual (see the Research Sample section) using the following variables: eye contact (0 = not looking at the camera, 1 = looking at the camera), smiling (0 = no smile, 1 = slight smile with no teeth showing, 2 = full smile with teeth showing), and sunglasses (i.e., hiding emotional expression: 0 = eyeglasses or no eye cover, 1 = sunglasses). Smiling serves a social display of emotion (Fridlund 1994; Kraut and Johnston 1979). Smiling with bared teeth has a unique significance in humans and other primates in that it communicates submissive, nonaggressive intentions (e.g., De Waal 1989; Sapolsky 2004) and functions to increase affiliation (Parr and Waller 2006).


Facebook Usage In addition to the image analysis data, for each account we recorded the following data representing Facebook usage of the account owner: Number of Facebook friends; number of “likes” (i.e., the total number of times the user “liked” a commercial firm or any other type of organization, political party, service, product, brand, or celebrity); adoption of the Timeline format; and disclosure of data in the public realm (for relationship status, number of friends, number of likes, and year of birth). Demographics From the basic information in the Facebook profile we recorded the user's reported gender and year of birth. We categorized the user's relationship status into not committed (0 = single, it's complicated, in an open relationship, widowed, separated, or divorced) and committed (1 = in a relationship, engaged, or married). Data Analysis Segmentation Rinne and Swinyard (1995) divided market segmentation methods into two main approaches. In the “a priori” or “management-driven” approach, segmenting variables are determined in advance (e.g., Gonzalez-Benito 2002). This method is more methodologically controlled and can utilize past knowledge, but may miss unfamiliar or creative aspects of the subject. In the “analysis-based” or “market-driven” approach, data are segmented during the data-analysis process, usually by cluster analysis (e.g., Strizhakova, Coulter, and Price 2012). This approach is not directed by preexisting assumptions and so is more open to potentially unexpected aspects of the subject. Given the innovative nature of Facebook, we employed the market-driven approach in order to “listen to the data” and avoid a priori assumptions about the behavior of Facebook users. We conducted a cluster analysis (CA) aimed at differentiating Facebook users based on how they present themselves in their PP. As described above, this is the first image of a Facebook account owner normally seen by other users, and it therefore forms the basis of others' first impressions. We used a two-step CA that combined Ward's hierarchical analysis method with a nonhierarchical k-means clustering procedure in order to optimize the cluster solutions (Clatworthy et al. 2005). A two-step CA was most suitable for this study since it enables the analysis of categorical variables (e.g., the three smile categories). The seven variables relevant for PPs showing only one person (style of photo, presence of an object, dress style, situational context, smile, eye contact, sunglasses) – operationally representing the four segmentation constructs – were analyzed together. The CA was based on the characteristics of the person depicted in the PP, presumably the user. For this reason, only those PPs showing just one person (n = 321, 64%) were analyzed at this stage. In order to avoid a priori assumptions, we used the CA option that sets the number of clusters based on the data analysis (instead of setting the number of clusters arbitrarily). For that purpose, hierarchical clustering was used to determine the optimal number of clusters and to choose the most

59

appropriate solution based on changes in the agglomeration schedule, dendrogram, and cluster membership. The optimal number of clusters was determined using the point at which a noticeable increase in the clustering coefficient (agglomeration) occurs (Grove, Fisk, and Dorsch 1998). Then, a k-means CA was performed to determine the significance of each of the seven variables to the formation of the clusters. Facebook Usage and Demographics by Segment Once we had segmented the account holders based on their PPs, we examined whether and how the segments differed in their demographics and Facebook usage using cross-tab and ANOVA analyses. The demographic variables were gender, age, and relationship status, and the usage variables were number of friends and likes, adoption of the Timeline format, and level of data disclosure. Brand Engagement by Segment We applied Hollenbeck and Kaikati's (2012) qualitative approach to identifying brand engagement in Facebook pages. For this analysis, we randomly sampled eight Facebook pages for each of the five segments and examined whether and how the segments differed in the brand names and product categories displayed in profile interests, likes, and uploaded photos. Results Segmentation Categories Of the full sample of 500 profile images, 321 featured a single person, presumably the account holder. We segmented these images based on their features (e.g., smile, eye contact, and dress style). The CA produced three clusters (agglomeration coefficient = 3.52), which we termed Aloof, Affectionate, and Go-getter. Table 2 presents the distribution of each variable for the three segments. As shown in the table and in Fig. 2, the three clusters are similar in size, with each comprising about a third of the segmented accounts. The clusters significantly differed in dress style, situational context, smiling, eye contact, and sunglasses, but not in the processing of the image (style of image) or the presence of objects. We termed the first segment Aloof. These users' PPs showed them wearing low-end dress (minimal or weekday casual) in an indoor setting (e.g., at home or in an office), and with very little emotional expression (no smile or eye contact). For the most part, these individuals seemed to aim for an image of themselves as not needing to appear friendly or likeable, or to impress others via their dress or surroundings. We designated the second segment Affectionate. These users differed from the first group in that their photos showed positive emotions, making them seem friendly and open to communication. All were smiling, either fully (53%) or partially (47%). None wore sunglasses, and almost all (94%) were looking at the camera, as if making eye contact with the viewer. The third segment was termed Go-getter. These users presented themselves as tough and high-status. All were photographed

60


Table 2 Clusters of Facebook users based on profile photo design choices a (n = 321). b Cluster name

Aloof

Affectionate

Go-getter

Size of cluster

31.9%

34.1%

34.1%

Total look Style Unprocessed image Processed image

82.8% (− 1.3) 17.2% (1.3)

88.7% (.4) 11.3% (− .4)

90.3% (.9) 9.7% (− .9)

N.S.

Status Presenting an object No Yes Formality of dress style Minimal Sportswear Weekday casual Smart casual Semi-formal or formal

68.3% (− 2.2) 36.2% (2.2)

79.0% (1.1) 21.0% (− 1.1)

79.0% (1.1) 21.0% (− 1.1)

13.8% (3.4) 0% (− 1.4) 62.1% (2.6) 19.0% (− 2.6) 5.2% (− 2.1)

1.6% (− 1.7) 3.2% (.7) 50.0% (.4) 33.9% (.4) 11.3% (− .4)

1.6% (− 1.7) 3.2% (.7) 32.3% (− 3.0) 41.9% (2.1) 21.0% (2.4)

Activeness Situational context Indoors Outdoors

100% (8.9) 0% (− 8.9)

58.1% (1.2) 41.9% (− 1.2)

0% (− 10.0) 100% (10.0)

93.1% (6.2) 6.9% (− 3.2) 0% (− 4.4)

0% (− 12.0) 53.2% (7.7) 46.8% (7.0)

90.3% (5.9) 1.6% (− 4.6) 8.1% (− 2.6)

39.7% (2.6) 60.3% (− 2.6)

6.5% (− 4.5) 93.5% (4.5)

35.5% (1.9) 64.5% (− 1.9)

91.4% (.1) 8.6% (− .1)

100% (3.0) 0% (− 3.0)

82.3% (− 3.1) ⁎ 17.7% (3.1)

Expression of emotion Smile No smile Smile without teeth Smile with teeth Eye contact No Yes Sunglasses (hiding eyes) No Yes

Chi2

N.S. 29.3 ⁎⁎⁎

121.6 ⁎⁎⁎

141.7 ⁎⁎⁎

20.3 ⁎⁎⁎ 12.2 ⁎⁎

⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001. a Figures in parentheses are adjusted standardized residuals (ASRs). An |ASR| larger than 1.96 indicates that the number of cases in that cell is significantly different from the expected if the null hypothesis were true (p b .05). b A sub-sample of photos with only one person in the profile photo.

outdoors, some engaging in outdoor activities such as hiking or diving, and others driving. Of the three segments, they had the highest rate of sunglasses use (18%). As with the Aloof segment, most users in the Go-getter group were shown not smiling (90%), and slightly more than a third (36%) did not look at the camera (i.e., no eye contact). On average, the users in this group tended to dress more formally than those in the other segments. After completing the CA and identifying the three segments (Aloof, Affectionate, and Go-getter), we were left with 179 PPs that were excluded from the CA since they showed either no people or more than one person. We divided these profiles into two additional segments based on the number of people shown (i.e., none or multiple). PPs showing two or more people (e.g., family, friends; 19% of the full sample) were categorized as a fourth segment, which we termed Sociable based on the main attribute available about these users — the importance they ascribed to other people. PPs showing no people (i.e., displaying objects or abstract graphics; 17%) were categorized as a fifth segment. We termed this segment Cryptic because they provided

no physical information about their appearance, expressions, or clothing (see Fig. 2). In fact, personal testimonies show that some SNS users choose to minimize information disclosure in order to “create mystery” (Labrecque, Markos, and Milne 2011). Facebook Usage Patterns and Demographic Characteristics by Profile Segment Of the full sample, 54% disclosed data regarding their relationship status in their public profile, but only 10% disclosed their year of birth. Disclosure of Facebook usage parameters was more prevalent; 80% disclosed the number of their Facebook friends (M = 565 friends), and 79% disclosed their likes (M = 185 likes). Almost all users (95%) had adopted the Timeline format. No significant differences were found between the five segments in Facebook usage patterns (number of friends and likes, adoption of the Timeline format, and the disclosure or non-disclosure of data for relationship status, friends, likes, and year of birth).


61

Fig. 2. Five segments of Facebook users based on profile photo design choices.

The results in Table 3 show significant differences between the five segments in gender and in reported relationship status. Regarding gender, males comprised 61% of the full sample, but formed a significantly higher proportion of two segments, Go-getter and Cryptic, at 86% and 73% respectively, and a significantly lower proportion of two others — Sociable and Affectionate, at 47% and 39%. With regard to relationship status, 42% of the full sample reported being in a committed relationship, compared with a high of 53% in the Sociable group and a low of 14% in the Aloof group. As less than 10% of the sample reported their year of birth, we were not able to characterize the segments by age. Brand Engagement by Segment Application of Hollenbeck and Kaikati's (2012) qualitative approach to identifying brand engagement in Facebook pages revealed several differences between the segments. As can be seen in Table 4, profiles in the Aloof segment tend to feature branded alcoholic beverages, sports teams and athletes, fast food, cars, movie stars known for action and adventure movies (e.g., Vin Diesel, Bruce Lee), characters from such movies (e.g., Rambo, Spiderman), electronic devices, and other brands or names associated with a “masculine” lifestyle. Brands shown in the Affectionate segment tend to not only cluster in areas associated

with shopping and retail experiences, such as beauty, fashion, and sweet foods and snacks, but also include exciting and high-status vehicle brands such as BMW, Ferrari, Jaguar, and Harley Davidson. Theoretical models (e.g., the FCB grid; Vaughn 1980) supported by empirical studies (e.g., Claeys, Swinnen, and Abeele 1995) suggest that brands in these product categories – fashion, snacks, luxury cars, etc. – are particularly associated with emotion, in that their purchase is driven by the urge to satisfy emotional needs such as ego gratification, social acceptance, and sensory stimulation. Put differently, these are “feel” products, whose consumption is based on affective information processing centered around possibilities for self-enhancement, subjective meanings, and emotional impressions (as opposed to “think” products – kitchen appliances, food staples like bread, detergents, etc. – whose consumption is based on cognitive processing centered around the product's functional performance). The Go-getter can be characterized mainly by branded sports outfits and shoes, fast food and snacks, soft drinks and beers, suggesting a lifestyle associated with young people who are well-adapted to today's fast and demanding culture. Given that Facebook is a digital platform, unsurprisingly, electronic apps and devices feature to some degree in all five segments. However, digital applications and electronic games are the focus of interest for brand engagement in the Cryptic segment. It appears that those users who choose not to show

Table 3 Demographic characteristics of types of Facebook users based on profile photo design choices a (n = 494). Demographics

Overall sample

Aloof

Affectionate

Go-getter

Sociable

Cryptic

Chi2

Gender Male Female

61.4% 38.6%

67.2% (1.0) 32.8% (− 1.0)

38.7% (− 4.0) 61.3% (4.0)

85.5% (4.3) 14.5% (− 4.3)

47.4% (− 3.3) 52.6% (3.3)

73.3% (2.4) ⁎⁎ 26.7% (− 2.4)

41.9 ⁎⁎⁎

Relationship status Not committed Committed

57.7% 42.3%

86.2% (3.4) 13.8% (− 3.4)

54.1% (− .5) 45.9% (.5)

58.6% (.1) 41.4% (− .1)

47.5% (− 2.0) 52.5% (2.0)

54.3% (− .4) 45.7% (.4)

12.6 ⁎

⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001. a Figures in parentheses are adjusted standardized residuals (ASRs). An ASR larger than 1.96 indicates that the number of cases in that cell is significantly different from the expected if the null hypothesis were true (p b .05).

62


Table 4 Brand engagement in Facebook profiles by PP segments. Gender Country Aloof M M

USA India

F F M

USA USA Brazil

M

USA

M F

Romania USA

Affectionate F USA M M M M F F

India India USA India Philippines USA

F

Mauritius

Go-getter M Nigeria M Philippines M USA M USA M India M Israel M Botswana M Rajasthan Cryptic M F M M M M

Sudan Philippines N.A. Iran United Arab Emirates Hong Kong

M F

USA Nigeria

Sociable M USA F USA M F M

Bangladesh USA Romania

F F F

USA Nigeria Malaysia

International brands Bud Light, Walmart, iPhone, Mustang, Nike, Texas Rangers, Dallas Cowboys FC Barcelona, CBF Brazil, YouTube, Android, Samsung, Hyundai, Teachers whisky, National Geographic, NASA, Rambo, Spiderman, Scott Kelly (astronaut), Jet Li, Will Smith, Jason Statham, Vin Diesel, Neymar, Tom & Jerry McDonald's, Nesquik, Pizza Hut, Jelly Belly, Tic Tac, Pokemon, Batman, Tom & Jerry Herbalife, Miami Heat, Boston Celtics, San Antonio Spurs Lacoste, Heineken, Johnnie Walker, Tsunami (energy drink), Askov (vodka), Selvagem Comary (wine), Candy Crush, Sao Paulo FC, Manchester City FC, Bayern Munchen FC, Chicago Bulls, Real Madrid, Will Smith, Captain America, Harry Potter, Neymar Visa, Venus Williams, The Hobbit, Star Trek, Denzel Washington, Will Smith, Harry Potter, Jules Verne, J.K. Rowling, National Geographic None Purina, Victoria's Secret, Colorado Rockies, Kansas City Chiefs, Candy Crush, Bruce Lee, Venus Williams, Muhammad Ali, Jackie Chan, Van Damme, Vin Diesel, Jason Statham, Paul Walker, Hugh Jackman, Rambo, Mickey Mouse, Donald Duck, Snoopy Walmart, Nintendo Wii, Bud Light, Dove, Colgate, Quaker Oatmeal Squares, YouTube, FritoLay, Disney, Candy Crush, Adam Sandler Levi's, KFC, Nokia, BBC News, Candy Crush, Lionel Messi Van Heusen, McDonald's, BMW, Ferrari, Jaguar, Reebok, Acer, Sony, Microsoft, Harry Potter, NASA iPad, Dropbox Disney, Harley Davidson, National Geographic, Neymar Nescafe, Dell, Converse, Unilever, Louis Vuitton, Burberry Activia, Disney, Unilever, Walmart, Amazon, Visa, Milky Way, Pillsbury, Acer, Lancôme, Tide, Duracell, Yoplait, Skippy, Lipton, Olay, Sisley, Ghirardelli (Lindt Chocolate), Smashbox Cosmetics, Green Island (Rum), Angry Birds Sprite, Sony, Smirnoff Copa Lagos Beach Soccer, Manchester City FC Cadbury, Warner Brothers, Converse, Nestle, Ford, Instagram, Yahoo!, NBA Nike, Diadora, Ford, Corona, Budweiser Kellogg's, Dunkin Donuts, JCPenney, Pay Pal, Yahoo, ESPN, Walmart, Candy Crush, SEGA, X-Men, Los Angeles Lakers None Waze (GPS), Tuborg, 7 Up, Carlsberg, Samsung, Schweppes, Candy Crush, Angelina Jolie Kaizer Chiefs FC, DBS (clothing), Will Smith, Jason Statham, Paul Walker, Oprah Winfrey iPhone, Microsoft, Intel, WhatsApp None Google Chrome, Gmail, Microsoft, Google, Mozilla Firefox, Hello Kitty Canon SuperMario, BBC News, National Geographic, Bruce Lee Lexus 7-Eleven, Disney, iPhone6, Samsung, Olympus, iPad, Lego, M&M's, 20th Century Fox, Google Chrome, National Geographic, Sony, Microsoft, Ferrari, BMW, CNN, Captain America, Spiderman, Superman, Batman, Power Rangers, The Hobbit, Harry Potter, Bruce Lee, Paul Walker Nintendo, Disney, Sony PlayStation, Amazon, SEGA, Pokemon, Digimon, Spiderman, Batman, Cartoon Network, Dragon Ball Z None Netflix, FOX Sports, PlayStation, Hunger Games, Kobe Bryant Adidas, Michigan U, Budweiser, National Geographic, Vogue, Cleveland Browns, Cincinnati Bengals, Cincinnati Reds, Venus (clothing), Shoeaholics, Snoopy Pizza Hut, PC Gamer, Mitsubishi, Sims, KFC, Angelina Jolie, Nicole Kidman Toys R Us, Honda Vans, M&M's, Quiksilver, Etnies, Ralph Lauren, National Geographic, Levi's, Adidas, DC (shoes), Coca Cola, Reebok, Salitos, The Simpsons None Shell, Sterling, GlaxoSmithKline Kit Kat, Nestle, Hershey's, Coca-Cola, Pepsi-Cola, Casio, Disney, iPhone, HushPuppies, Winnie the Pooh, Dora

themselves or other people in their photos do so not for a lack of digital literacy, though it is unclear whether their main motivation is a desire to remain private and unexposed to the public, or to communicate a sophisticated message. Finally,

brands featured in the Sociable segment relate mainly to products that fill social needs and impress others (fashion and branded clothing) and products associated with hanging out (snacks and fast food).


Discussion and Conclusion SNSs in general, and Facebook in particular, offer marketers access to millions of potential customers. But to make advertising on these platforms more effective, marketers need to find ways to target ads to the people who are most likely to be interested in a particular product or service. We suggest that the images users post on Facebook or other social media platforms for impression management purposes (Nadkarni and Hofmann 2012) can provide valuable data on their attributes, interests, and attitudes — data that can be used to segment account holders based on the marketer's particular needs. The first aim of the current study was to suggest a method for segmenting SNS users based on visual cues instead of self-reports, and to illustrate this method on a sample of Facebook users. We achieved this by developing a segmentation of Facebook users on the basis of the impression management decisions expressed in their posted PPs. First, we documented five types of Facebook users based on the way they convey a concern with status, activeness, and emotional expression in their Facebook PPs. The segmentation reflects psychological characteristics such as sociability, creativity, empathizing or systemizing tendencies, multidimensionality, emotional expressivity, status seeking, a need to impress, and attitudes toward technology. Given that these characteristics are implicitly embedded in the images posted by Facebook users, our segmentation avoids the pitfalls faced by previous segmentations of Facebook users, which were based on self-reports (e.g., Alarcon-del-Amo, Lorenzo-Romero, and Gomez-Borja 2011; Underwood, Kerlin, and Farrington-Flint 2011). Although there are differences between SNSs in the type of user-generated content they present and the ways they present it (Smith, Fischer, and Yongjian 2012), the study results and segmentation can be applied to any SNS that uses photos (e.g., LinkedIn, Myspace). Our second objective was to identify relevant visual cues present in virtual impression management. The segmentation of the PPs suggests that five different categories of Facebook users (Aloof, Affectionate, Go-getter, Sociable, and Cryptic) can be distinguished based on only a few variables: the number of people in the photo (none, one, or multiple), the setting (indoors or outdoors), emotional expressions (the presence or absence of a smile; whether the subject is looking at the camera), and the relative formality of the subject's dress. These findings support the significance of PPs in general, and the specific visual cues studied in particular, as reducing uncertainty (Berger and Calabrese 1975) and increasing social presence (Gefen and Straub 2004), and show that visual cues have the potential to create interesting and useful segmentations. The segmentation results support P1 and P2 regarding user variation in visual impression management in Facebook, and the potential for segmentation. The identification of the segments led to our third goal, namely identifying the factors that motivated each segment in its SNS conduct. Using self-reports, previous studies suggested two motives for Facebook use: the need to belong and the need for self-promotion (Nadkarni and Hofmann 2012; Underwood, Kerlin, and Farrington-Flint 2011). Based on their characteristics,

63

it appears that Affectionate and Sociable users are motivated mainly by a need for belonging, and Go-getter users by a need for self-promotion. The needs for belonging and self-promotion are very similar to two of the three higher-order needs in Maslow's (1954) hierarchy — Belonging and Esteem. (Maslow's lower-order needs relate to existence and are thus less expected to be communicated in impression management.) Based on the characteristics of our identified types, we suggest adding a third Facebook motive to parallel Maslow's third higher-order need (self-actualization): the need for self-expression. This is the desire to present yourself to the world as you are, ignoring social norms and prejudice. In the current study, this tendency was apparent in Aloof users, who made no effort to present an ideal image to the world, and indeed displayed images which might reveal unpopular characteristics. It is debatable whether self-expression should be considered a separate need (as suggested by Maslow's theory), or as encompassed within another need. On the one hand, the need for self-expression may have specific characteristics that distinguish it from other needs. Hemetsberger (2005) described the motivations of contributors to open-source projects on the Internet. He claimed that their actions are motivated not by a desire for self-promotion, but rather by a need for self-realization — the wish to become one's own self. On the other hand, self-expression can be seen as serving various broader needs (e.g., Kenrick et al. 2010). For instance, Zinkhan et al. (1999) suggest that Internet users who create a personal Web site are motivated by a number of factors, including the need for “personal portrayal” or the need “to communicate different dimensions of your personality.” Yet in their view, this need is one component of the need for affiliation, and does not stand alone. Presumably, to some degree or another, people express all of their needs through their Facebook activity: belonging, selfpromotion, and self-expression. However, our typology, like Maslow's theory, suggests that one dominant need usually emerges. In this, our findings confirm our third proposition — namely, that human needs in general, and users' motives for Facebook impression management in particular, can be linked to the use of visual elements in PPs. The findings also support the added value of nonverbal communication (Burgoon, Guerrero, and Floyd 2010) over verbal communication for understanding impression management. Practical Implications The results of the current study can be used to inform a number of business practices: customer relationship management, customer recommendation networks, and, of course, customer segmentation. The results of our qualitative analysis show that the segments differ in the type of brands, product categories, and cultural icons they engage with in their Facebook profiles. We found that the Aloof segment prefers “masculine” lifestyle brands, while the Affectionate segment prefers brands that seek to engage the emotions. The brands favored by the ambitious Go-getters reflect the fast and demanding rhythms of modern life. The Cryptics seem to live in a virtual-digital world,

64


keeping busy with digital applications and electronic games. The Sociables, in contrast, are stimulated by social interactions, and favor branded clothing and fast food fit for quick get-togethers. The match between Facebook segments and brands is not surprising. Previous research has shown that Facebook users use brands to express themselves (Hollenbeck and Kaikati 2012), and may even identify themselves with a specific brand (Labrecque, Markos, and Milne 2011). Segmenting users by their impression management patterns combined with their brand engagement preferences allows marketers to offer people in each segment deals and bargains designed specifically to attract those individuals. For instance, the Sociable segment can be offered deals on products or services that appeal to families or groups of friends, such as amusement parks, while the Go-getter segment can be targeted for sports accessories. Followers of specific brands in the Aloof segment can be targeted for premium benefits, while followers in the Affectionate segment can be targeted for consumer engagement programs and emotional customer experiences. The information gleaned from the suggested segmentation can be further combined with specific data on individual users, such as whether the user has “liked” or displayed a particular brand, for even more individually tailored marketing. Yet crucially, the segmentation gives marketers a powerful sense of users' preferences even in the absence of specific information directly provided by the user. Hollenbeck and Kaikati (2012) found that only about 25% of Facebook users display their self-identity through the symbolic act of “liking” a brand, while McGrath (2015) found that 40% of Facebook users are passive users who only browse the site without posting or liking any content. The Facebook profile photo, therefore, serves as a basic common denominator for all users, in contrast to other data which requires higher levels of involvement or activity. This point is important in light of the increasing development of automated techniques for targeted marketing via Big Data analyses. Facebook, for instance, enables the activation of third-party apps that are based on user data (Wang, Xu, and Grossklags 2011). Google analytics uses past online activity as a means of learning about the customer. Finally, Social Network Analysis (SNA) is a mathematical-statistical approach which analyzes the flow of information from one participant to the other, but cannot analyze the content of the information (van Aalst 2012). The present findings offer practical value even in the era of SNA and data mining. First, SNA and data and text mining for target marketing are still complicated techniques that are not often used in business practice (Bonchi et al. 2011). This may be especially true for small-sized businesses which make up approximately 50% of the U.S. Gross Domestic Product (Kobe 2012). A segmentation approach that focuses on profile images may provide a more accessible managerial tool for such firms. Second, in SNA, SNS users are characterized based on their relations with others (i.e., location and connections in the network), without taking their attributes (e.g., motives, emotions, preferences) into account (Marin and Wellman 2011). Sentiment analysis might add the required psychological insights to SNA, but this is text-based analysis with major limitations, such as

topic-specific interpretations (Thelwall and Buckley 2013). Moreover, text-based analysis in general is likely to become less valuable given the rapid rise in prominence of visual content in SNS (Lee 2014). Thus, marketers would do well to adopt tools and techniques which can use this visual information for marketing purposes. Eventually, of course, visual data may be incorporated into SNA as well (van Aalst 2012). Yet even then, while SNAs can serve as useful tools for segmentation, any additional tool that can increase their predictive power may translate into large cumulative earnings. Our suggested segmentation has further practical implications. An important part of any marketing communications strategy is the choice of the message strategy, in general, and the type of endorser, in particular. Huber et al. (2013) found that a “typical user” endorser can be an effective way to influence consumers' brand perceptions. They suggest that different groups, or segments, of people are differently affected by the same endorser, based on the level of fit between the endorser and the consumer's mental image of typical brand users. Our Facebook profile photo segmentation provides a tool for better identifying the stereotypic “typical user” for each of the user segments. A high fit between the average user profile of a segment and the endorser can improve the effectiveness of ads. A final consideration involves the Buyer Persona concept. The Buyer Persona is “a semi-fictional representation of the firm's ideal customer based on market research and real data about existing customers” (Kusinitz 2014); such data usually include gender, age, profession, education, financial situation, and buying motivations (Handy 2014). Unfortunately, many marketers define their customers' personas based on intuition and not facts, leading to errors in strategy and content creation (MacFarland 2014). The notion of the Buyer Persona suggests uses for our Facebook profile photo segmentation even beyond targeted marketing, in that it can help marketers characterize their ideal customers based on psychological, emotional, and lifestyle elements — thus creating “a picture of the persona.” MacFarland (2014) suggests that marketers should employ Big Data technologies to expand the information on which they base their customer personas. We believe that profile photos can form a crucial source of such information. The suggested extended theoretical framework of motivations for impression management through Facebook has important practical implications that we could not test in the current study. Specifically, we suggest that advertising appeals, sales tactics, and customer relationship management strategies can be adapted to the type of user according to the dominant need directing his or her Facebook activity: belonging, self-promotion, or self-expression. Here we suggest applying Bacile, Ye, and Swilley's (2014) approach to consumer-contributed marketing communications, an area that is attracting growing interest in the age of social media. For instance, users motivated mainly by the need for belonging are better targeted with emotional appeals (e.g., a promise for happiness if visiting a specific mall), users motivated chiefly by the need for self-expression could be encouraged to offer new ideas for product innovation, suggest improvements in service processes, or express their feelings toward a specific brand, while


users motivated by the need for self-promotion can be recruited to present new products in their Facebook accounts, turning them into public opinion leaders. Future research should try to validate these suggested links between Facebook impression management motivations and the attitudes and behaviors of users for the purpose of consumer-contributed marketing. Future research might also employ reciprocity and assortativity analyses to determine whether users in the same PP segment are more strongly connected to other users in the same segment in other ways — e.g., whether they tend to “like” and friend others in their segment more than people in other segments. Such knowledge would be valuable in designing consumer-contributed marketing strategies.

Special Considerations and Future Research The approach outlined in this paper raises a number of special considerations, some of which offer scope for future research. The first of these involves issues of privacy and ethics in both research and marketing (see for example Crawford 2014). Is it acceptable to collect data about an account holder's emotional expressions, dress, and other details available in photographs without obtaining explicit permission from the user? In this regard, it can be argued that images uploaded to social media are intentionally chosen and publicly presented as a means of impression management or personal branding (Labrecque, Markos, and Milne 2011). More important, Facebook requires that account holders' names and PPs be accessible to all other users, but since users differ in the level of personal privacy they need or prefer (Such, Garcia-Fornes, and Botti 2013), Facebook allows them to restrict access to other verbal and non-verbal information via the site's privacy settings (Zhao, Grasmuck, and Martin 2008). Thus, PPs and any other visual data not specifically restricted by the user can be equated to the visual information people convey during a visit to the mall, beach, or other public area — and, therefore, as equally legitimate sources of data for analysis, whether for research or marketing purposes (Hurworth and Sweeney 1995). The second consideration deals with the authenticity of Facebook information. Do the features of the PP truly represent the Facebook user? Do a user's brand “likes” represent his brand engagement? Past studies have shown high congruency between the psychological traits of individual Facebook users and their personalities as perceived by observers (Gosling, Gaddis, and Vazire 2007). Kosinski, Stillwell, and Graepel (2013) even succeeded in predicting the demographics of Facebook users through their “likes”. Moreover, some Facebook users consider authenticity to be a successful self-branding strategy, expected from oneself and others (Labrecque, Markos, and Milne 2011). Even so, we do not know if a given user really has a dog or likes hiking — all we can see are the data presented by the user as her public persona. On their own, brand “likes”, brand name mentions, brand photo uploads, brand links, etc., may be weak-signal manifestations of actual brand engagement. Yet the accumulated evidence about a Facebook user, composed of all

65

the above, may be a significant road map for marketers in their attempt to reach their target market. Future studies can further assess the authenticity of the suggested segments. Labrecque, Markos, and Milne (2011) presented Facebook profiles of twelve users to others and asked for their impressions and evaluations. They then showed these impressions to the profile owners and asked them to rate their accuracy. Using their methodology, one can visually segment Facebook PPs and then ask the PP owners to rate these evaluations. Experimentally, fictive Facebook profiles can be created based on the design characteristics of each of the five segments. These Facebook profiles can then be evaluated by marketers and customers, allowing for comparison between the impressions created by each profile and the current study findings about each segment's “persona”. A third concern relates to stability over time. Some Facebook users change their PPs based on the time of year (e.g., holidays) and current life events (e.g., birthdays, anniversaries). Personal branding decisions also seem to change over time (Labrecque, Markos, and Milne 2011). These changes can reflect either transient shifts or more stable maturation processes. Future research using a longitudinal approach can study the dynamics of PP choices and change (frequency of changes, design congruency over time, etc.). An analysis of the congruency between the PP and the cover photo of the user can further contribute to our understanding of this process. Beyond the ethics, validity and reliability of the proposed segmentation process, there remains a fourth question — can it accurately predict consumer behavior? Here we conducted some qualitative explorations showing that the segmentation does indeed coincide with certain consumer tendencies. This, however, was only a preliminary trial, and further quantitative tests should follow. Future research can use the visual segmentation applied in this paper to segment Facebook PPs of known users who agree to participate in a survey and report their life style, purchasing habits, media consumption, level of exposure to advertising, and other consumer behavior information. Such data can link the PP segmentation to crucial consumer behavior variables and increase predictability. A fifth issue concerns the practicalities of the work presented here. The current study was based on manual coding. Such coding is labor intensive and although relatively objective in nature (e.g., counting the number of people in a photo, or coding a smile with or without teeth), it might be subject to bias in the interpretation of some variables, such as the formality of someone's clothing. This is especially relevant in an international sample where images of people from different nations and cultures are rated. With the passage of time, software is increasingly becoming able to process and analyze aspects of images (e.g. ImageJ, http://rsbweb.nih.gov/ ij/), including faces (e.g., Luxand Face recognition software, http://www.luxand.com). Future research might seek to replicate the current findings using such software, examining its efficacy in segmenting social media users based on their photographs. Finally, the market segmentation approach we employ in this paper was developed as a tool for targeting groups of consumers when mass communication was the main customer relations

66


instrument available. Some might argue that this approach is outdated in the Big Data era, when marketers can directly target ads for specific merchandise to individuals based on their browsing history. As technological advances alter the nature of marketing channels and consumers become increasingly sophisticated in their shopping behavior, market segmentation schemes are growing ever more refined, to the point where each consumer may be considered a “segment of one” (Dibb 2001) or “market of one” (Gilmore and Pine 2000). While much work in this area remains to be done, we suggest that the move from mass production and standardization to mass customization and differentiation (Gilmore and Pine 2000) does not obviate the value of photo analysis. Rather, advanced technologies for analyzing images and identifying human features may allow photo analysis to be used either independently or in conjunction with other available data (e.g., a person's browsing history) to improve the identification of consumers as “segments of one”. The choices people make with regard to their Facebook PPs provide significant information about their creativity, resistance to change, sociability, and more — information which reflects the user's interests and personality. The current paper is an exploratory study that presents a proof of concept aimed at identifying conceptual segmentation constructs. Future research should aim to link the image-based segmentation paradigm suggested in this paper with the “segment-of-one” approach, especially as photo analysis software becomes more refined.

Appendix A. An Illustration of a Facebook Profile

Conclusion Facebook reports 725 million daily active users (Facebook 2013) and had a revenue of $2.02 billion for the third quarter of 2013 (Mourdoukoutas 2013). As such, it is a goldmine for online marketing. While traditional segmentation builds on explicit data such as demographics, along with self-declarations and posts, we attempted to segment Facebook users using the implicit cues revealed through their PP. We used these implicit data to segment users based on the way they convey status, activeness, and emotions in their PPs. The findings suggest a potential new way of segmenting social network users which may be useful for marketing managers who wish to use social networks as a marketing tool.

Acknowledgments We would like to thank Naama Bayev, Chen Salamania, Betty Greenberg, and Diane Polakow for their assistance in data collection, and Meira Ben-Gad for language editing. We are grateful for the helpful insights received from colleagues in academia and industry. We thank Zvika Jerbi, Gadi Hayat, Daphne Raban, Benny Bornfeld, Tali Efrat, Hila Barda, Amit Rechavi, Itay Tsamir, Sheizaf Rafaeli, and Yoel Asseraf as well as two anonymous reviewers for their constructive comments. We are also grateful for the financial support of the Ruppin Academic Center (Research grant 22023_2014-15).


References Adams Jr., Reginald B. and Robert E. Kleck (2005), “Effects of Direct and Averted Gaze on the Perception of Facially Communicated Emotion,” Emotion, 5, 1, 3–11, http://dx.doi.org/:10.1037/1528-3542.5.1.3. Alarcón-del-Amo, Maria-del-Carmen, Carlota Lorenzo-Romero, and Miguel-Ángel Gómez-Borja (2011), “Classifying and Profiling Social Networking Site Users: A Latent Segmentation Approach,” Cyberpsychology, Behavior and Social Networking, 14, 9, 547–53, http://dx.doi.org/:10.1089/ cyber.2010.0346. Alexa Internet, Inc. (2013), “Top Sites: The Top 500 Sites on the Web,” Retrieved December 1, 2013, from http://www.alexa.com/topsites. Andersen, Peter A. (2008), Nonverbal Communication: Forms and Functions. Long Grove, IL: Waveland. Aragon, Kathryn (2014), 3 Can't-miss Tips for Optimizing Social Media Content. The Daily Egg (Retrieved February 12, 2015, from http://blog.crazyegg.com/ 2014/04/29/optimizing-social-media-content/?utm_content=buffer91ea2&utm_ medium=social&utm_source=twitter.com&utm_campaign=buffer). Bachrach, Yoram, Michal Kosinski, Thore Graepel, Pushmeet Kohli, and David Stillwell (2012), “Personality and Patterns of Facebook Usage,” 24–32. (Retrieved from http://dl.acm.org/citation.cfm?id=2380722). Bacile, Todd J., Christine Ye, and Esther Swilley (2014), “From Firmcontrolled to Consumer-contributed: Consumer Co-production of Personal Media Marketing Communication,” Journal of Interactive Marketing, 28, 2, 117–33, http://dx.doi.org/:10.1016/j.intmar.2013.12.001. Back, Mitja D., Juliane M. Stopfer, Simine Vazire, Sam Gaddis, Stefan C. Schmukle, Boris Egloff, and Samuel D. Gosling (2010), “Facebook Profiles Reflect Actual Personality, Not Self-idealization,” Psychological Science, 21, 3, 372–3744, http://dx.doi.org/:10.1177/0956797609360756. Baumeister, Roy F. (1989), “Motives and Costs of Self-presentation in Organizations,” in Impression Management in the Organization, R.A. Giacalone, P. Rosenfeld, editors. Hillsdale, NJ: Lawrence Erlbaum Associates. 57–71. ——— and Mark R. Leary (1995), “The Need to Belong: Desire for Interpersonal Attachments as a Fundamental Human Motivation,” Psychological Bulletin, 117, 3, 497–529, http://dx.doi.org/:10.1037/0033-2909.117.3.497. Belk, Russell W. (2004), “Men and Their Machines,” Advances in Consumer Research, 31, 273–8. Bellezza, Silvia, Francesca Gino, and Anat Keinan (2014), “The Red Sneakers Effect: Inferring Status and Competence from Signals of Nonconformity,” Journal of Consumer Research, 41, 35–54, http://dx.doi.org/:10.1086/ 674870. Berger, Charles R. and Richard J. Calabrese (1975), “Some Explorations in Initial Interaction and Beyond: Toward a Developmental Theory of Interpersonal Communication,” Human Communication Research, 1, 2, 99–112, http://dx.doi.org/:10.1111/j.1468-2958.1975.tb00258.x. Berry, Leonard L. and Terry Clark (1986), “Four Ways to Make Services More Tangible,” Business, 36, 53–5. Bonchi, Francesco, Carlos Castillo, Aristides Gionis, and Alejandro Jaimes (2011), “Social Network Analysis and Mining for Business Applications,” ACM Transactions on Intelligent Systems and Technology, 2, 3, 22–37. Bose, Indranil and Xi Chen (2010), “Exploring Business Opportunities from Mobile Services Data of Customers: An Inter-cluster Analysis Approach,” Electronic Commerce Research and Applications, 9, 3, 197–208, http:// dx.doi.org/:10.1016/j.elerap.2009.07.006. Brewer, Devon D. (2000), “Forgetting in the Recall-based Elicitation of Personal and Social Networks,” Social Networks, 22, 1, 29–43, http:// dx.doi.org/:10.1016/S0378-8733(99)00017-9. Burgoon, Judee K., Laura K. Guerrero, and Kory Floyd (2010), Nonverbal Communication. Boston: Allyn & Bacon. Cardon, Peter W. and Ephraim A. Okoro (2009), “Professional Characteristics Communicated by Formal Versus Casual Workplace Attire,” Business Communication Quarterly, 72, 3, 355–60, http://dx.doi.org/:10.1177/ 1080569909340682. Chen, Yubo, Scott Fay, and Qi Wang (2011), “The Role of Marketing in Social Media: How Online Consumer Reviews Evolve,” Journal of Interactive Marketing, 25, 2, 85–94, http://dx.doi.org/:10.1016/j.intmar. 2011.01.003.

67

Citrix report (2014, January), “Mobile Usage Trends,” Retrieved February 12, 2015, from http://www.slideshare.net/citrix/citrix-mobileusage. Claeys, C., A. Swinnen, and P. Vanden Abeele (1995), “Consumer's Means– End Chains for “Think” and “Feel” Products,” International Journal of Research in Marketing, 12, 3, 193–208, http://dx.doi.org/:10.1016/01678116(95)00021-S. Clatworthy, Jane, Deanna Buick, Matthew Hankins, John Weinman, and Robert Horne (2005), “The Use and Reporting of CA in Health Psychology,” British Journal of Health Psychology, 10, 329–58, http://dx.doi.org/: 10.1348/135910705X25697. comScore (2013), “Facebook Vaults Ahead of Google Maps to Finish 2012 as #1 U.S. Mobile App,” Retrieved December 1, 2013, from http://www.comscore. com/Insights/Blog/Facebook_Vaults_Ahead_of_Google_Maps_to_Finish_ 2012_as_number_1_US_Mobile_App. Constantinidesa, Efthymios and Marc C. Zinck Stagnoa (2011), “Potential of the Social Media as Instruments of Higher Education Marketing: A Segmentation Study,” Journal of Marketing for Higher Education, 21, 1, 7–24, http://dx.doi.org/:10.1080/08841241.2011.573593. Cooper, Belle Beth (2013), How Twitter's Expanded Images Increase Clicks, Retweets, and Favorites. (Retrieved February 12, 2015, from https://blog. bufferapp.com/the-power-of-twitters-new-expanded-images-and-how-tomake-the-most-of-it). Crawford, Kate (2014). When Big Data Marketing Becomes Stalking: Can Data Brokers be Trusted to Regulate Themselves? Scientific American. Retrieved Jan 28, 2014, from http://www.scientificamerican.com/article/when-bigdata-marketing-becomes-stalking/ De Jong, Martijn G., Rik Pieters, and Jean-Paul Fox (2010), “Reducing Social Desirability Bias Through Item Randomized Response: An Application to Measure Underreported Desires,” Journal of Marketing Research, 47, 1, 14–27, http://dx.doi.org/:10.1509/jmkr.47.1.14. De Waal, Frans (1989), Peacemaking Among Primates. Cambridge, MA: Harvard University Press. Dibb, Sally (2001), “New Millennium, New Segments: Moving Towards the Segment of One?” Journal of Strategic Marketing, 9, 193–213, http:// dx.doi.org/:10.1080/713775742. Dijksterhuis, Ap, Pamela K. Smith, Rick B. van Baaren, and Daniel H.J. Wigboldus (2005), “The Unconscious Consumer: Effects of Environment on Consumer Behavior,” Journal of Consumer Psychology, 15, 3, 193–202, http://dx.doi.org/:10.1207/s15327663jcp1503_3. DiMicco, Joan Morris and David R. Millen (2007), “Identity Management: Multiple Presentations of Self in Facebook,” . Duggan, Maeve (2013), Photo and Video Sharing Grow Online. Pew Research Center . Retrieved February 12, 2015, from http://www.pewinternet.org/files/ old-media//Files/Reports/2013/PIP_Photos%20and%20videos%20online_ 102813.pdf. Ekman, Richard J., Richard J. Davidson, and Wallace V. Friesen (1990), “The Duchenne Smile: Emotional Expression and Brain Physiology: II,” Journal of Personality and Social Psychology, 58, 2, 342–53, http://dx.doi.org/: 10.1037/0022-3514.58.2.342. Facebook (2013), “Facebook Newsroom,” Retrieved December 1, 2013, from http://newsroom.fb.com/Key-Facts. Farrag, Dalia A., Ismail M. El Sayed, and Russell W. Belk (2010), “Mall Shopping Motives and Activities: A Multimethod Approach,” Journal of International Consumer Marketing, 22, 2, 95–115, http://dx.doi.org/:10.1080/ 08961530903476113. Fleiss, Joseph L. (1986), The Design and Analysis of Clinical Experiments. New York: John Wiley & Sons. Fletcher, Clive (1989), “Impression Management in the Selection Interview,” in Impression Management in the Organization, R.A. Giacalone, P. Rosenfeld, editors. Hillsdale, NJ: Lawrence Erlbaum Associates, 269–82. Fridlund, Alan J. (1994), Human Facial Expression: An Evolutionary View. San Diego, CA: Academic Press. Gefen, David and Detmar W. Straub (2004), “Consumer Trust in B2C eCommerce and the Importance of Social Presence: Experiments in e-Products and e-Services,” Omega, 32, 407–24. Gilboa, Shaked (2009), “A Segmentation Study of Israeli Mall Customers,” Journal of Retailing and Consumer Services, 16, 135–44, http://dx.doi.org/: 10.1016/j.jretconser.2008.11.001.

68


Gilmore, James H., Pine, B. Joseph, editors. Markets of One: Creating Customerunique Value Through Mass Customization, Boston: Harvard Business Review Press. Gonzalez-Benito, Oscar (2002), “Geodemographic and Socioeconomic Characterization of the Retail Attraction of Leading Hypermarket Chains in Spain,” International Review of Retail Distribution & Consumer Research, 12, 1, 81–103, http://dx.doi.org/:10.1080/09593960110103869. Gosling, Samuel D., Sam Gaddis, and Simine Vazire (2007), “Personality Impressions Based on Facebook Profiles,” . 26–8. Grove, STEPHEN J., RAYMOND P. Fisk, and MICHAEL J. Dorsch (1998), “Assessing the Theatrical Components of the Service Encounter: A Cluster Analysis Examination,” The Service Industries Journal, 18, 3, 116–34, http://dx.doi.org/:10.1080/02642069800000035. Han, Young Jee, Joseph C. Nunes, and Xavier Drèze (2010), “Signaling Status with Luxury Goods: The Role of Brand Prominence,” Journal of Marketing, 74, 4, 15–30, http://dx.doi.org/:10.1509/jmkg.74.4.15. Handy, Bethany (2014, May 4), The Importance of Creating Buyer Personas for Your Sales Team. Thinkhandy (Retrieved February 7, 2015, from http:// blog.thinkhandy.com/buyer-personas-for-your-sales-team). Hargittai, Eszter (2007), “Whose Space? Differences Among Users and Non-users of Social Network Sites,” Journal of Computer-Mediated Communication, 13, 1, 276–97, http://dx.doi.org/:10.1111/j.1083-6101.2007.00396.x. Hemetsberger, Andrea (2005), “Creative Cyborgs: How Consumers Use the Internet for Self-realization,” Advances in Consumer Research, 32, 1, 653–60. Henschen, Doug (2013). Facebook on Big Data Analytics: An Insider's View. Information week. Retrieved May 7, 2013, from http://www.informationweek. com/cloud-computing/platform/facebook-on-big-data-analytics-an-inside/ 240150902. Hill, Shawndra, Foster Provost, and Chris Volinsky (2006), “Networkbased Marketing: Identifying Likely Adopters via Consumer Networks,” Statistical Science, 21, 2, 256–76, http://dx.doi.org/:10.1214/ 088342306000000222. Hirschman, Elizabeth C. (2003), “Men, Dogs, Guns, and Cars: The Semiotics of Rugged Individualism,” Journal of Advertising, 32, 1, 9–22, http:// dx.doi.org/:10.1080/00913367.2003.10601001. Hollenbeck, Candice R. and Andrew M. Kaikati (2012), “Consumers' Use of Brands to Reflect Their Actual and Ideal Selves on Facebook,” International Journal of Research in Marketing, 29, 395–405, http:// dx.doi.org/:10.1016/j.ijresmar.2012.06.002. Huber, Frank, Frederik Meyer, Johannes Vogel, Andrea Weihrauch, and Julia Hamprecht (2013), “Endorser Age and Stereotypes: Consequences on Brand Age,” Journal of Business Research, 66, 2, 207–15, http://dx.doi.org/: 10.1016/j.jbusres.2012.07.014. Hurworth, Rosalind and Martin Sweeney (1995), “The Use of the Visual Image in a Variety of Australian Evaluations,” Evaluation Practice, 16, 2, 153–64, http://dx.doi.org/:10.1016/0886-1633(95)90024-1. Ivcevic, Zorana and Nalini Ambady (2012), “Personality Impressions from Identity Claims on Facebook,” Psychology of Popular Media Culture, 1, 1, 38–45, http://dx.doi.org/:10.1037/a0027329. Jiang, Pingjun and Siva K. Balasubramanian (2014), “An Empirical Comparison of Market Efficiency: Electronic Marketplaces vs. Traditional Retail Formats,” Electronic Commerce Research and Applications, 13, 2, 98–109, http://dx.doi.org/:10.1016/j.elerap.2013.11.003. Johnson, Kim K.P., Jeong-Ju Yoo, Minjeong Kim, and Sharron J. Lennon (2008), “Dress and Human Behavior: A Review and Critique,” Clothing & Textiles Research Journal, 26, 1, 3–22, http://dx.doi.org/:10.1177/ 0887302X07303626. Kenrick, Douglas T., Vladas Griskevicius, Steven L. Neuberg, and Mark Schaller (2010), “Renovating the Pyramid of Needs: Contemporary Extensions Built Upon Ancient Foundations,” Perspectives on Psychological Science, 5, 3, 292–314, http://dx.doi.org/:10.1177/ 1745691610369469. Kobe, Kathryn (2012), Small Business GDP: Update 2002–2010. (Retrieved from https://www.sba.gov/sites/default/files/rs390tot_0.pdf). Koernig, Stephen K. (2003), “E-scapes: The Electronic Physical Environment and Service Tangibility,” Psychology & Marketing, 20, 2, 151–67, http:// dx.doi.org/:10.1002/mar.10065.

Kosinski, Michal, David Stillwell, and Thore Graepel (2013), “Private Traits and Attributes are Predictable from Digital Records of Human Behavior,” Proceedings of the National Academy of Sciences, 110, 15, 5802–5, http:// dx.doi.org/:10.1073/pnas.1218772110. Kraut, Robert E. and Robert E. Johnston (1979), “Social and Emotional Messages of Smiling: An Ethological Approach,” Journal of Personality and Social Psychology, 37, 1539–53. Kusinitz, Sam (2014, March 8), The Definition of a Buyer Persona [in Under 100 Words]. HubSpot Blogs/Marketing (Retrieved February 7, 2015 from http://blog.hubspot.com/marketing/buyer-persona-definition-under-100-sr). Labrecque, L.I., Ereni Markos, and George R. Milne (2011), “Online Personal Branding: Processes, Challenges, and Implications,” Journal of Interactive Marketing, 25, 1, 37–50, http://dx.doi.org/:10.1016/j.intmar.2010.09.002. Landers, Richard N. (2011), Computing Intraclass Correlations (ICC) as Estimates of Interrater Reliability in SPSS. NeoAcademic (Retrieved February 18, 2015 from http://neoacademic.com/2011/11/16/computingintraclass-correlations-icc-as-estimates-of-interrater-reliability-in-spss/). Leary, Mark R. and Robin M. Kowalski (1990), “Impression Management: A Literature Review and Two-component Model,” Psychological Bulletin, 107, 1, 34–47, http://dx.doi.org/:10.1037/0033-2909.107.1.34. Lee, Kevan (2014, May 27), A Complete Guide to Visual Content: The Science, Tools, and Strategy of Creating Killer Images. (Retrieved February 7, 2015 from https://blog.bufferapp.com/a-complete-guide-to-creating-awesomevisual-content). MacFarland, Scott (2014, Aug 11), Big Data and Marketing Personas — A Perfect Match? Huff Post Business (Retrieved February 7, 2015 from http://www.huffingtonpost.com/scott-macfarland/big-data-and-marketingpe_b_5777660.html). Marin, Alexandra and Barry Wellman (2011), “Social Network Analysis: An Introduction,” in The SAGE Handbook of Social Network Analysis, J. Scott, P.J. Carrington, editors. London: Sage. 11–25 .(Ch. 2). Maslow, Abraham H. (1954), Motivation and Personality. New York: Harper & Row. McAndrew, Francis T. and Hye Sun Jeong (2012), “Who Does What on Facebook? Age, Sex, and Relationship Status as Predictors of Facebook Use,” Computers in Human Behavior, 28, 6, 2359–65, http://dx.doi.org/:10.1016/j.chb.2012.07.007. McGrath, Felim (2015, January 20), “4 in 10 Facebookers Now Browsing the Site Passively,” Retrieved from: http://www.globalwebindex.net/blog/4-in10-facebookers-now-browsing-the-site-passively. Moore, Nina-Jo, Mark Hickson III, and Don W. Stacks (2010), Nonverbal Communication: Studies and Application. 5th ed. New York: Oxford University Press. Mourdoukoutas, Panos (2013), “Facebook Is Getting its Business Model Right. Forbes: Markets,” Retrieved: Feb 4, 2014, from http://www.forbes.com/ sites/panosmourdoukoutas/2013/10/30/facebook-is-getting-its-businessmodel-right/. Nadkarni, Ashwini and Stefan G. Hofmann (2012), “Why Do People Use Facebook?” Personality and Individual Differences, 52, 243–9, http:// dx.doi.org/:10.1016/j.paid.2011.11.007. Nellisen, Rob M.A. and Marijn H.C. Meijers (2011), “Social Benefits of Luxury Brands as Costly Signals of Wealth and Status,” Evolution and Human Behavior, 32, 6, 331–49, http://dx.doi.org/:10.1016/j.evolhumbehav.2010.12.002. Oksenberg, Lois, Lerita Coleman, and Charles F. Cannell (1986), “Interviewers' Voices and Refusal Rates in Telephone Surveys,” Public Opinion Quarterly, 50, 97–111, http://dx.doi.org/:10.1086/268962. Parr, Lisa A. and Bridget M. Waller (2006), “Understanding Chimpanzee Facial Expression: Insights into the Evolution of Communication,” Social Cognitive and Affective Neuroscience, 1, 3, 221–8, http://dx.doi.org/:10.1093/scan/nsl031. Parkinson, Mike (2012), The Power of Visual Communication. Do-it-yourself Billion Dollar Graphics. PepperLip, Inc. (Retrieved February 12, 2015, from http://www.billiondollargraphics.com/infographics.html). Pempek, Tiffany A., Yevdokiya A. Yermolayeva, and Sandra L. Calvert (2009), “College Students' Social Networking Experiences on Facebook,” Journal of Applied Developmental Psychology, 30, 3, 227–38, http:// dx.doi.org/:10.1016/j.appdev.2008.12.010. Rafaeli, Anat and Robert I. Sutton (1987), “The Expression of Emotion as Part of the Work Role,” Academy of Management Review, 12, 1, 23–37, http:// dx.doi.org/:10.5465/AMR.1987.4306444.

I. Vilnai-Yavetz, S. Tifferet / Journal of Interactive Marketing 32 (2015) 53–69 Rinne, Heikki and William R. Swinyard (1995), “Segmenting the Discount Store Market: The Domination of the Difficult Discounter Core,” International Review of Retail, Distribution & Consumer Research in Marketing, 5, 2, 123–45, http://dx.doi.org/:10.1080/09593969500000010. Sapolsky, Robert M. (2004), “Social Status and Health in Humans and Other Animals,” Annual Review of Anthropology, 33, 393–418, http://dx.doi.org/: 10.1146/annurev.anthro.33.070203.144000. Short, John, Ederyn Williams, and Bruce Christie (1976), Social Psychology of Telecommunications. London: Wiley. Smith, Andrew N., Eileen Fischer, and Chen Yongjian (2012), “How Does Brand-related User-generated Content Differ Across YouTube, Facebook, and Twitter?” Journal of Interactive Marketing, 26, 2, 102–13, http:// dx.doi.org/:10.1016/j.intmar.2012.01.002. Smith, Catharine (2012), Facebook Timeline Becoming Mandatory for All Users. Huffington Post (Retrieved February 11, 2013, from http://www. huffingtonpost.com/2012/01/24/facebook-timeline_n_1228800.html). Stead, Bette Ann and George M. Zinkhan (1986), “Service Priority in Department Stores: The Effects of Customer Gender and Dress,” Sex Roles, 15, 11-12, 601–11, http://dx.doi.org/:10.1007/BF00288217. Steele, Fitzgerald Jr., David C. Evans, and Rachel K. Green (2009), “Is Your Profile Picture Worth 1000 Words?” .(Retrieved March, 28, 2014, from https://www.aaai.org/ocs/index.php/ICWSM/09/paper/viewPaper/216). Strizhakova, Yuliya, Robin A. Coulter, and Linda L. Price (2012), “The Young Adult Cohort in Emerging Markets: Assessing Their Glocal Cultural Identity in a Global Marketplace,” International Journal of Research in Marketing, 29, 43–54, http://dx.doi.org/:10.1016/j.ijresmar.2011.08.002. Such, Jose M., Ana Garcia-Fornes, and Vicent Botti (2013), “Automated Buyer Profiling Control Based on Human Privacy Attitudes,” Electronic Commerce Research and Applications, 12, 6, 386–96, http://dx.doi.org/: 10.1016/j.elerap.2013.05.001. Thelwall, Mike and Kevan Buckley (2013), “Topic-based Sentiment Analysis for the Social Web: The Role of Mood and Issue-related Words,” Journal of the American Society for Information Science and Technology, 64, 8, 1608–17, http://dx.doi.org/:10.1002/asi.22872. Tifferet, Sigal, Ofir Gaziel, and Yoav Baram (2012), “Guitar Increases Male Facebook Attractiveness: Preliminary Support for the Sexual Selection Theory of Music,” Letters on Evolutionary Behavioral Science, 3, 1, 4–6, http://dx.doi.org/:10.5178/lebs.2012.18. ——— and Iris Vilnai-Yavetz (2014), “Gender Differences in Facebook Selfpresentation: An International Randomized Study,” Computers in Human Behavior, 35, June, 388–99, http://dx.doi.org/:10.1016/j.chb.2014.03.016. Underwood, Jean D.M., Lianne Kerlin, and Lee Farrington-Flint (2011), “The Lies We Tell and What They Say About Us: Using Behavioral Characteristics to Explain Facebook Activity,” Computers in Human Behavior, 27, 1621–6, http://dx.doi.org/:10.1016/j.chb.2011.01.012. van Aalst, Jan (2012), “Commentary: Combining Content Analysis and Social Network Analysis,” Open and Interdisciplinary Journal of Technology, Culture and Education, 7, 2, 84–94. Vaughn, Richard (1980), “How Advertising Works: A Planning Model,” Journal of Advertising Research, 20, 5, 27–33. Vilnai-Yavetz, Iris and Sigal Tifferet (2009), “Images in Academic Web Pages as Marketing Tools: Meeting the Challenge of Service

69

Intangibility,” Journal of Relationship Marketing, 8, 148–64, http:// dx.doi.org/:10.1080/15332660902876893. ——— (2013), “Promoting Service Brands via the Internet,” The Service Industries Journal, 33, 15-16, 1544–63, http://dx.doi.org/:10.1080/ 02642069.2011.636423. Wang, Shaojung Sharon, Shin-I Moon, Kyounghee Hazel Kwon, Carolyn A. Evans, and Michael A. Stefanone (2010), “Face Off: Implications of Visual Cues on Initiating Friendship on Facebook,” Computers in Human Behavior, 26, 2, 226–34, http://dx.doi.org/:10.1016/j.chb.2009.10.001. Wang, Na, Heng Xu, and Jens Grossklags (2011), “Third-party Apps on Facebook: Privacy and the Illusion of Control,” Proceedings of the 5th ACM Symposium on Computer Human Interaction for Management of Information Technology, ACM CHIMIT11, December 4, 2011, Boston, MA, USA. 4. Wu, Roung-Shiun and Po-Hsuan Chou (2011), “Customer Segmentation of Multiple Category Data in E-commerce Using a Soft-clustering Approach,” Electronic Commerce Research and Applications, 10, 3, 331–41, http:// dx.doi.org/:10.1016/j.elerap.2010.11.002. Xu, Qian (2014), “Should I Trust Him? The Effects of Reviewer Profile Characteristics on eWOM Credibility,” Computers in Human Behavior, 33, 136–44, http://dx.doi.org/:10.1016/j.chb.2014.01.027. Ye, Qiang, Guoxin Li, and Bin Gu (2011), “A Cross-cultural Validation of the Web Usage-related Lifestyle Scale: An Empirical Investigation in China,” Electronic Commerce Research and Applications, 10, 3, 304–12, http:// dx.doi.org/:10.1016/j.elerap.2010.11.001. Zhao, Shanyang, Sherri Grasmuck, and Jason Martin (2008), “Identity Construction on Facebook: Digital Empowerment in Anchored Relationships,” Computers in Human Behavior, 24, 5, 1816–36, http://dx.doi.org/: 10.1016/j.chb.2008.02.012. Zinkhan, George M., Margy Conchar, Ajay Gupta, and Gary Geissler (1999), “Motivations Underlying the Creation of Personal Web Pages: An Exploratory Study,” Advances in Consumer Research, 26, 1, 69–74 (doi:acrwebsite.org/volumes/8227/volumes/v26/NA-26). Iris Vilnai-Yavetz received her PhD in Management from the Technion, Haifa, and her BA in Psychology and her MBA from the Hebrew University, Jerusalem. She is a senior lecturer of marketing at the Ruppin Academic Center, Israel. Before her graduate studies Iris held a senior managerial position in a marketing research and consulting firm. Her current research interests are in the area of environmental psychology, with implications for consumer behavior, organizational psychology, services, and marketing. She has published in outlets such as the Journal of Service Research, European Journal of Marketing, Service Industries Journal, Service Business, Environment & Behavior, Computers in Human Behavior, Electronic Commerce Research and Applications, and Organization Science.

Sigal Tifferet received her PhD in Psychology from the Hebrew University, Jerusalem. She is a senior lecturer at the Ruppin Academic Center. Sigal is interested in the implementation of evolutionary theory to the behavior of consumers and families.