1 Introduction

Since the emergence of relationship marketing, practitioners have been trying to create strong ties between brands and consumers. Indeed, consumers develop feelings towards brands that are similar to those they have towards humans (Fournier and Alvarez 2012). To bring their brand closer to consumers, practitioners develop their brand’s presence on social media sites such as Facebook or Twitter. Nowadays, consumers interact with brands on a wide range of online platforms (Hamilton et al. 2016). Online brand-consumer interactions have a positive impact on consumers attitudes (Hudson et al. 2015) and behaviors (Kumar et al. 2016). However, research generally considers this interaction as a whole, without being specific about the different kinds of interactions.

Conversation, a particular form of interaction, has emerged over the last few years as a trending topic in the marketing field (Berthelot-Guiet 2011; De Montety and Patrin-Leclère 2011). We define ‘brand conversation’ as a series of public online messages exchanged between several individuals, at least one of them being a brand representative (Andriuzzi 2017). These conversations can be initiated by brands, e.g., when brands post a message on their Facebook page, or initiated by consumers, e.g., when they complain about a brand’s products or services on Twitter.

Only a few authors are beginning to study these conversations. For example, within the recent ‘webcare’ research field, researchers have been discussing the circumstances in which brands should respond to consumers’ comments (e.g., Van Noort and Willemsen 2012), adopting a reactive point of view on interaction. Although the issue of effectively managing consumer complaints is crucial, it seems important to study brand conversations initiated by brands, as well as those initiated by consumers.

Another problem for brand managers is that only a minority of consumers react to brand messages on social media (Campbell et al. 2014). As social media budgets are dramatically increasing, it could be disappointing for marketers to think that they are addressing only a small part of consumers. However, existing research helps to put this situation into perspective by showing that the online audience is made of both posters who participate in conversations and lurkers who simply observe them (Schlosser 2005). It is therefore necessary to understand to what extent these conversations have an impact on both types of web users. In addition, by using new advertising tools offered by social media sites such as Facebook and Twitter, brands can push their posts to a wider audience, beyond their regular followers. Since these messages can be modified by consumers’ comments, brand-consumer interactions have a significant potential impact. With this in mind, this research addresses two main themes: the lack of knowledge about conversations initiated by brands, as well as the lack of knowledge about their effect on lurkers.

Using an experimental method, we test the impact of brand-consumer conversations on consumers who just look at these interactions on social media. As lurking consumers can have different experiences with tested brands, we study the potential moderating effect of brand attachment on the relationship between conversational strategies and consumers’ perceptions. Using Goffman’s face-work (1955) as a theoretical lens, we find that consumers with high attachment do not react to a brand showing appreciation to other consumers. Conversely, we find that consumers with low attachment feel like the brand is more human when the brand does show appreciation to others.

2 Conceptual Development

2.1 Face-Work Theory

Verbal and unscripted interactions between brands and consumers typically involve a human brand spokesperson, e.g., a community manager (Griffiths and Mclean 2015). Therefore, in order to study brand conversation, we mobilize a theoretical framework that usually applies to interactions between individuals: face-work. Sociology and linguistics-rooted, face-work theory was developed by Goffman (1955, 1967) before being extended by Brown and Levinson (1987) and others like Kerbrat-Orecchioni (2005). Face-work explains how individuals behave when they meet others by assuming that, during an interaction, each of the participants are committed to carrying out two simultaneous and continuous actions: maintaining their own face while ensuring other participants do not lose face (Goffman 1955). To achieve this dual mission, they use a number of strategies aimed at avoiding or minimizing face-threatening acts (FTAs; Goffman 1973), as well as producing face-flattering acts (FFAs; Kerbrat-Orecchioni 2007).

According to face-work theory, any interaction potentially threatens people’s desire for freedom and autonomy (Brown and Levinson 1987). For instance, in an online context, encouraging people to participate in the conversation by asking for their opinions or making suggestions, or even just simply addressing them, could be considered an FTA. To counteract the potential negative impact of social media posting, and thus improve the perception of Internet users, face-work theory suggests speakers could try to minimize their FTA, for example by being indirect (e.g., “one could do…” instead of “you should do”). To counteract FTAs’ negative effects, one could also enhance consumers’ face by producing FFAs. An FFA can consist of flattering people, paying respect to them or more generally by producing appreciative expressions (Brown and Levinson 1987; Kerbrat-Orecchioni 2007). In an online marketing context, brands could be considered as producing FFAs when showing consideration or appreciation to consumers, while responding to their questions or comments on social media.

2.2 Appreciative Expressions Impact on Brand Humanization

Even if a very scarce amount of marketing research taps into face-work as a theoretical framework, FFAs have already shown a positive effect on consumers within previous works. For example, when a firm rejects consumers’ ideas, consumers feel less face-threatened when the firm uses face enhancement at the same time (Fombelle et al. 2016). Such face enhancement could include recognizing consumers’ past-submitted ideas, including consumers within a community or apologizing. Furthermore, as FFAs are more efficient when produced publicly compared to privately (ibid.), we could suggest that face-work strategies have an impact on consumers when produced on public social media sites such as Facebook or Twitter.

When developing such face-work strategies, one could say that brands conform to human conversational norms as face-work is considered as an invariant of human communication (Brown and Levinson 1987). Indeed, brands tend to mimic human-to-human communications on social media, instead of just adopting an “advertising” posture. Brands and their agents rely on a wide range of rhetoric tactics in online posts such as using personal pronouns (see Packard et al. 2018), second-person pronouns when addressing consumers (see Cruz et al. 2017) or using a familiar language (see Gretry et al. 2017). In addition, as consumers can interpret social cues online, brands use nonverbal communications such as emoticons, a technique that can provide consumers with a feeling of warmth towards brands (Li et al. 2018). Therefore, one could think that mimicking human interactions norms, such as adopting face-work strategies, drives consumers to feel that brands have human characteristics.

The consumers’ propensity to attribute human characteristics to inanimate objects, i.e., anthropomorphism, is an important concept in marketing. Indeed, it is considered to be the pillar of the existence of brand-consumer relationships, a foundation for profitable consumer’s attitudes and behaviors (Aaker and Fournier 1995). Even if anthropomorphism is an individual psychological characteristic (Epley et al. 2007), brands’ strategies can have an impact on the way consumers attribute more or less human traits to them (MacInnis and Folkes 2017). The adoption of human language in interaction, a human characteristic if any, could lead to such outcomes. Thus, as the production of face-flattering acts during a conversation is supposed to generate a positive attitude towards the speaker, we make the following hypothesis:

  • H1: Brands are perceived as more human when they use appreciative expressions, compared to when they do not use them.

2.3 Moderating Role of Consumer’s Brand Attachment

Although face-work theory predicts a positive effect of appreciative expressions, research shows that practitioners should be careful not to go too far when interacting with consumers online. For instance, a brand that warmly thanks consumers for their positive contributions is not always well received: consumers may think that the brand is trying to manipulate them by using flattery (Wang and Chaudhry 2018). Such results contrast with face-work theory, where positive comments should rather lead to positive outcomes. Yet, the body of work from Brown and Levinson (1987) shows a potential explanation: when in a close relationship, people do not need sophisticated face-work strategies for their interactions to be successful. For example, an interaction between old friends would be more likely be made of “bald on record” speech acts rather that to contain sophisticated FTA mitigation or FFA production. In other words, the type of existing relationship between speakers moderates the relationship between face-work and its outcomes.

As we know that the relations between brands and consumers sometimes imitate person-to-person relationships (Fournier and Alvarez 2012), one could think that brand-consumer relationships could have an impact on the brand’s face-work perception by consumers. As brand attachment is a good way to evaluate the strengths of brand-consumer relationships (Whan Park et al. 2010), one could think that appreciation from the brand would be more or less well perceived depending on the degree of brand attachment. Indeed, as they are already close to the brand, attached consumers probably do not need the brand to “over-emphasize” the relationship.

Psychology research supports the existence of different perceptions depending on the context: people usually like better those who flatter them rather than those who flatter others (Vonk 2002). This occurs because, on the one hand, most people have positive self-esteem and therefore are likely to think their ingratiator is sincere. On the other hand, observers lack information on the target of ingratiation and therefore could question the ingratiated judgement. Moreover, ego considerations would lead people to think that they deserve to be appreciated more than others:

“Being motivated to assume, as most people are, that they are better than others, observers may be reluctant to uncritically accept lavish praise about another participant who just happened to be there at the same time” (Vonk 2002, p. 525).

For similar reasons, in a marketing context, consumers may prefer to be flattered rather than to see strangers being flattered. In addition, because of their emotional bond to the brand, attached consumers would negatively over-react to brand appreciation targeting others. Indeed, attached consumers would likely claim to be the subject of brand appreciation, as they would think that they deserve it more than unknown consumers would. Conversely, unattached consumer should not wait for specific treatment from the brand and therefore not react as negatively to brand flattery addressed to others. In contrary, as they do not have specific expectation from the brand, unattached consumers may be pleasantly surprised to see that the brand respects the rules of interpersonal communication by paying attention to others. Indeed, a surprise effect can play a positive role in brand evaluation (Schamari and Schaefers 2015). For all these reasons, we formulate a second hypothesis:

  • H2: When consumers are attached to the brand, the use of appreciative expressions addressed to others influences brand anthropomorphism less than when consumers are not attached to the brand.

3 Method

To test the causal relationship between brand interaction strategies and consumers’ attitudes, we ran an experimental design. We tested the impact of a face-work lever (i.e., appreciative expressions) on consumers’ attitude by measuring to what extent they see the brand as imbued with human characteristics (i.e., on perceived anthropomorphism) in the context of high or low brand attachment. Thus, we conducted a 2 (appreciation: yes/no) × 2 (brand attachment: high/low) between-subjects experiment.

3.1 Participants and Procedure

One hundred and eighty-eight participants (mean age = 36.60, SD = 11.77; 50% female) living in France were recruited and received monetary compensation. Participants belong to a consumer panel managed by an online research firm. They were screened by attesting they were holding a Facebook account.

Participants were randomly assigned to the different experimental cells. They received an email with a link to a self-assessed survey where they were asked questions about a car brand (n = 97) or about a coffee brand (n = 91). After these preliminary questions, we asked participants to imagine they were browsing their Facebook newsfeed and we exposed them to a screenshot of a fictitious brand-consumer conversation on Facebook, designed by us. Chosen brands are well known in France, thus the Facebook brand-consumer interaction stimuli could be seen as credible by participants. Finally, we asked them to answer questions about this conversation and again about the car or the coffee brand. Among other variables (see below), we chose brand personification as a control variable, as personification can have an impact on perceived anthropomorphism (Cohen 2014; MacInnis and Folkes 2017). Therefore, we designed four stimuli by combining two modalities of appreciation (yes/no) and of brand personification (yes/no).

3.2 Appreciation and Brand Attachment Manipulation

Appreciation was manipulated by including appreciative expressions to the brands’ posts and answers in the conversations (e.g., “What a nice question, thank you again!”; appreciation = yes) or not including them (appreciation = no). Second, we created ex-post groups of consumers based of their brand attachment (M = 4.56). We ended with a group of weakly attached consumers (n = 88, Attachment < 4.56) and a group of strongly attached consumers (n = 99, Attachment > 4.56).

3.3 Measures

As manipulation checks, participants answered four items using seven-point Likert scales indicating to which extent they found the brand used appreciative expressions (e.g., “[Brand] cast people in a good light”, “[Brand] make people look good”; α = .92; adapted from Kerssen-Griep et al. 2008). As for brand attachment, we used the measure by Lacoeuilhe (2000) as cited and adapted by Gouteron (2011). Participants answered five seven-point Likert scales (e.g., “I have a lot of affection for [Brand]”, “I feel connected to [Brand]”; α = .95). Finally, consumers indicated to which extent they anthropomorphized the brand by answering six seven-point semantic differential scales (e.g., “conscious/unconscious”, “artificial/natural”; α = .89; adapted from Hudson et al. 2015).

3.4 Control Variables

As face-work theory predicts that incentive expression (e.g., “Tell us what you think!”) can have an impact on consumer’s perception, we controlled incitation. Thus, we chose a non-incitation condition by only showing participants conversation without explicit incentive expressions as we thought it would be more neutral. To measure incitation level perception, we asked participants to answer three seven-point Likert scales, indicating to which extent they found the brand used incentive expressions (e.g., “[Brand] incites people to answer questions”, [Brand] pushes people to give their opinion”; α = .85). We also controlled the conversation topic by only showing conversations that were related to the brand’s core product (a new coffee format launched by the coffee brand; a new electric car launched by the car brand). To check their understanding of the conversations’ topics, we asked participants if they thought the conversation was about the brand’s products. Finally, we manipulated brand personification by exposing participants to conversations where the brand was posting from a generic branded account (i.e., identified by the brand’s name and logo) or from a dedicated employee account (“Stéphanie – [brand’s name] client service” plus the picture of a young woman). To check brand personification manipulation, we asked participants if they thought the brand was represented by its logo or by the picture of an employee. To control brand personification as well as the brand itself, we included them as covariates in data analysis (see below).

4 Results

4.1 Manipulation Check

A one-factor ANOVA revealed a significant effect of appreciation (yes/no) on perceived appreciation (F (1, 186) = 6.01, p < .05). Participants in the appreciation condition found that appreciation was higher (M = 5.53, SD = 0.94, n = 94) than participants in the no-appreciation condition (M = 5.19, SD = 0.98, n = 94). Moreover, we found a significant difference in brand attachment between our groups of high and low attached consumers (p < .001). Consumers in high attachment condition scored higher in attachment (M = 5.65) than consumers in low attachment (M = 3.35). Therefore, we considered our manipulations were successful. As for control variables, first, participants judged the brand’s incentive level low, as expected (M_perceived_incitation = 3.62). A one-factor ANOVA showed that differences of perceived incitation within cells were non-significant (F (3, 184) = 1.72, p = .165). Therefore, incitation was controlled at a weak level. Second, as for the conversation topic, all participants found that the conversation was about a product sold by the brand (n = 188). Finally, as for brand personification, all participants in the “logo” condition (n = 89) said they thought the brand was represented by its logo, where all participants in the “employee” condition (n = 99) said they thought the brand was represented by an employee.

4.2 Hypotheses Testing

We conducted a two-factor ANCOVA (appreciation: yes/no; brand attachment; high/low) on anthropomorphism, with brand personification and brand name as covariates (Table 1).

Table 1. Results

On the one hand, findings show no direct effect of appreciation on anthropomorphism (F (1, 182) = 1.33, p = .250) nor significant effect of covariates - neither from brands (F (1, 182) = 0.80, p = .372) nor from brand personification (F (1, 182) = 2.97, p = .086). On the other hand, findings show a direct effect of brand attachment on anthropomorphism (F (1, 182) = 26.57, p = .000) as well as an interaction effect between appreciation and attachment on anthropomorphism (F (1, 182) = 7.19, p = .008). Because of this interaction effect, we can say brand attachment has a moderating effect on the relation between appreciation and anthropomorphism.

However, a simple effects analysis shows that appreciation has a significant positive effect on consumers with low brand attachment (F = 7.10, p = .008) but not on consumers with high brand attachment (F = 1.08, p = ns). Consumers who are not particularly attached to the brand feel like the brand is more human when the brand uses appreciative expressions (vs. when the brand does not use appreciative expressions), but appreciation has no effect on consumers who are attached to the brand.

Therefore, we partially validate our first hypothesis: brands are perceived as more human when they use appreciative expressions, compared to when they do not use appreciative expression, yet we found a significant effect for weakly attached consumers only. Moreover, we validate our second hypothesis: when consumers are attached to the brand, the use of appreciative expressions addressed to others influences brand anthropomorphism less than when consumers are not attached to the brand. Indeed, we found no influence of appreciation on anthropomorphism when brand attachment is high, versus a positive influence when brand attachment is low.

5 Discussion and Conclusion

While there is a lot of research on the effects of social media in general, little is known about the actual interactions taking place between brands and consumers. Using an experimental method, we identify a causal relationship between brand interaction strategies and consumers’ attitude. This research contributes to two streams of literature: brand-consumer interactions and brand anthropomorphism. It also provides managerial implications in the area of community management.

First, this research shows that face-work theory applies to brand-consumer interactions on social media, revealing the potential impact of face-flattering acts on consumers’ perception in interactive brand communications. Face-flattering acts such as positive expression or compliments are supposed to respond to one of the fundamental aspirations of individuals during an interaction: the desire to be accepted (Goffman 1967). Consistent with face-work theory, consumers react positively to the brand’s appreciative expressions, yet we found this effect only when brand attachment is low.

What’s more, we found a specific manifestation of face-work in a marketing context. Findings show, in a somehow non-intuitive way, that consumers who are attached to the brand do not react in a positive way when they see the brand showing appreciation to others. Do consumers who are strongly attached to the brand develop a form of jealousy towards other consumers? Research in psychology supports the fact that, because of their emotional bond to the brand, attached consumers may prefer to be appreciated by the brand rather than to see the brand showing appreciation to strangers (Vonk 2002).

Our results could also be examined in the context of advertising literacy. Consumers who are advertising literate show knowledge, expertise and a critical viewpoint toward brands’ communications (O’Donohoe and Tynan 1998). Moreover, a consumer’s level of advertising literacy has an impact on the way they respond to brand messages (Livingstone and Helsper 2006). Can we therefore say that consumers who score higher on brand attachment also score higher on advertising digital literacy, at least toward the object of their attachment? Indeed, advertising literacy can emerge from cumulative experience with brands’ communications (O’Donohoe and Tynan 1998). Therefore, attached consumers, with greater knowledge and experience of the brand they favor, could better decode the brand’s social media strategies compared to non-attached consumers.

Second, this research highlights the impact of interaction strategies on anthropomorphism, a tendency for individuals to see humanlike characteristics or feelings in nonhuman agents (Epley et al. 2007). Anthropomorphism is important for marketing researchers and practitioners alike as it is the foundation of brand personality (Aaker 1997) and consumer-brand relationships (Fournier and Alvarez 2012). In a literature review on brand humanization, MacInnis and Folkes (2017) stress that visual, verbal and rhetorical clues can help to humanize brands. Therefore, we contribute to the brand anthropomorphism literature by showing that language in interactions can generate anthropomorphic reactions, as long as the brand conforms to interpersonal interaction norms. In addition, we contribute to the literature on brand personification (see Cohen 2014), stressing the idea that the use of a human spokesperson does not always have positive effects (Fleck et al. 2014) and is not sufficient to humanize a brand (Andriuzzi 2016). Indeed, findings show that in brand conversation it is important for the spokesperson to act as a human by respecting interpersonal communication rules.

As for managerial implications, our research shows that brand representatives, such as community managers, should carefully adapt their interaction style to the context. For example, they could segment their conversational platforms depending on whether they are talking to loyal customers or whether they are talking to new customers or prospects. These recommendations are important when managing sponsored posts campaigns where community managers may have to deal with multiple audiences, a persistent issue in online communications (Schlosser 2005).

In conclusion, this research shows the impact of brand conversation on consumers’ attitudes. Marketers face the constant development of new interactive tools, such as connected objects or virtual reality devices. In addition, innovation in the field of artificial intelligence leads to the emergence of increasingly sophisticated automated conversation assistants such as chatbots. As verbal interactions between brands and consumers occurs in a growing number of environments, our work encourages new research on the role of conversation in marketing.