Parvatiyar and Sheth (2000) stated that CRM and relationship marketing are terms that are frequently used in an indistinctive manner in the marketing literature. Actually, they are highly correlated themes, since CRM can be seen as the systematic and automatized application of relationship marketing concepts. However, the origin of such concepts has occurred in different times, since relationship marketing has appeared before CRM. According to Gebert et al. (2003), CRM was derived from the combination of a new approach to manage customers with an information systems (IS) orientation, particularly with technologies for transaction management and sales forces automation.

CRM is a customer oriented model where an individual sale is just the beginning of a process, opposing to the product oriented view, where the sale is the end of the process (RYGIELSKI; WANG; YEN, 2002; LIN; SU; CHIEN, 2006). In the CRM perspective, organizations seek to maximize customers’ value by analyzing their lifecycle and exploiting business analytics and relationship management technologies (RYGIELSKI et al., 2002; SEDIGHI et al., 2012). The intensive usage of customer information is quintessential to the correct and efficient relationship management (BRETZKE, 2001), emphasizing the strategic contribution of information and knowledge management to CRM initiatives.

In our study, five different classifications were found to typify CRM: operational CRM, analytical CRM, collaborative CRM, strategic CRM and electronic CRM (Table 1).

Table 1

References	CRM Types
	Operational	Analytical	Collaborative	Strategic	Electronic
Schwede (2000)	X	X	X
Peppers and Rogers (2001)	X	X	X
Bose and Sugumaram (2003)	X	X
Henning et al. (2003)	X	X	X
Xu and Walton (2005)	X	X	X		X
Geib et al. (2006)	X	X	X
Iriana and Buttle (2007)	X	X		X
Richards and Jones (2008)	X			X
Ngai et al. (2009)	X	X
Khodakarami and Chan (2014)	X	X	X

Source: Elaborated by the authors.

Majorly, CRM types are grouped in three categories which will be further considered in our CKM model development:

- Operational CRM: its main function is to automatize the CRM process, increasing efficiency and productivity. Customer care and support units, call centers, sales force systems and business rules automation belong to this category;

Customer knowledge is a constituent part of CRM, despite the scarcity of this discussion in the Information System literature (KHODAKARAMI; CHAN, 2014). Table 2 summarizes how some scholars have classified customer knowledge in three categories, depending on the origin and the purpose of the managed knowledge.

Table 2

Knowledge Category	Source	Purpose	Example
About the Customer	Customer transactions and customer database	Used to understand customer motivation towards a personalized relationship	Demographical and register data, payment history, characteristics of the purchased goods and services
From the Customer	Customer relationship	Used to provide enhancements in products, services and processes	Claims, praises, suggestions, opinions about products and services and perceptions about competitors
To the Customer	Firm	Used to comply with customer request for knowledge related to products and services supplied by the firm	Information about products and services, alerts about deals and payment dates

Source: Developed by the authors based upon Garcia-Murillo and Annabi (2002); Park and Kim (2003); Smith and Mckeen (2005); Geib et al. (2006).

Indeed, CRM processes are knowledge oriented processes, because the three types of knowledge (about, from and to the customer) are part of a firm’s intellectual capital (GEBERT et al., 2003). The usage of those types of customer knowledge in the knowledge creation process, and the adoption of KM tools to support it, are discussed in the next section.

There are many definitions of KM because different perspectives of what knowledge is may bring different perceptions about KM (ALAVI; LEIDNER, 2001). For the purposes of this work, KM is understood as a set of methods to generate, organize and share knowledge (DAVENPORT; PRUSAK, 1998). Knowledge creation reflects the firm’s capacity to generate, disseminate and internalize new knowledge (NONAKA; TAKEUCHI; 1997). The epistemological dimension distinguishes tacit from explicit knowledge. Tacit knowledge is usually personal, corporal, hard to transfer and embedded into individual cognitive connections. On the other hand, explicit knowledge can be transferred by formal and systematized language, assuming material forms such as in documents, databases and images (NONAKA; TAKEUCHI, 1997).

Knowledge creation occurs by the dynamic interaction of tacit and explicit knowledge represented in a knowledge spiral (NONAKA; TAKEUCHI; 1997). Four types of knowledge conversion emerge in this cycle, which can be exemplified in the CRM context as follows:

- Socialization: comprises the sharing of personal experiences, such as interactions among call center employees, salespersons and customers. Empowered by CRM, these professionals possess knowledge about the product portfolio and customer profiles, offering a more customized attendance and experience. Furthermore, part of the tacit knowledge derived from interacting with customers (e.g. claims, quality issues that differentiate products and services) may also contribute to the customer knowledge base;

Carvalho (2001) classified KM software into eight categories according to their technical features and usage. For the purpose of this paper, the following five suitable tools to customer KM were considered: (i) intranets; (ii) groupware; (iii) workflow; (iv) CBR (Case-Based Reasoning) systems; and (v) Business Intelligence (BI).

Knowing customers is indispensable for a customized relationship, which is a CRM basic premise. The application of KM tools and techniques is essential to capture, store, analyze and allow an organization and its customers to exchange knowledge (LOPEZ-NICOLAS; MOLINA-CASTILLO, 2008), deeply correlating CRM and KM.

Aware of such possibilities, organizations tend to pursue the integration of CRM and KM processes. KM plays a substantial role in a successful CRM strategy (SALOMANN; DOUS; KOLBE; BRENNER. 2005), since the CRM process is complex and deals with semi-structured or even unstructured knowledge.

The alignment of CRM and KM results in a hybrid perspective (CKM), which is centered on the management of knowledge about, from and to the customers. Salomann et al. (2005) define CKM as the systematization and management of knowledge obtained from the diverse points of interaction with customers. CKM may also be understood as the result of CRM and KM integration because CKM’s role is to manage the gathering, codification and dissemination of relevant customer knowledge (KOLBE; GEIB, 2005) in order to assure an effective customer management, promoting enduring customer relationships (LIN et al., 2006). Indeed, the application of CKM results in a win-win relationship. By better knowing their customers, organizations may promote lasting relationships. On the other hand, customers benefits from personalized interactions, particular deals and customized products and services.

The research’s unit of analysis was a Brazilian private high education group that holds several institutions and brands providing face-to-face and distance learning services. The institution is present in all states (twenty seven) of Brazil and is currently the leader in the Brazilian higher education market, with more than seven hundred thousand (700,000) students attending more than 70 programs in 450 distance learning facilities and 50 face-to-face learning campi (data from 2014). The group is an open market organization and its stocks are traded at the São Paulo stock exchange (BM&F BOVESPA) since 2007 (BOVESPA, 2014).

The educational group is structured with nine vice presidencies (VP), one of which is in charge of marketing and sales. The customer retention and loyalty area, with is responsible for the CRM and also the DBM (database marketing) activities responds to this vice presidency. This area designs the student marketing campaigns in order to increase reenrollment every semester. This team is also responsible for the deployment and management of CKM initiatives. One of the authors of this paper was the leader of that team at the time the theoretical model was built, which provided us with a good practical situation to develop the field experiment.

The CKM model (Fig. 1) was deployed at the first time in the educational institution from December 2013 to September 2014. The model guided the control of all student reenrollment campaigns and the actions addressed to ex-students (term used to describe inactive students who evaded during previous semesters). The following steps were taken to deploy the CKM model.

The first step (Analytical CRM phase) was the mapping and integration of diverse data sources. This step occurred from December 2013 to February 2014 and involved the mapping of financial, demographic and registration data from students. Most of the data sources were dispersed across many departments in “information islands”. It was possible to identify excessive and sometimes unappropriated marketing approaches to students, resulting in dissatisfaction, due to redundant contacts. Data were captured, organized and integrated into a single data repository, called student retention database. Business Intelligence (BI) back-end tools were used to support the ETL (extraction, transformation and load) process.

Still in the Analytical CRM phase, a data mining process was conducted from February to May 2014 to discover relevant knowledge about the students. Based on the analysis of the retention database, it was possible to identify patterns concerning the academic and financial engagement of students, i.e., measurements of how dedicated the students were to their academic life and the financial responsibilities associated to the continuity of their studies. Such information was made available through BI analytical portals to the retention and loyalty area, in charge of the marketing actions of the educational institution. The BI analytical portals were used to support the combination of the knowledge generated by data mining and the knowledge internalization by the marketing team.

The marketing strategy was based on the segmentation of offerings, channels, frequency and content of the messages directed to students and ex-students according to their profile. Despite the scarcity of studies concerning the development and usage of segmentation in higher education marketing, segmentation has already proven to be a fundamental tool for marketing strategy, especially in the service sector (JIMÉNEZ-CASTILLO; SÁNCHEZ-FERNANDEZ; INIESTA-BONILLO, 2013). The segmentation based upon the learning modality is justified by the differences of the academic model, business model and student profile between face-to-face program, distance program 1 and distance program 2, which are different brands of the higher education group. Concerning the student type, segmentation is also necessary since different target publics will receive specific offers. Moreover, for each segment, the educational group uses specific managerial indicators.

The next step (Operational CRM phase) comprised the construction of student relationship rules, defining workflows to organize student relationship actions in a sequential and conditional manner. This approach is sequential because it distributes relationship actions across a predetermined period of time, detailing the clock for marketing actions. It is also conditional, because it has decision points that are influenced by the financial and academic engagement of the student and the student’s response to each previous interaction.

Next, marketing actions were performed involving the students and ex-students through diverse communication channels, selected by the previously mentioned workflows. Call center and face-to-face interactions were used in these actions, supported by CBR (which are part of Operational CRM), and also web portals, social networks and direct marketing (SMS, e-mail, direct mail), which are part of Collaborative CRM.

The interactions happened in a consecutive manner according to the student profile and his/her responses to each interaction. There were variations in the content, in the frequency and in the communication channels that were used to interact with students. All interactions were registered and their results (knowledge from students) were included again in the student retention database, so they could, once more, be analyzed to determine the next type of interaction with the student.

Marketing actions were performed according to the type of the program (face-to-face or distance learning), the student’s status (active or not) and the phase during the reenrollment period, resulting in approximately 130 different marketing actions. The content and style of the messages of the marketing action observed the student profile which compass academic, financial, demographic and enrollment information and also the answer patterns for each marketing interaction.

The number of active students and ex-students exposed to the marketing campaigns was about six hundred thousand individuals. Control groups were randomly extracted from this universe to measure the effectiveness of marketing actions, disregarding students who reenrolled by themselves, without the influence of marketing campaigns. Control groups represented seventeen thousand students, which were not exposed to any type of message from the institution, letting the reenrollment happen in an absolutely spontaneously way. In the quantitative analysis of the proposed CKM model, it was expected that the students of the control group (spontaneous reenrollment) would present a lower reenrollment rate than the students impacted by the CKM based marketing campaigns. The difference between the two rates was considered the business result of the marketing actions.

The reenrollment indexes for the second term of 2014 were used for the quantitative analysis of the CKM model. The reenrollment process was called reenrollment campaign. It took place from June to September 2014 and comprised the reenrollment in face-to-face and distance learning programs. The index of ex-students, who had evaded in previous terms and came back now, was also measured in a campaign that was called “win back”.

A contrast scenario to verify the effectiveness of the marketing campaigns was created for the following two groups: case group (students exposed to marketing campaigns designed by the CKM model) and control group (students not exposed to marketing campaigns). A chi-square test (χ²) was used to verify the dependence of the dependent variable on the independent one, as proposed in Soares and Siqueira (2002). The purpose of the test was to check if the rate of enrollments of ex-students (win back) and reenrollment of current students was influenced or not by the marketing actions designed based on the CKM model. The hypotheses verified by the test were if the rate of enrollments of ex-students (win back) and reenrollment of current students is positively influenced by the marketing actions designed based on the CKM model or not.

Next, the test results will be presented separately for different programs (face-to-face program, distance learning brand 1 and distance learning brand 2) and depending on the student profile (current student exposed to reenrollment campaign or ex-student exposed to “win back” campaign), resulting in six combinations (Tables 4, 5 and 6). This separation is necessary not only because of the fact that the program could affect the effectiveness of the campaign (distance learning students are somewhat different to face-to-face students and the business models used for each of those businesses are also distinct), but also because of the differences between current students and ex-students.

Table 4

	Face-to-face - Ex-Students			Face-to-face-  Current Students
	Potential Public	Number of Enrollments	% Enrollment	Potential Public	Number of Reenrollments	% Reenrollment
Case group	54,206	8,388	15.47%	179,152	148,226	82.74%
Control group	2,235	373	16.69%	1,844	1,279	69.36%

Source: research data.

Table 5

	DP 1 - Ex-Students			DP 1 - Current Students
	Potential Public	Number of Enrollments	% Enrollment	Potential Public	Number of Reenrollments	% Reenrollment
Case group	83,781	7,043	8.41%	303,781	257,958	84.92%
Control group	2,936	242	8.24%	6,000	4,671	77.85%

Source: research data.

Table 6

	DP 2 - Ex-Students			DP 2 - Current Students
	Potential Public	Number of Enrollments	% Enrollment	Potential Public	Number of Reenrollments	% Reenrollment
Case group	46,696	3,788	8.11%	75,954	57,222	75.34%
Control group	1,923	151	7.85%	2,156	1,269	58.86%

Source: research data.

The marketing campaigns were not significant to the enrollment (win back) of ex-students (Tables 4, 5 and 6). This result was further confirmed by the chi-square dependency verification (Table 7). For p-values greater than 0.05, is not rejected (Table 7), i.e., the verified rates for the case group and the control group are not statistically different. On the other hand, the marketing campaigns have made significant differences on the reenrollment indexes of current students. For this group, p-values were lower than 0.05, resulting in the rejection of , i.e., showing that the marketing campaign was important to keep students loyal.

Table 7

Type of Learning	Student Type	P-value
Face-to-face	Ex-Students	0.19
	Current Students	0.00*
DP 1	Ex-Students	0.77
	Current Students	0.00*
DP 2	Ex-Students	0.71
	Current Students	0.00*

Note. The symbol * indicates significant differences between proportions.

The marketing campaign directed to ex-students has not obtained the same success of the campaign targeted to current students, reinforcing the importance of CKM for better understanding of relationship patterns. For the current students, it was possible to monitor the academic routine, the monthly payments and the satisfaction with the services and use updated demographic and register data. In the counterpart, for ex-students, the campaign design were based only demographic and register data. The richness and variety of data concerning the current students have leveraged marketing results, since the campaigns have exploited the existing customer knowledge.

It was possible to measure the return on investment (ROI) of the marketing campaigns, since there was an increase in reenrollment among current students. Next, the calculation of the marketing investments and the financial return is explained.

The following items were considered for the estimation of marketing investments: (i) payroll costs (taxes included) of the team involved in developing the marketing campaign (application of the CKM model); (ii) call center cost during the campaign; (iii) personal attendance cost; (iv) direct marketing costs (e-mails and cell phone messages). The financial return of the marketing campaign was calculated based on the increase of reenrollments. The increase was obtained from the difference between the total of reenrollments (case group) and the total of students that would be expected to reenroll if the same reenrollment rate that happened for the control group happened also for the remaining of the potential public (Tables 8, 9 and 10).

Table 8

	Face-to-face Education Program - Current Students
	Case Group	Control Group	Increase
Potential public	179,152	179,152	179,152
% Reenrollment	82.74%	69.36%	13.38%
Number of reenrollments	148,226	124,260*	23,966

Note. The symbol * in this table and in the following ones denotes the number of students estimated to reenroll if no marketing campaign had occurred (based on what happened with the control group).

Table 9

	Distance Learning Program 2- Current Students
	Case Group	Control Group	Increase
Potential public	75.954	75.954	75.954
% Reenrollment	75.34%	58.86%	16.48%
Number of reenrollments	57,222	44,706*	12,516

Source: research data.

Table 10

	Distance Learning Program 2 - Current Students
	Case Group	Control Group	Increase
Potential public	75.954	75.954	75.954
% Reenrollment	75.34%	58.86%	16.48%
Number of reenrollments	57,222	44,706*	12,516

Source: research data.

After calculating the increase of students for each program, it was possible to multiply this number by the respective average monthly fee paid by students of each program (face-to-face, distance learning programs 1 and 2) for a six month period (six fees for one term). That product was then divided by the marketing investment. There were no additional Information Technology (IT) investments for the CKM initiative, since the technological infrastructure was already available because the institution uses it for other business processes. Due to the secrecy agreement for this research, the absolute values of the marketing investments and the financial return cannot be presented here. However, the following ratio represents the calculated ROI for the marketing actions.

ROI = Financial Return MKT Investment , where : Return Financial = Icrease in Student sEnrollment ′ × Average Monthly Fee × 6 Marketing Investments = ∑ Marketing cos ts

Even considering that the simplified calculation does not take into account the cost of delivering the educational services (which means that the actual ROI could be a little less), the calculated return is amazingly high: for each US$ 1 invested in marketing, the expressive ROI of US$ 148 was obtained.