Credibility, audacity and joy: Brand personalities that connect users to social media

The main purpose of this article is to evaluate the influence brand personality has on customer loyalty in the context of social networks as brands. We conducted a survey of 268 social networks users in Brazil and analyzed data through Structural Equation Modeling. As a result, brand personality dimensions Credibility, Audacity and Joy were found to be predictors of loyalty, the main aspect governing customer relationship perception. Credibility is the main predictor of customer loyalty, shedding light on network trustworthiness, user privacy and personal information safety, while Audacity indicates the power of innovation, and Joy points out the relevance of social networks’ entertainment atmospheres. Together, these dimensions are strategic aspects to be considered in the development of relationships with users in the digital world. Our finding contributes to the body of knowledge interested in relationship marketing and its relation with strategy and sustainable competitive advantage. We also investigated social networks as brands, a topic of major impact in the development of a literature on the digital world. Lastly, the results serve as a source of information about social network user behavior, helping companies enhance their communication strategies and achieve customer loyalty.


Introduction
Over the years, the way people communicate with each other has changed. Social networks were widely adopted as social interaction platforms, a practice that created a new reality for businesses, that may use these networks to promote their products, acknowledge their public and develop a closer relationship with clients. Social networks transformed the way people interact globally. This digital environment can be described as a complex context still lacking deeper investigation, given the novel character of this content in scientific literature (Kleineberg & Boguñá, 2016).
In this context, the use of social networks as tools in customer relationship development has become a frequent phenomenon in companies, bringing them closer to their clients (Qualman, 2010). This tool creates customer value through advertising, public relations, content creation, sales, customer service and user support, besides the opportunity of collecting information that will enable the development of new offers (Culnan, Mchuch, & Zubilla, 2010). Companies benefit from this in terms of user targeting, viral marketing, cost reduction, profitability, user satisfaction and customer retention.
Because of this scenario, competition among social networks demands new differentiation strategies, in order to conquer customer preference and loyalty. In this context, the scientific production relating social network and marketing strategy has been valuable for marketing research companies, organizations, and brands in different market segments (Mostafa, 2013), showing that social networks are a useful platform for firms/brands and user interaction despite how no attention being given to social networks as brands, and how they relate to their customers: the users. Additionally, competition between social networks engenders a literature gap, and its relevance lies in the fact that these companies constitute building blocks in innovation development and in the transformations lived by consumer societies (Kleineberg & Boguñá, 2016). Specifically, technological advances create new customer profiles, demanding companies learn how to sensitize them (Kimura, Basso, & Martin, 2008).
Thus, the comprehension of customer dynamic with reference to new technologies is paramount to marketing research. Each social network is recognized by a particular characteristic that makes them stand out among the others, although this has not been enough to avoid competition in the sector, considering that it is a segment that allows the entry of competitors at all times and in which technologies can be easily copied (Pacanhan, Chiusoli, & Stahl, 2007). Moreover, for Kleineberg and Boguñá (2016), a brand only exists in digital context if it is able to attract and keep user attention.
So, it is plausible to say that each social network is a different brand competing in a pool of innovation, and for this reason, we decided to investigate social networks as brands under relationship marketing theory, particularly in terms of the aspects that make users loyal to these brands. Grönroos (2009) and Sheth and Parvatiyar (2002) corroborate by presenting relationship marketing as a strategic approach placing the customer first and changing the marketing role from manipulating customers to making a real commitment to them. The authors emphasize the retention of profitable customers, multiple markets and an approach regarding multifunctional marketing, where customer loyalty plays an important role as a relevant indicator of well-built relationship marketing.
Although the body of knowledge on relationship marketing covers several relevant aspects that tell us the way brands and customers relate, loyalty is a building block in this literature (Agariya & Singh, 2011;Berry, 1995;Ndubisi, 2007;Oliver, 1999). Oliver (1999, p. 33) stated that satisfaction was not the "king" anymore, and researchers should investigate other mechanisms that guide customer loyalty. Considering our research context, we wonder: what makes users loyal to social networks?
As proposed by Fournier (1998, p. 343), loyalty is "a long-term, committed and affect-laden partnership", and brand personality is a construct that can help bring together brands and customers, given the characteristics of the brand seen as positive by customers in a long-term relationship. In other words, brand personality enables customers to legitimate a brand as a partner. This construct represents a set of human characteristics associated with a brand, being a measure of the emotional bond that brands and customers have (Aaker, 1997).
Brand personality is a construct that assumes a symbolic function (Keller & Lehman, 2006), which can be used in examining brand relationships (Fournier, 1998;Hankinson & Cowking, 1993). In addition, customers frequently use brands that are part of people's lives as a way to deliver a message about themselves to others (Jerónimo, Ramos, & Ferreira, 2018). These arguments showed us a possible relation between relationship marketing and brand personality, given that brand identification may open an opportunity of a closer relationship between brand and consumer. There have been efforts at understanding social networks as a marketing tool in organizations (Carneiro, Simões, & Felipe, 2013;Soares & Monteiro, 2015;Thackeray, Neiger, & Keller, 2012), but little attention has been dedicated to the study of social networks as brands looking for competitive advantages.
Considering the complexity of digital environment competition, it is imperative to understand how relationship marketing and brand personality relate in the context of social networks, and for this, we propose to investigate brand personality as a driver of customer loyalty. Thereafter, our main purpose is to evaluate the influence of brand personality on customer loyalty in the context of social networks.

Brand Personality
From the moment that brands started to be considered a competitive advantage source, literature has concentrated efforts on identifying brand-related aspects that create brand preference (Biel, 1993;Brito, 2010;Gardner & Levi, 1955;Högström, Gustafsso, & Tronvoll, 2015;Scussel & Demo, 2016;Sirgy, 1982). Among these efforts, there is the concept of brand personality, a set of characteristics that define a brand in the same way as perceiving someone's personality (Aaker, 1997).
Brand personality represents the set of human characteristics customers perceive in a brand, in a similar way they would describe someone's personality (Batra, Lehmann, & Singh, 1993). Although brand personality has its roots in psychology and it engenders a useful metaphor to describe customers' perceptions towards a brand (Caprara, Barbaranelli, & Guido, 1998), brand personality is a construct belonging to a market perspective, with the aim of understanding the impact of brand personality on consumer behavior (Azoulay & Kapferer, 2003;Scussel & Demo, 2016).
The relevance of this construct in marketing research has been proven. Studies suggest the relation between brand personality and consumer self-concept (Jerónimo et al., 2018); the use of brands for customers to express their beliefs and personalities (Diamantopoulos, Smith, & Grime, 2005;Park & John, 2012); and the influence of brand personality on product and services preference (Swaminathan, Stilley, & Ahluwalia, 2009). Brand personality was found to be a component of competitive strategy formulation (Malär, Nyffenegger, Krohmer, & Hoyer, 2012), an element used in brand positioning (D. H. Kim & Sung, 2013) and a relevant construct used by advertising in consumer persuasion (Park & John, 2012). Our paper will be based on the second perspective.
It was Jennifer Aaker, in the end of the 1990's, who changed the research tradition on brand personality (Azoulay & Kapferer, 2003). In her work, Aaker (1997) conceptualized brand personality in the context of marketing literature, operationalizing the construct through a measurement scale of brand personality, an instrument that has guided the latest findings in this content (Scussel & Demo, 2016). Aaker (1997) identified five brand personality dimensions in the North American context, namely, Sincerity, Excitement, Competence, Sophistication, and Ruggedness. Although the scale is a reliable and valid instrument, Aaker (1997) advises the need of scale validation when using the scale in different social and cultural contexts.
In this sense, these authors consider the scale from Muniz and Marchetti (2012) a building block in brand personality research in the Brazilian context, enabling relational studies with other marketing constructs, such as customer loyalty. Delmondez, Demo and Scussel (2017) have confirmed there is a relation between brand personality and customer relationship perception in the food and beverage sector, revealing important aspects of consumer behavior. Their findings reinforce that customer loyalty is a subject of major importance in relationship marketing studies, and they took the first step into investigating brand personality as a driver of customer loyalty.
Considering the necessity of understanding consumer behavior in the digital environment context and the absence of investigations on user and brand relationships, we contribute by associating brand personality and customer loyalty in using social networks as brands. McKenna (1991) presented the relationship marketing concept in his seminal work as a paradigm shift on marketing concepts, a change on marketing orientation from just attracting customers to having customer retention and, more importantly, customer loyalty. According to Berry (1995), the main purpose of competitive organizations and brands is to build a long-term relationship with customers, transforming them into loyal clients.

Customer Loyalty
The most adopted concept of customer loyalty is a deep commitment to repeatedly buy or recommend a certain product or service in spite of situational influences and marketing efforts potentially capable of causing behavioral changes (Oliver, 1999). Besides, the assessment of customer loyalty drivers is a topic of great interest, since satisfaction has proven insufficient at keeping clients loyal (Oliver, 1999;Reichheld, 2003), and in the business arena, customers can be loyal to a diverse range of brands (Sirdeshmukh, Singh, & Sabol, 2002). For companies, this means the development of long-term relationships with clients and influencing their behavior and choices (Watson, Beck, Henderson, & Palmatier, 2015). Scussel, Petrol, Semprebon and Rocha (2017), in an attempt to unify concepts, state customer loyalty as a defining construct of relationship marketing, a business philosophy based on the holistic interaction between customers and firms/brands, creating customer value through the engagement between them.
The most prominent feature of the customer loyalty field of study is its theoretical and operational diversity (Russo, Confente, Gligor, & Autry, 2016;Toufaily, Ricard, & Perrien, 2013;Wolter, Bock, Smith, & Cronin, 2017). Consequently, these researchers agree that customer loyalty should be studied contextually, in other words, it should be operationalized according to the setting in which the construct is investigated. In this sense, literature shows that customer loyalty is studied as repurchase (Bove & Johnson, 2009), attitude towards a brand or a firm (Chacon & Mason, 2011) or by the level of loyalty that customers demonstrate (Dick & Basu, 1994;Ngobo, 2017).
Although these works find resonance in relationship marketing on a theoretical basis, none of them aim at measuring the most relevant aspects that make customers engage in a solid relationship with brands. Additionally, we must also recognize that customer loyalty can be temporary, since clients are loyal if they perceive benefits, and only if these benefits outweigh those proposed by competition ( , 2008). There is no loyalty if customers don't perceive they have a relationship with a brand; if they do not recognize their interactions and the benefits they get from it (Grönroos, 2009).
This debate drives us to customer relationship perception, a concept that comprises those aspects considered by customers when deciding to engage in a relationship with a company or a brand (Demo & Rozzett, 2013;Rozzett & Demo, 2010). Literature agrees that customer relationship perception influences customer willingness to interact with companies (Becker-Olsen, Taylor, Hill, & Yalcinkaya, 2011;Lafferty, Goldsmith, & Hult, 2004;Souza & Mello, 2009;Wong & Sohal, 2002), based on technical, functional, behavioral and emotional aspects (Grönroos, 2017).
Recently, Ngobo (2017) revealed that customer loyalty can't be measured the same way over time, since contextual conditions, such as social and economic, change. Considering the setting in which this research is localizedsocial networksand the need to explore the relationship between these brands and their users, we have decided to use the customer relationship perception scale validated by Demo, Watanabe, Chauvet and Rozzett (2017). This instrument has proven its stability and internal structure and external validity in Brazil (Rozzett & Demo, 2010), US (Demo & Rozzett, 2013), and France (Demo, Watanabe, Chauvet, & Rozzett, 2017), with reliable psychometric indices and theoretical support, which proves its ability in measuring customer relationship perception.
However, we only used the items related to the customer loyalty factor, since the items from the customer service factor do not apply to the context of social networks. Given the brands studied in this paper are related to social networks, attendants don't have personal contact with customers. In consequence, the items belonging to the customer loyalty factor, from the customer relationship perception scale , compose the measurement instrument used in our paper. The same strategy has been used by Lima, Demo and Scussel (2017).

Methods
To analyze the influence of brand personality on customer loyalty in the context of social networks, we conducted a survey, considering brand personality in its five dimensions for the Brazilian customer (Muniz & Marchetti, 2012)credibility, joy, audacity, sophistication, and sensitivityas independent variables and customer loyalty, as proposed in Demo et al. (2017), as the dependent variable.
Social network users composed the sample for this research. To estimate the minimum sample in order to test the predicted relationship between variables, we resorted to Cohen (1992), who considers four criteria in sample calculations: sample size (N), criterion of significance (α), effect size (ES) and statistical power. Considering a 0.8 statistical power, a medium effect of 0.25, and 0.05 of significance, the minimum sample calculated was 118 subjects. The final sample was composed by 268 questionnaires, with a statistical power of 0.99.
Regarding data collection, we used an online Typeform platform questionnaire through social media. The first question of the questionnaire asked the subject to indicate a social network to be evaluated, followed by 28 items from the brand personality scale (Muniz & Marchetti, 2012) and 13 loyalty items from the customer relationship perception scale . Finally, sociodemographic questions were proposed in order to characterize the sample.
In relation to data analysis, we chose to perform Structural Equation Modelling, since Hair, Black, Babin, Anderson and Tatham (2009) affirm it is the most adequate technique for dependence relation analysis when variables have multiple dependent and independent relations. More specifically, we performed Confirmatory Factor Analysis (CFA) and Path Analysis (Byrne, 2016 The first step of our analysis concerned data treatment, following Tabachnick and Fidell's (2013) recommendations, starting with frequency distribution analysis. For missing values, we performed the listwise procedure, excluding 38 questionnaires in this stage. To identify outliers, we used the Mahalanobis method: based on the chi-square table and using a significance index of p<0.001, the value of X² was 80.077. Hence, 12 outliers were identified and eliminated, generating a final sample of 268 subjects, formed mostly by women (60%) between 18 and 28 years old, with college education (73%). The subjects are social network users for at least five years (65%), accessing their favorite social network daily (93%). Social networks most cited by our sample were Facebook (31%), Instagram (24%) and WhatsApp (21%).
Lastly, we generated normal probability and residual plots. In this way, we verified linearity, homoscedasticity and the normality of the distribution of error terms. As a result, confirmation of all assumptions was obtained.

Results and Discussion
First, we performed a Confirmatory Factor Analysis (CFA) for both instruments used in this articlethe Brand Personality scale and the Loyalty factor items from the Customer Relationship Perception Scalein order to identify the measurement model and, from this, proceed with path analysis. In this step, we used the maximum likelihood method following Brown's (2014) guidelines.
In Table 1, we present the results from the CFA performed for the Brand Personality scale. It shows the brand personality dimensions, estimates, standard errors (S. E.), critical ratio (C.R.) and the standardized regression weight. All items in the scale were significant (p-value <0.01).  Next, Table 2 illustrates the results from the CFA performed for the Customer Loyalty factor items from the Customer Relationship Perception scale.  Hair et al. (2009), factor loadings must be higher than 0.5. The items L5 (This social network is fast in problem solutions), L9 (The advertisements of this social network is in line with its offers), L12 (This social network offers me personalized service), L13 (This social network's prices are fair) and L14 (This social network is the best option when compared to its competitors) achieved loadings under 0.5. For this reason, they were taken out of the analysis.
According to Tables 1 and 2, the correlations were significant (p-value<0,01; CR>|1,96|). Thus, the 37 items from both scales were distributed over five brand personality dimensions and one customer relationship perception dimension (loyalty). Our results corroborate the findings from Muniz and Marchetti (2012) and Demo et al. (2017).
The following stage was concerned with verifying the modification index (M.I.), following Kline's (2011) guidelines. The M.I. between C3 and C2 was 95.0 and between J3 and J4 was 45.05. For this reason, a double arrow was inserted between these items, indicating a positive correlation between them.
Next, we verified index fit. As stated by Byrne (2016), the ratio between chi-square and the degrees of freedom shall not exceed 5 (CMIN/DF). The Comparative Fit Index (CFI) rates between 0 and 1, and the closer to 1 that a value is, the better the fit is. The parsimony fit index Root Mean Square Error of Approximation (RMSEA) indicates an adequate fit model when its value is less than 0.1. Finally, the absolute fit index Standardized Root Mean Square Residual (SRMR) is the difference between observed normalized correlation and predictable correlation, which must be less than 0.1 (Marôco, 2010). Therefore, we can ensure that the obtained structure in this research has a satisfactory fit (CMIN/DF=1.89; CFI=0.89; RMSEA=0.05; SRMR=0.06). In this sense, the measurement model was confirmed, enabling path analysis. The next step concerned verifying the convergent and discriminant validity of brand personality scale. Regarding convergent validity, Table 1 shows that most items have a standardized estimative over 0.50, as recommended by Hair et al. (2009). Besides, Jöreskog's rho was higher than 0.7 for all dimensions (Credibility=0.86; Joy=0.91; Sophistication=0.90; Audacity=0.82; Sensitivity=0.87). Therefore, convergent validity was confirmed.
Concerning discriminant validity, the square root of the extracted variance must be greater than the correlations between the dimensions, as stated in Hair et al. (2009). Table 3 shows that discriminant validity was confirmed as well. The extracted variance from of the brand personality's dimensions were above 0.5, reaching Fornell and Lacker's (1981) criteria. As a last step, we verified convergent validity for the Customer Loyalty scale used in this study. The Jöreskog's rho was 0.83, above recommendations from Hair et al. (2009), and the items with factor loadings under 0.5 were deleted from the model. The remaining items had loadings above 0.5, however BAR, Rio de Janeiro, RJ, Brazil, v. 15, n. 4, art. 5, e180088, 2018 www.anpad.org.br/bar the extracted variance was 0.35, below what is recommended (Fornel & Lacker, 1981). Despite this value, the following analyses were not compromised, since two criteria of convergent validity were reached.
Considering loyalty as the dependent variable and the five brand personality dimensions as independent variables, we calculated the initial structural model, as Table 4 shows.  Table 4 shows that Credibility, Joy and Audacity significantly predict consumer loyalty. It is important to mention that brand personalities Sophistication (p-value=0.25) and Sensitivity (p-value=0.27) were not significant in the explanation of loyalty to a social network. This means that characteristics such as chic, fancy, glamorous or romantic do not interfere in consumer loyalty to social media.
As proposed by Muniz and Marchetti (2012), brands with sophisticated characteristics are known to offer expensive products and services to build their elegant and exclusive image. Social networks follow an opposite logic: registration is usually free, and costs are optional only if users want to use special features. Additionally, social networks are used globally for millions of users, in contrast with the exclusive atmosphere provided by sophisticated brands.
According to Muniz and Marchetti (2012), brands perceived as sensitive are connected to emotions and sensitive elements, like delicate, romantic and charming characteristics. This helps us to understand the low result for Sensibility dimension, since social networks indicate more fun and enthusiastic brands, considering that users connect to social networks to have a good time, focusing on entertainment based on fun (Ferreira & Arruda, 2015).
Next, we performed a new path analysis. This time, only the three significant dimensions were used. Table 5 illustrates the final structural model. From the data on Table 5, we observe that Credibility, Audacity and Joy were significant in loyalty prediction, with Credibility being strongest (ß=0.50). Standardized coefficients or beta BAR, Rio de Janeiro, RJ, Brazil, v. 15, n. 4, art. 5, e180088, 2018 www.anpad.org.br/bar coefficients (β) enable the comparison between coefficient and dependent variable, revealing the magnitude and the direction of the relationship between predictors and the dependent variables (Hair, Black, Babin, Anderson, & Tatham, 2009). Also, the model had a R² of 56%, which means that 56% of customer loyalty in the context of social networks as brands can be explained by brand personality; more specifically, by Credibility, Audacity and Joy dimensions.
In accordance with Cohen (1992), the coefficient of determination's (R²) statistical significance can reveal a small effect (2%), medium effect (13%) or major effect (26%). Based on the results from Table 5, it is possible to affirm that brand personality dimensions influence 56% of the explication for loyalty variable, which is a major effect. Lastly, fit indexes were checked once again, confirming the satisfactory fit (CMIN/DF=2.04; CFI=0.90; RMSEA=0.06; SRMR=0.07). The structure model is presented by Figure 1.

Figure 1. Structural Model
The analysis revealed Credibility (β= 0.50) as Loyalty's best predictor, showing that credibility is the most important aspect perceived by customers when considering being loyal to a social network. From this, it is reasonable to declare that users want to feel safe using social networks. Literature confirms these findings: when a brand is seen as trustworthy, customers tend to trust their personal information to the brand, increasing the chances of a long-term relationship (Berry, 1995;Rambalducci, Borinelli, & Oliveira, 2012). Our results also reinforce the finding from Mosteller and Poddar (2017) and Huang and Chen (2018). The first study confirmed the positive influence of trust in user engagement in social networks, a fundamental aspect in customer relationship developing. In a more specific way, Huang and Chen (2018) confirmed there is a relation between consumer trust and loyalty on Facebook.
Under this perspective, social networks users are more susceptible to creating profiles in social networks known for their trustworthiness and safety (β= 0.50). It is relevant to notice that Credibility is a dimension in constant need of monitoring, since any problem related to information safety can negatively influence brand image, especially considering the sharing characteristic of social networks (Bentivegna, 2002). As this author proposes, the impacts of a situation like this would directly affect consumer loyalty, resulting in user, advertiser and sponsorship loss. BAR, Rio de Janeiro, RJ, Brazil, v. 15, n. 4, art. 5, e180088, 2018 www.anpad.org.br/bar Given the above-mentioned reasons, it became common for social networks to adopt safety policies related to user personal information. In consonance with Bertot, Jaeger and Hansen (2012), the connection made via social network can generate new ideas, increase service offers and collect information about individuals that influence companies and brands' decision-making processes. Nonetheless, these authors argue that online interaction sets challenges related to privacy, safety, information management and accessibilityand the way these safety policies are presented to users are different, which may interfere in consumer perception of credibility. In this sense, the more consumers perceive these policies being put into practice, the more credibility is associated with the social network brand.
The dimension Audacity (β= 0.28) was the second-best predictor of loyalty in the context of social networks as brands. In agreement with Muniz and Marchetti (2012), brands that share the Audacity dimension have traits like modern, fearless, creative and updated. These brands also tend to frequently innovate, leading to consumer preference through creative and original offers and aggressive communication. Based on this, the presence of such attributes in a social network increases the possibility of a user being more loyal to its brand. In other words, new and updated tools proposed by competition can easily outshine a social network that does not innovate or show different and modern offers.
Pursuant to Culnan, McHugh and Zubillaga (2010), the information technology sector, which embraces social networks, is known for companies' innovation capacity, adopting the most recent technologies available. In this context, technology adoption and implementation involve risks that may jeopardize brand market value, reinforcing the idea we defend in this article that social networks, as active elements of this complex online environment, consist of modern and innovative brands, which justifies the results for the Audacity dimension.
Similarly, Formiga Sobrinho and Barbosa (2014) assert that what makes a social network attractive and capable of user retention is how creative it is in interacting with users and proposing new tools to connect people in digital worlds. According to them, creativity connects users and social networks, demanding companies innovate on a regular basis if they want to build and keep a relationship with their clientsthe users. Besides this, the sense of belonging, interactivity and innovativeness, or affinity, are basic user expectations when connecting to a social network (Krishen, Berezan, Agarwal, & Kachroo, 2016).
Joy, the third predictor dimension of loyalty (β = 0.18), is a brand attribute that catches consumer attention and develops a more informal relationship with clients, through a more relaxed atmosphere. Following Muniz and Marchetti (2012), brands with the Joy dimension are perceived as happy, fun and humorous. In this sense, De Toni and Schuler (2007), while studying the development of technological products, identified that consumers value products able to inform, solve problems and provoke feelings like freedom, pleasure, joy and companionship. Moreover, Ferreira and Arruda (2015) explain that users resort to social networks for hedonistic reasons, with the objectives of having fun, interacting with people and creating and sharing ideas unpretentiously. Finally, Vries, Peluso, Romani, Leeflang and Marcati (2017) emphasize that consumers engage in activities, such as creating their own online content in social networks, in a search for self-expression and socialization, something perceived as pleasurable for them.
The question that guided our work in this paper wondered about the relation between brand personality and customer loyalty in the context of social networks, and our findings reveal a strong relation between the variables (R²= 0.56), showing, according to Cohen (1992), a strong effect of credibility, audacity and joy in the way customers perceive their relationship with social networks. This result finds theoretical support in Culnan et al. (2010), who affirm that when individuals identify themselves with a social network, they develop a sense of responsibility for this online community where they interact with other users, stimulating their presence in the network and, in this sense, the development of a virtual relationship.
When Cheung and Lee (2012) discuss what drives consumers to choose a social network, they reveal the sense of belonging created in the online social environment, the construction of a reputation in their social groups and the entertainment promoted by the digital world are important matters. Considering our research context, we might interpret that social networks are seen as partners in the user lives, which denotes the existence of a relationship between the customers and the social networks used. We can also imply that there are some characteristics of social networks helping this relationship grow, in other words, users will connect and be loyal if the social network is a fun place to stay, if they identify with its purpose and, most of all, if it is reliable in usage.
Along these lines, we interpret that the more users perceive social networks as trustworthy, innovative and fun, the higher their loyalty will be. Our findings highlight the role of brands in the development of marketing strategies. We went beyond this by confirming the power brand personalities have in influencing relationships perception as well. Furthermore, we present credibility, audacity and joy as important antecedents of customer loyalty, especially in a context where there is no money exchange and satisfaction cannot be evaluated by service quality only. Lastly, our findings innovate by indicating that customers perceive social networks as not only the context in which they communicate with their groups, companies and brands, but also as respectable, modern and lively brands they are willing to relate with.

Conclusion
The main purpose of this article was to investigate the relation between brand personality and customer loyalty in the context of social networks as brands. As a result, we found that the dimensions Credibility, Audacity and Joy are predictors of the perception users have about their loyalty to social networks. In this way, it is possible to say that brands that transmit these personality traits have higher chances of developing long-term relationships with their users, conquering their loyalty.
Regarding academic contributions, this article is an advance in the study of relationship marketing, a body of knowledge of great interest for scholars and marketing practioners, given its relation with strategy and sustainable competitive advantage. Customer loyalty is the foundation of this theory, but also its main goal. Following this, we have revealed an antecedent of this construct, reinforcing that customer perceptions, evaluations and feelings must be at the top of the marketing research agenda.
An important advancement of this study is the consideration of social networks as brands. Social networks are usually investigated as the context in which transactions between users and firms/brands happen and relationships grow. Here, we studied social networks as active actors in this relationship, as since they have personality traits, they are subject to customer attention, preference and perception. Hence, our paper sheds light onto the way customers perceive the social networks they engage with and how they evaluate their relationship with these online environments -topics of major impact in the development of a literature on the digital world, including shared economy services.
The first managerial implication of this paper concerns social networks as brands. This document is a source of information about what drives consumer loyalty and could be used to develop new tools to engage with users and to provide a better user experience, personalize customer accounts, charge for special features and communicate special offers to users. This would help social networks enhance their relationship with customers and improve their brand positioning to face competition.
Other companies and brands, which use social networks as the context to interact with customers, can use our findings to better acknowledge user behavior, not only improving their relationships with customers, but reaching new audiences by promoting the characteristics perceived as important by them credibility, audacity and joyin their marketing strategies, especially in advertisement and BAR, Rio de Janeiro, RJ, Brazil, v. 15, n. 4, art. 5, e180088, 2018 www.anpad.org.br/bar communication. Likewise, brand managers can benefit from our results by using brand personality dimensions to build an image with their public, connecting with the users through brand identification.
A limitation faced in this study is the cross-sectional nature of the data, which gives us an overview of one data collection and does not allow data generalization. In this sense, the development of a time-series database would be relevant to test the development of customer relationship perception through different points of interaction, considering the fast-changing scenario where social networks belong and the innovative capacity of this environment. Additionally, we believe that the quantitative nature of our data only enables statistical inferences, and a deeper analysis of user behavior in the online context, considering social network consumption, would provide a better diagnosis of this phenomenon.
Considering the relationship between social networks as brands and their users is a recent but expanding phenomenon, we suggest that future investigations resort to other research approaches and qualitative methodologies in order to achieve a greater understanding of this relationship. In addition, multi-method surveys, which combine qualitative and quantitative approaches, including methodological triangulation, are very welcome since they produce a better understanding of the phenomenon, while efforts to either understand or measure it are engendered. Research results such as those presented here can help companies develop marketing strategies that connect users more effectively with social networks in order to improve customer loyalty, which will possibly translate into better organizational results.

Contributions
The first author was the creator of the paper. Thus, she contributed to the development of the introduction and theoretical background, and the analysis and discussion of the results. In addition, she participated in the final review. The second author contributed to the development of the introduction and theoretical background. She was responsible for data collection. The third author participated in the data analysis and discussion of the results. In addition, she contributed to the final review. The fourth author contributed to the discussion of the results and was responsible for paper translation. In addition, she contributed to the final review.