A recent definition states that relationship marketing is as a business philosophy that aims to promote interaction between customers and companies in a holistic way, creating value to the customer by the engagement between them (SCUSSEL et al., 2017). The central concept of this definition is interaction, and managing these interactions enables firms to develop long-term relationships with customers. According to Grönroos (2009), companies need to understand their needs, habits and lifestyles in order to create attractive value proposals, and relationship marketing and its interactional and holistic approach gives companies the tools to achieve that.

Recently, Grönroos (2017) proposed a relationship marketing theoretical model composed by three processes. The first one concerns customers' value creation processes, a set of all encounter points between company and customer that enables relational benefits to arise. The second one is the interaction process, regarding the actions and interactions that aim to maintain and enhance their relationship. Lastly, there is the communication process, where sales, marketing communication, digital marketing, mass communication and formal offers are presented to the customers so they can build the image of the company and their perceptions about it.

From a consumer perspective, relationship marketing is formed by three main defining constructs: customer trust, customer satisfaction and loyalty (SCUSSEL et al., 2017). The authors say that companies need to show customers the relational benefits they can have, in order to help them to decide if they want to relate with the brand. As Grönroos (2009) explains, the decision about relationship with a brand is a customer decision. To this scholar, companies can offer customers value proposals, provided by relational benefits customers perceive, but only customers can say if they relate or not with a company or a brand. In this sense, interactions are the basis of a relationship, once they are opportunities for companies to have access to customers' information, which will guide companies' efforts on value creation (GRÖNROOS, 2009).

Considering the above, customer relationship perception presents itself as an important construct to relationship marketing literature. Research on this topic acknowledges that both parties - customer and company - need to perceive their share in a relationship (WONG; SOHAL, 2002; SOUZA NETO; MELLO, 2009). As stated by Grönroos (2009), a relationship does not exist because the firm says so, but only when the customer recognizes the nature of this bond. Literature on consumer behavior endorses this: when customers perceive to have a relationship with a brand and feel they can benefit from it, they are more likely to buy than when they do not feel in a relationship with the brand (LAFFERTY; GOLDSMITH; HULT, 2004; BECKER-OLSEN; TAYLOR; HILL; YALCINKAYA, 2011). Recent findings show the way companies manage their interaction with customers affects not only the way customers perceive their relationship but also their willingness to continue the relationship (GRÖNROOS, 2017).

Rozzett and Demo (2010) conducted studies in Brazil and in the United States (DEMO; ROZZETT, 2013) to develop and validate a scale to measure customer's perception of their relationships with companies, obtaining reliable psychometric indices. Recent literature points out studies of validation of their work on specific sectors, such as amusement parks (VASCONCELOS; DEMO, 2012), electronic games (DEMO; BATELLI; ALBUQUERQUE, 2015) and the beverage sector (DEMO; LOPES, 2014). These studies certify the internal structure of the scale, proving its ability to measure customer relationship perception.

According to Fournier (1998), people increasingly tend to build relationship with brands based on the compatibility between consumer and brand, attributing brands with human characteristics. On this matter, Aaker and Joachimsthaler (2000) admit that one way that consumers can identify a brand in their minds is through the perspective of the brand as a person, which places brand personality as a central element.

Since brands are targets of preferences, expectations, and attributions, Caprara, Barbaranelli, and Guido (2001) argue that it is plausible to use the same heuristic that leads to the formation of an impression about a person to the formation of an impression about a brand. For Aaker (1997, p. 347), brand personality is the "set of human characteristics associated with a brand," where the symbolic use of brands is explained by the fact that consumers assign brands with human personality traits that identify and differentiate those brands.

Under marketing studies, recent empirical production highlights brand personality as an important component of competitive differentiation strategies (MALAR; NYFFENEGGER; KROHMER; HOYER, 2012), bringing out the influence of symbolic consumption in the design of business strategies (NOBRE; BECKER; BRITO, 2010). In addition, we found a body of research on the effects of brand personality on brand positioning strategies (KIM; SUNG, 2013), the development of persuasion in advertising (PARK; JOHN, 2012), and the use of celebrities for brand endorsement, exploring the influence of brand personality traits (ARSENA; SILVERA; PANDELAERE, 2014). These findings reveal brand personality as a predictor of a series of marketing and consumer behavior constructs, yet no studies were found on the relation between brand personality and relationship perception.

Regarding brand personality measurement, Aaker (1997) proposed the development of Brand Personality Dimensions, a scale in which different types of brand personality can be identified. The results indicate that individuals in the United States perceive brands in terms of five personality dimensions: Sincerity, Excitement, Competence, Sophistication, and Ruggedness. Literature agrees that the research tradition in brand personality took a new direction with Aakers's scale, noting that brand personality consists of a specific construct for each culture (AZOULAY; KAPFERER, 2003).

Muniz and Marchetti (2012) conducted a validation study of Aaker's original scale (1997) in Brazil, identifying five brand personality dimensions in the Brazilian context: Credibility, Joy, Audacity, Sophistication, and Sensitivity. Nonetheless, there are still only a few studies on brand personality in Brazil (SCUSSEL; DEMO, 2016). These authors justity the need to develop works in this topic, once it can help develop consumer beharvior understanding in this context. They also state that the instrument validated by Muniz and Marchetti (2012) is a first step on the development of brand personality studies in Brazilian marketing literature, allowing case studies and relational researches with other marketing constructs.

The luxury industry engender a highly competitive market, which has expanded the last decade, with notable participation of BRIC economies and a shift in luxury consumption habits (KO; PHAU; AIELLO, 2016). In this sense, it is paramount that investigations on the luxury segment consider the effects of the country on customer's perceptions, attitudes and willingness to buy luxury brands, especially those focused on fashion apparel (CHEAH; PHAU; ZAINOL, 2016). In Brazil, luxury brands set a promising market, once hedonic values are very important to Brazilian customers, whose sense of identity and self-image are associated with the possession of high prestige and power products (CAMPANARIO; STREHLAU, 2016).

In luxury fashion brands segment, value creation is associated with the development of strong bonds with customers through unique interactions, based on emotional and cognitive aspects (CHOI et al., 2016). This means that luxury customers' evaluations are created from their perceptions, especially those related to service quality, the anticipation of their needs and the experience in retail stores (KIM et al., 2016). In addition, luxury customers value hedonic benefits and tend to be more loyal to brands that fulfill their consumption goals and needs (STATHOPOULOU; BALABARIS, 2016).

Nevertheless, empirical research is lacking on relational aspects on luxury fashion brands market (CHOI et al., 2016). Along these lines, Ko, Phau and Aiello (2016) affirm that luxury brand market configures a growing research arena, especially in the fast-changing global economy and the changes in luxury consumption in emergent cultures.

In the LCRP Scale development, guidelines provided by Churchill (1979) and Rossiter (2002) were followed. According to Diamantopoulos (2005) and Finn and Kayande (2005), both procedures have strengths and limitations. In face of this, Oliveira and Veloso (2015) proved that a combination of both brings advantages to the scale development process. Hence, Study 1 was planned in seven steps: construct definition; empirical understanding of the construct; generation of a pilot version of the scale; scale purification; new scale purification, with a new sample; and proposition of final items.

The first step is construct definition, which takes us back to the theoretical references visited on relationship marketing to define Customer Relationship Perception. For the empirical comprehension of the construct, both Churchill (1979) and Rossiter (2002) agree that to acess construct domain, interviews can provide the comprehension of the phenomena under investigation. Considering it is difficult to access luxury brands clients, we resorted to the snowball strategy, a non-probabilistic sampling procedure that uses reference chains and is usefull to research inpenetrable social groupings (ATKINSON; FLINT, 2001). According to these authors, snowball sampling has an explorative nature and offers practical advantages to conduct interviews, which makes it adequate to our study's objective.

Hence, we conducted ten interviews with frequent buyers of luxury fashion brands aged between 29 and 49 years, all with college education. The number of respondents was defined by the saturation criterion, which occurs when the responses begin to repeat and no new data is generated, determining that the sample is sufficient (BARDIN, 2014). We used an interview scrip to promote a conversation with them, following the guidance of Castillo-Montoya (2016), with interview questions related to their relationship perception on luxury fashion brands. With the authorization of participants, interviews were recorded, transcripted and analyzed through thematic categorical content analysis as proposed by Bardin (2014).

For the generation of a pilot version of the scale, we used two methods: a previously validated scale (DEMO; ROZZETT, 2013) and the results from content analysis made on the previous step. Following the instructions from Churchill (1979) and Rossiter (2002), the pilot version was submitted to semantic analyses and judges' analyses. For the semantic analyses, a sample of 20 luxury fashion brands consumers assessed the clarity of the items. Simultaneously, eight specialists in marketing conducted judges' analysis to determine if the proposed items referred to the construct relationship perception specifically in the luxury brands context.

As Churchill (1979) states, the next step is scale purification, when the items from the LCRP Scale pilot version were submitted to exploratory factor analysis (EFA). For this, we followed the recommendations by Hair et al. (2009) and Tabachnick and Fidell (2013), who suggest that the sample size for EFA is between 5 and 10 subjects per item as a rule of thumb.

The subsequent move proposed by Churchill (1979) is a new purification of the scale. We performed confirmatory factor analysis (CFA); using covariance based structural equation modeling (CB-SEM), maximum likelihood estimation method, with the use of AMOS statistical program. Based on Coltman, Devinney, Midgley and Venaik's (2008) criteria, we built a reflective measurement model, once the customer relationship perception is an existing construct and we expect its dimensions reflect customers' perception about their relationship with brands and organizations.

Hair et al. (2009), Kline (2011), and Byrne (2009) agree that there must be between 10 and 20 subjects per variable for the CFA, but a minimum of 200 subjects is recommended. To conduct the CFA, the model validated with EFA was used. The last step concerns the proposition of final items of the scale.

To analyze the relation between brand personality and relationship perception on luxury fashion brands context, we used a brand personality scale validated in Brazil by Muniz and Marchetti (2012) and the final version of the LCRP Scale developed and validated in Study 1.

To verify the prediction between variables we used path analysis, also known as a structural model test, which aims to estimate relationships between variables through structural equations modeling, using maximum likelihood estimation criterion to check the fit of the model. The main function of the path analysis is the specification and estimation of linear relationships between variables (KLINE, 2011).

For behavioral sciences, it is important to select a minimum sample that has greater statistical power than 0.80 (COHEN, 1992). Utilizing the GPower 3.1 program and considering the predictor variable, brand personality, with five dimensions, we obtained a minimum sample of 138 subjects, considering α= 0.05 and 5 predictors.

Data was collected for both studies at the same time, once the questionnaire was composed by the pilot version of LCRP Scale (developed on Study 1) and 28 items from Brand Personality Scale (MUNIZ; MARCHETTI, 2012). Four sociodemographic items were added in order to characterize the sample. Final questionnaire had 72 items.

To proceed with this paper, we investigated frequent luxury brand female buyers, once research demonstrated there are differences between genders in luxury consumpsion, revealing that women represent 60% of consumers of the luxury market (ROUX; TAFANI; VIGNERON, 2017). Only Brazilian women who had already purchased a luxury brand could complete the online questionnaire. The first question requested them to write the name of a luxury fashion brand they were clients, so they would answer according to their relationship with that brand. We used TypeForm as the tool for data collection.

A total of 622 valid responses were obtained after removing incomplete responses and outliers. We devided randomly this 622 subjects sample in two independent samples. A sample of 300 subjects was used on the first purification of the LCRP Scale (EFA). The respondents in this sample were between 29 and 39 years old (49%), with college education and MBA (47%) and with most of them having a relationship with the brand for five years (30%), buying their products at least once a year (56%).

Another sample of 322 responses was destinated to the new purification of the scale (CFA) and also used on Study 2. Most women in this sample were between 18 and 28 years old (75%), with college education (47%), clients of the brand for at least one year (68%), with at least one purchase a year (42%).

Thus, sampling procedures to the LCRP Scale development and validation process met all literature criteria. Figure 1 illustrates the methodological procedures of this research.

Figure 1

Given that the scale development began with a construct definition, an extensive literature review on relationship marketing enabled us to define customer relationship perception and the aspects perceived by customers as relevant in their relationship with luxury fashion brands. The empirical understanding of this construct was achieved though interviews with luxury female consumers, which helped us to develop an initial list of 32 items associated with customer relationship perception in luxury brands context.

To create the LCRP Scale pilot version, we added to these results 14 items of a previously validated scale that measures relationship perception (DEMO; ROZZETT, 2013). This pilot version of LCRP Scale had 46 items. After semantic analyses and judges' analyses, the final pilot version of the LCRP Scale had 40 items, which was submitted to scale purification.

Firstly, we performed the principal component analysis to verify the viability of using exploratory factor analysys (EFA) in our sample, or factorability, checking the correlation matrix and the measure of sampling adequacy proposed by Kaiser-Meyer-Olkin (KMO). Results indicated siginificant correlations between variables and high values for communalities, confirming the matrix's factorability for our sample. The KMO index reached 0.92 and was classified as "wonderful" according to Kaiser (1974). The criteria used were the eigenvalues, the percentage of explained variance, the scree plot graph, and the parallel analysis. All criteria indicated three factors to be extracted.

From this, we iniciated the EFA, with Promax oblique rotation, once behavioral studies assume correlation between variables (PASQUALI, 2012). Next, we tested the correlation between factors, observing a high correlation between them, what can be an evidence of a second-order factor. Pasquali (2012) states that if the arithmetic mean of the correlation of the factors is above 0.3, there is the presence of a second-order factor, which is our case, once the mean of our three factors was 0.45.

Considering that the three extracted factors are indicative factors of a second-order factor, a new principal component analysis was performed, revealing one factor, followed by another analysis conducted with only one factor. As a result, the scale obtained three first order factors. As a result, LCRP Scale obtained three first order factors, namely, Brand Experience, Customer Loyalty and Brand Image. The second-order factor was named Customer Relationship Perception and it is formed by the three first-order factors above.

Next, we evaluated the psychometric indices of the scale by its validity, reliability and total explained variance (HAIR et al., 2009). Scale validity was analized by checking the factor loadings of its items. Comrey and Lee (1992) indicate 0.55 as a minimum loading, explaining that loadings between 0.55 and 0.62 are classified as good; between 0.63 and 0.7 are very good; and higher than 0.71 are considered excellent. In this first purification, 17 from the 40 items of the pilot version of the scale were excluded, once they did not reach 0.55 loading. Hence, the LCRP Scale is an instrument composed of 23 items, with 10 being excellent, 2 very good, and 11 good items. Scale reliability was calculated through Cronbach's alpha, with good indicators: 0.89 for Brand Experience; 0.84 for Customer Loyalty; and 0.74 for Brand Image. Results above 0.70 are considered reliable and above 0.80 very reliable (NUNNALLY; BERNSTEIN, 1994). In addition, the three first order factors explain 45.8% of construct variance, which is considered a good factor solution by Hair et al. (2009).

The next step concerns a new purification of the LCRP Scale, when the 23 items from the previous stage are submitted to confirmatory factor analysis.

In order to check the dimensions of the scale, the one-factor and the multifactor models were tested and compared following parsimony recommendations (BYRNE, 2009). The one-factor model showed worse rates of adjustment (NC = 6.18, CFI = 0.61, RMSEA = 0.13) when compared to the three-factor multifactorial model (NC = 2.31, CFI = 0.90, RMSEA = 0.06). The three-factor model was found to outperform the one-factor model on all measures. In addition, the chi-square difference was significant (Δχ² ₍₃₎ = 661.27, p<0.001), also indicating that the multifactorial model was indeed better than the unifactorial one. These results allow us to state that the scale is a good fit since all parameters meet the recommendations by Kline (2011).

According to Coltman et al. (2008), the direction of causality from construct to items is the first evidence of a reflective model, as showed in Figure 2, which illustrates the measurement model obtained in the confirmatory factor analysis and its parameters. Other evidence are the high positive intercorrelations between items, observed in terms of factor loadings and reliability; and convergent and discriminant validity.

Figure 2

Note. χ2(166)=382.72; p<0.001; NC=2.31; CFI=0.90; RMSEA=0.06

The relation between brand personality and customer relationship perception is illustrated by Figure 3.

Figure 3

Note. χ2(3)=73.78; p<0.001; NC=24.59; GFI=0.94; CFI=0.90; RMSEA=0.27