Relationship Marketing results : proposition of a cognitive mapping model

Resultados – We established an aggregate mental map that represents the RM structural model. Model analysis identified that CLV is understood as the final result of RM. We also observed that the impact of most of the RM elements on CLV is brokered by loyalty. Personalization and quality, on the other hand, proved to be process input elements, and are the ones that most strongly impact others. Finally, we highlight that elements that punish customers are much less effective than elements that benefit them.

In a scenario that is highly competitive and reveals rapid erosion of competitive advantages based on products, relationship marketing (RM) emerges as an alternative way of developing superior value by understanding customer needs and offering products that are suitable to each individual customer.RM is actually a set of practices that aim to guide companies towards an understanding of their customer bases (in the most individual way possible) and towards adjustment of supply, so as to retain current customers (Gordon, 1998;Vavra, 1993).
However, RM actions require large investments.It is necessary to adapt organizations to receive, store and disseminate customer information (Gordon, 1998;Vavra, 1993); to develop ways of adapting supply to customer needs (Gordon, 1998;McKenna, 1992); and, perhaps the most complex and profound effort of all, to change the focus and culture of companies, transforming them into companies that are focused on relationships (Vavra, 1993).These changes are only justified if companies can get back all the benefits that RM actions can bring.And, for this to be so, we need clear understanding of how these results can be achieved.
Therefore, it would be adequate to develop a model that reflects the various constructs involved in RM activities and their relationships with each other.A formal model of relationship marketing is useful to visualize the elements that make up RM, the interactions between them and the results of this activity.Only through deep understanding of how RM activities' actions' results are achieved can these actions be properly guided and, moreover, assessed and corrected.
There are several methodological possibilities for developing an RM model.Here, we chose to use a cognitive mapping method.The focus is, therefore, on identifying how stakeholders involved with the relationship activity understand its dynamics so as to, based on the cognitive models of various stakeholders, map a possible RM model.
Thus, the objectives of this research are: (a) to identify, through a systemic approach, the interactions between the elements that make up the main results of relationship marketing, especially CLV, according to the vision of the surveyed professionals, and (b) to propose an aggregate cognitive map that represents the overview of the surveyed individuals as to how CLV is formed through the interaction between relationship marketing elements.

RElAtIonshIP MARkEtIng
RM is a set of practices that aim to retain current customers through narrowing of the company-customer relationship.To this end, it is necessary to be able to identify each of the individual customers, to sustain with them a constant, two-way dialogue and, through the accumulated information provided, to adapt supplies to the needs expressed by customers (Gordon, 1998;Vavra, 1993).
However, in order to justify companies' engagement in RM, it is crucial that they perceive solid results.Therefore, a first step in the analysis of the RM activity is understanding what results it may bring.
Since the search for continuing negotiations with the same customers is a pillar of the RM activity, the importance of encouraging loyalty is clear (Edvardsson, Johnson, Gustafsson, & Strandvik, 2000;Gordon, 1998;Vavra, 1993;Zeithaml & Bitner, 2003).It is common to find in literature (Anderson & Mittal, 2000;Berger & Nasr, 1998;Berry, 1995;Edvardsson et al., 2000;Mulhern, 1999;Peppers & Rogers, 2005) evidence that increased customer loyalty brings about increased company profitability.According to the authors, increased profitability is due to the increased individual profitability of each customer over time.What is highlighted is that loyal customers would be more profitable to companies Relationship Marketing results: proposition of a cognitive mapping model than customers who are not loyal, due to their increased Customer Lifetime Value, or CLV.
Therefore, here we understand customer loyalty and the increase in their CLV as the main results of the RM activity.However, CLV is highlighted as a second-degree result, or as a later consequence of loyalty in the following sequence: RM elements → loyalty → CLV.Thus, we understand that there is in fact one great end RM result: CLV.This end result is in turn brokered by loyalty.
Having defined the main objective of RM actions, we must then identify which initial customer-company relationship assessment elements could, together, lead to its end results.As we said before, however, although CLV is understood as the end of the process, it is highlighted as a second-degree result, or as later consequence of loyalty.Therefore, in order to identify which initial elements of RM lead to its end results, we need to understand how loyalty is formed.
Different studies point to several possible relationships between the initial elements of RM and its ultimate goals, as can be seen in Figure 1.Multiple analyzes of the interactions between RM elements can be carried out from the study of existing models.In this paper, however, the focus is only on identifying the antecedents of loyalty.Analyzing all the found models, we are able to identify the elements that have a direct impact on loyalty.Table 1 shows these antecedents.In Table 1, we can observe that, among the elements with a direct impact on loyalty, the most mentioned are satisfaction, quality, commitment, trust, personalization and switching costs.These are, then, the initial RM elements considered potential paths towards loyalty.
To analyze the relationships between RM elements it is also necessary to have a clear definition of each construct.For this reason, based on extensive analysis of existing literature, Table 2 presents the definition of how each element in this study was understood.
We should note that certain activities referring to RM, especially Customer Relationship Management (CRM) activities, will not be discussed.This is because we strove to identify how the results of RM actions communicate with each other.Activities such as CRM work, the learning relationship and the segmentation of the customer base are actions that allow these RM results to occur, that is, they enable its operation and lead to the narrowing of the companycustomer relationship.The focus of this work is on the relationship of results and not in activities themselves.Therefore, constructs identified in Table 2, below, will be considered for this study.

Relationship Marketing results: proposition of a cognitive mapping model tABlE 2 -List of considered constructs subdivision definition source
Satisfaction -Satisfaction is the pleasant feeling that the package of consumption events concerning a given product or brand has met a certain goal or desire.Satisfaction occurs when a product provokes positive expectations in consumers, meets these expectations and receives positive affect.Satisfaction can occur both as to products themselves and to consumers' relationships with companies/employees.Oliver (1980); Surprenant and Solomon (1987); Vavra (1993); Fornell et al (1996); Garbarino and Johnson (1999); Oliver (1999); Szymanski and Henard (2001); Zeithaml & Bitner (2003) Commitment Affective Affective commitment occurs when consumers in fact like a given supplier, and thus resist abandoning it.

Calculative
Calculative commitment occurs when consumers feel that terminating their relationship with a given supplier may lead to economic or social sacrifices.

Procedural
Costs that may deter consumers from switching to a different supplier because of the time spent in doing so.Burnham, Frels and Mahajan (2003); Caruana (2004); Jones et al (2007)

Financial
Costs that may deter consumers from switching to a different supplier because of financial losses (such as loyalty program points, or initialization costs, such as enrollment or adaptation of specific assets).

Relationship
Costs that may deter consumers from switching to a different supplier because of the emotional losses inherent to doing so.

Trust -
Trust is the intention a party has to present himself as vulnerable to actions by another party.Trust occurs when one realizes that the other party possesses skills (competencies), benevolence (tendency to do good to the other party involved) and integrity (respects principles and values).

Personalization
Personalization is the supplying company's effort towards making products be as close as possible to the individual needs of each customer.Personalization can occur both in the product itself as well as in the contact/communication between customers and suppliers.Surprenant and Solomon (1987); Gordon (1998) Ball, Coelho and Vilares (2006); Vesanen (2007); Tong, Wong and Lui (2012)

Quality -
To what extent a product (both the product itself and the way it is carried out and delivered -the relationship between suppliers and customers) approaches the ideal and presents few errors.Gronroos (1995); Fornell et al (1996); Zeithaml and Bitner (2003); Baptista (2005)

loyalty -
Loyalty refers to brand preference, even in situations where there is pressure towards substitution.
The following are consequences of loyalty and, therefore, are ways to ascertain if it does in fact exist: i) repurchase; ii) indication; iii) preference; and iv) willingness to pay more.

FIEld REsEARCh MEthod
In order to achieve the outlined goals, we chose to develop a model obtained through cognitive mapping.Despite the different uses of the term "cognitive map", and the different approaches to its preparation and analysis, in a similar way to Markóczy and Goldberg (1995), Bastos (2002), Scavarda (2004) and Bouzdine-Chameeva (2007), in this research cognitive Iná Futino Barreto / Edson Crescitelli / Júlio César Bastos Figueiredo mapping is understood as the representation of the understanding of an individual concerning a researched subject.
Causal maps are particularly a subtype of cognitive map focused specifically on identifying constructs that have causal relationships with each other.Causal maps are widely used to capture the understanding of managers and researchers (Markóczy, 1994) and can be very useful for risk analysis, for developing strategy and construction research or theory testing (Markóczy & Goldberg, 1995;Scavarda et al, 2004).

stages of the causal mapping process
The first step in the development of a causal map is identification of constructs referring to the subject to be studied; in the case of this study, RM constructs and their end results.It is important to ensure that all constructs are clearly defined, so that they may be understood in the same way by everyone.The second step is presentation of the constructs, listed for the experts who are to be interviewed and will point out the relationships between them (Bouzdine-Chameeva, 2005; Clarkson & Hodgkinson, 2005;Markóczy & Goldberg, 1995;Scavarda, Bouzdine-Chameeva, Goldstein, Hays, & Hill, 2005).Next, the individual maps of respondents are developed (Bouzdine-Chameeva, 2005;Clarkson & Hodgkinson, 2005;Markóczy & Goldberg, 1995); Finally, we present a consolidated map of all respondents (Bouzdine-Chameeva, 2005;Scavarda et al., 2005).The sequence of stages is presented in Table 3.
In this study we followed the proposal of Markóczy (1994): using existing theory analysis, we developed a pre-defined list of constructs, which was then presented to respondents along with a brief explanation about each construct (as observed in Table 2).Using a pre-defined list, according to the author, allows all respondents to work with the same universe of constructs and, more importantly, to understand all the indicated constructs in the same way.
The list was developed in two stages.First, we created an initial list according to existing theory analysis (Table 2).Then, that list was presented to two experts in the field for content validation (namely, to a scholar with articles published in magazines classified as A1 or A2 by CAPES' Qualis system on the subject RM; and a market expert, a consultant involved in RM, whose agency was awarded the best agency of the year in CRM/DBM: a prize awarded by the Brazilian Association of Direct Marketing/ ABEMD).We asked the experts to verify whether they felt any constructs were missing and should be added to the list, if they correctly understood the meaning of all listed constructs and if they believed that any of the presented constructs should be joined to another.Thus, we validated the list with 11 constructs presented in Table 2, which was then used in the interviews.
Once the list of constructs had been defined, the next step of the process was to choose the respondents.As pointed out by Groesser and Bruppacher (2007), this choice does not intend to build a representative result, but should be guided by pursuit of the greatest possible variability.For this reason, we chose to focus on three categories of stakeholders: consultants/agencies focused on RM, managers and scholars involved in RM.
Relationship Marketing results: proposition of a cognitive mapping model However, there was no control to ensure the same amount of responders in each group.
To select the companies (and individuals) who would be targeted by the research, we considered, at first, RM service providers who are most remembered by Brazilian companies, according to research published by the Portal Do Mundo do Marketing (2013).They are as follows: Market Data; The Group; Accentiv'Mimética; Expertise; Casa Nova; Omnion; Rapp Digital; CSU; Total On Demand; E/OU.We contacted the directors, partners and/or presidents of each of these companies.Of these, six agreed to participate.From then on, we asked each respondent to suggest other names who were intensely focused on RM (Snowball technique).At the end, we had contacted 82 potential respondents, of which 35 actually participated in the study, as shown in Table 4 (in the Table, information that, at the request of respondents, had to be omitted, was replaced by XX).
Steps 3 and 4 (Identification of connecting arcs between constructs and Creation of individual maps) were carried out simultaneously.Since this research works with prior establishment of a list of constructs, listing the arcs was done using that same list.Despite the many proposed procedures that are present in theory, this study approaches what was presented by Markóczy (1994), Markóczy and Goldberg (1995) and Bouzdine-Chameeva et al (2001).To define which constructs (nodes) will be in the causal map, the authors suggest that respondents are requested to select, among the listed constructs, those they deem most important to the topic that is being studied.Markóczy (1994) and Markóczy and Goldberg (1995) suggest that respondents select the 10 most relevant constructs.Bouzdine-Chameeva et al (2001) limit the choice of respondents to a total 20 to 40 constructs.Crescitelli and Figueiredo (2010), who conducted a study based on the methodology proposed by Markóczy and Goldberg (1995), also work with a total 10 constructs selected by respondents.
In this project, however, since there are only 11 listed constructs, all were compulsorily used by the respondents.Since 11 constructs is a value that is very close to the lowest number of constructs used by the authors we consulted, the risk of the field procedure becoming excessively dull and time consuming is very small, and, at the same time, there is no forced exclusion of relevant constructs.If there were constructs that respondents judged as not having any impact on others, all they had to do was not point out any links to them.Thus, all respondents were given a list of constructs (as shown in Table 2); next, the interviewer presented and discussed each element with the respondent.After ensuring that there were no doubts, respondents were asked to evaluate the direct influence of construct 1 on construct 2, in order to identify the relationship links between the constructs.If the respondent indicated that there was an influence, he was asked about the strength of this influence in a scale of -5 (strong negative influence) to +5 (strong positive influence).To present the ordered pairs, we developed supporting software which presented the constructs under analysis randomly -software implemented in C from algorithms developed by Goldberg (1996).
The fifth and final step of the cognitive mapping process used in this paper is the development of an average map, to indicate the aggregate vision of all respondents, together.One issue to be defined, however, is how to identify the links between the nodes in the average map.To work around this point, some possibilities can be considered.The simplest possibility is to indicate in each link the average of the links mentioned by each of the respondents.In this case, however, virtually all possible links would be shown in the average map, since almost all of them were mentioned by at least one of the respondents, except for the link between the constructs "switching costs procedural" and "quality", which presented average zero.If we had chosen Relationship Marketing results: proposition of a cognitive mapping model this type of aggregation, the average map would have to have, in all, 109 links, which would make it confusing and inefficient for understanding the main relationships.Another possibility is to consider only the links listed by at least a given number of respondents, for example, those listed by at least half of the respondents.In this case, however, definition of what is the limit to be considered (10%, 40% or 50% of respondents) would be arbitrary.
There are also prominent points regarding nodes and links.Accordingly, first of all, we observed that CLV is, in fact, a final element of the RM activity.In the model, we see that CLV is essentially a system output.We confirmed that, as was pointed out in the literature review, CLV is the end result of RM work and should be regarded, therefore, as its primary goal.
As with CLV, relationship switching costs appears as a system output.However, unlike CLV, relationship switching costs is not a goal to be pursued by companies and thus is not a result to aim for.
Loyalty, on the other hand, is the node that most receives impacts from the others (it is impacted by five other constructs: satisfaction, affective commitment, trust, personalization and quality).Also, the strongest link in the entire model is the influence between fidelity and CLV, worth 4.6.Importantly, in addition to being the strongest link, it is also the link that withstood the most rigorous test of significance.When testing the links that, with a significance level of 0.01, had a modulus value of at least 4, the link between loyalty and CLV is the only one that reveals itself as significant.It is, therefore, a link that is considered consistently strong by the interviewed individuals.
Relationship Marketing results: proposition of a cognitive mapping model

FIguRE 2 -Average map diagram
This confirms what was pointed out by the meta-analysis carried out based on many different models in literature, that is, most of the RM elements have an impact on the CLV that is mediated by loyalty.Five elements (except for loyalty) have a direct or indirect and extremely intense impact on CLV, and all are in some way mediated by loyalty.Of these five, three have an impact on CLV only through loyalty (satisfaction, affective commitment and personalization); one, quality, has an impact on CLV mediated by loyalty and trust; and the fifth one, trust, has a direct impact on it and is mediated by loyalty.Therefore, it is quite reasonable to conjecture that the RM process should be understood by the sequence: RM elements → loyalty → CLV indicated in this paper.
At another extreme, early on in the process, we identified that personalization and quality are exogenous variables, that is, they do not receive impacts from any other elements.This result is important, since these are the two variables that actually face the decision and control of the company.All the other elements (trust, satisfaction, loyalty, CLV, etc.) must be earned by the company.Quality and personalization are, therefore, the starting point in the narrowing of the relationship between a company and a customer.Quality and personalization are also the nodes that have a strong influence on the widest array of other nodes.With specific regard to personalization, it is interesting to note this result, since, according to Ball, Coelho and Vilares (2006), the effects of personalization have not much researched.Similarly, in the researched literature after 2006, only a few formal models actually included personalization; we only highlight Tong, Wong and Lui (2012).
Iná Futino Barreto / Edson Crescitelli / Júlio César Bastos Figueiredo As for the less relevant nodes, the generated model shows that calculative commitment and switching costs (the three types) do not have a strong impact on any other construct.In fact, calculative commitment and procedural and financial switching costs do not impact and are not significantly affected by any other node, and were therefore excluded from the final model.This shows that, according to what we have observed, these elements are less important and should be less focused on in the search for the final results of the company-customer relationship.
Interestingly, positive elements (such as personalization, quality, satisfaction) have far greater impact than elements of penalty, such as switching costs and calculative commitment.This indicates that, according to the surveyed individuals, customer benefits tend to have better results than penalties.This does not mean, of course, that penalty elements have no impact on the CLV, but that, in an effort to prioritize and allocate scarce resources, the impacts of these elements will be smaller.
As well as identifying how the final elements of relationship marketing are formed, average map analysis also allows us to observe relationships between the mean elements.Accordingly, we can observe that: Satisfaction is strongly formed by quality, as widely noted and discussed (Baptista, 2005;Fornell, Johnson, Anderson, Cha, & Bryant, 1996;Kristensen & Eskilden, 2012;Ranjbarian, Sanayei, Kaboli, & Hadadian, 2012;Szymanski & Henard, 2001), and by pesonalization.Also as to satisfaction, this study validated the direct impact of satisfaction on loyalty, as pointed out by Fornell et al (1996) -Figure 1C  As pointed out in the theory (Jones et al, 2007;Fullerton, 2005), affective commitment was also shown here as an antecedent of loyalty (impact 3.7).As for the shapers, affective commitment is strongly formed by the junction of personalization, quality, satisfaction and trust.
Trust, on the other hand, appears as a direct shaper of loyalty, as indicated by Gabarino and Johnson (1999) and Ball, Coleho and Villares (2006) -Figure 1M, but is rejected by Baptista (2005) -Figure 1J and Brei and Rossi (2005) -Figure 1K.We also confirmed the commitmenttrust theory of Morgan and Hunt (1994), in which trust has an influence on commitment.We highlight the direct impact of trust on the CLV, which was not shown by researched theory.As for its shapers, in the model developed here, the only element with a strong impact on confidence is quality, as pointed out only by the Baptista model ( 2005) -Figure 1J.
Besides the main objective of analyzing the relationship between RM elements, it could be interesting to analyze whether there is some sort of difference in the thoughts of individuals with different characteristics.In order to explore this possibility, we carried out cluster analysis of individual maps of all 35 respondents.However, no formation of groups with different patterns of behavior was observed.In fact, after the formation of two small groups (the first comprised by eight respondents, and the second of three respondents), which then come together, we basically added one respondent at a time to one same larger group.Thus, there is no evidence, in this research, to suppose that the thoughts of scholars, managers and consultants is different from each other's; or that individuals with greater experience in the B2B or B2C markets make up groups that have different thoughts.
Relationship Marketing results: proposition of a cognitive mapping model loyalty and CLV; and CLV as the ultimate goal of relationship actions.Each one of these elements was analyzed in the light of theory.In addition, we listed existing and tested theoretical models to indicate the relationship between the surveyed elements.
Based on this survey, a final list of RM constructs was developed.In order to identify the relationships between these elements, we employed cognitive mapping, to find out how Brazilian professionals who are highly involved in relationship management with customers see the interactions between the formed factors of loyalty and, consequently, the ultimate goal: CLV.Besides individual maps expressing how each of the surveyed experts understand the relationships between the constructs, an average map was drawn.This average map shows the average pooled overview of all respondents and it is through it that we carried out our analysis.
As for the main findings, first, we emphasize that CLV, considered throughout the research as the ultimate goal of the RM activity, was confirmed as an output variable in the model.This means that the method adopted in this study validated the hypothesis that conceptual CLV should be understood as the final result of the RM activity.
This finding is important, since it is in line with the central concept that RM activity cannot exist if it does not bring higher returns.Furthermore, it allowed the inclusion of CLV in the structural models of the company-customer relationship.The inclusion of CLV in the structural model developed in this research is relevant because, although it is a central concept, CLV is not part of the structural models identified in the developed theoretical review.
The analyzes carried out allowed us to understand the interactions between the various RM elements and, in particular, how its final element is formed: CLV.Instead of analyzing formulas to estimate the individual CLV of each client, such as most CLV models research do, this study investigated the relationship between elements that form it.The identification of the aggregate map and of the elements with higher and lower importance in the formation of CLV can be useful to increase awareness on the issue and assist managers to better invest their efforts and optimize results.We were able to identify, for example, where better to invest to maximize results on CLV, or performance indicators of intermediate elements that can predict future variations of CLV in the customer base.
While this study was able to point out the results or output elements referring to RM, it was also able to identify key input elements.Personalization and quality appear as exogenous elements in the designed model.That is, even considering low intensity impacts, personalization and quality are much more strongly causative constructs than consequences, and at the same time have an important impact on the other process elements.These two elements should be, therefore, the starting points of relationship efforts with customers.Some limitations, however, need to be identified.The first refers to the type of research and to the research universe.Through the method we used, we only identified how the field experts see the interrelationships between constructs that are relevant to RM.We could not, however, assess whether consumers in fact behave the way the experts point out.To this end, it would be necessary to list the opinions of consumers.This type of verification would be quite rich to validate the model that was developed here and is highlighted as a suggestion for future studies.
The second limitation still refers to data collection.As pointed out by Markóczy (1994) and Ensslin and Montibeller (1999) Iná Futino Barreto / Edson Crescitelli / Júlio César Bastos Figueiredo Finally, the third limitation we consider important to emphasize refers to the variables analyzed.Only the constructs regarding the RM activity were analyzed, and that could somehow influence their results.There are, however, a number of other exogenous factors that can enhance the impact of these elements.As examples, we can mention the exogenous variables that we did not deal with: customer engagement with the purchase, number of previous purchases (Johnson & Sivadas;Gabarino, 2008) and the degree of market competitiveness (Lee & Lee and Feick, 2001).

Figure 1 -
Figure 1 -Main models surveyed on the subject of loyalty

Figure 2 -FIguRE 1 B
Figure 2 -Main models surveyed on the subject of loyalty
, causal maps cannot claim to be a perfect representation of reality, nor even a perfect representation of the mental model of respondents.The maps identify what a given respondent was able to access in a research meeting.It is possible that another interviewer, researching the same subject at another time, could establish a different causal map.Causal maps must bring the representation, as close as possible, of what the respondent believes and can express at the time of research.

tABlE 4 -
List of respondents

tABlE 5 -
Average of all links with at least 0,05 significance in relationship averages

tABlE 6 -
Average map association matrix