A Framework for Customer Relationship Management and Data Mining - Literature review Example

A framework for CRM and data mining Table of Contents 1.0 Literature Review …………………………………………………………. 3 1.1 Background …………………………………………..……………………. 3 1.2 Literature Review Foundations ………………………………………...….. 4 1.3 A critical review concerning the contribution of data mining to CRM ……. 4 1.4 Integration of data mining and CRM and an assessment of its impact ….… 6 1.5 Impact of Data Mining on CRM and Current Practices …………………… 7 1.5.1 Association ……………………………………………………………….. 7 1.5.2 Classification ………………………………………………………..…… 8 1.5.3 Clustering ………………………………………………………..……….. 9 1.5.4 Forecasting ………………………………………………………….….. 10 1.5.5 Regression ………………………………………………………………. 10 1.5.6 Sequence discovery ………………………………………………..…… 11 1.5.7 Visualization ……………………………………………………….…… 11 1.6 Recommendations represented by a framework for successfully integrating data mining into CRM …………………..……………………. 12 References …………………………………………………………..………… 14 Figure 1 – Seven Factors of Data Mining CRM Extraction and Analysis ……... 6 1.0 Literature Review 1.1 Introduction In looking into the contribution of data mining to customer relationship management (CRM), the starting point of the evaluation centered upon understanding the two terms as a basis for making an assessment. Williams (2009, p. 12) tells us CRM is “… an information industry term for methodologies as well as software, and usually Internet capabilities that help an enterprise manage customer relationships in an organised way”. To provide a full understanding, Schmitt’s (2008, p. 18) definition states CRM is “… a management philosophy according to which a company’s goals can be best achieved through identification and satisfaction of the customers' stated and unstated needs and wants”. Regardless of which definition one elects to adopt, the core principle is that CRM represents a tool to make more profitable and defined use of information gathered by companies on customers that can be used in marketing, customer service and other uses to improve and or increase maximization of customer data (Yu, 2001).   As a business function, CRM is an integral part of managing and maximizing customer information geared to more profitable operations. Data mining represents “the process of discovering new patterns from large data sets involving methods at the intersection of artificial intelligence, machine learning, statistics and database systems” (Fayyad et al, 1996, p. 3). As the above indicates, the purpose of data mining is to scour CRM databases for data sets, patterns and other programmed variables to be utilised for varied end purposes (marketing, sales, customer service, etc.) based on management objectives (Zhang et al, 2003). This examination sought to investigate a framework for CRM and data mining, thus the efficiency and effectiveness of both functions begin with the data collection completeness under Customer Relationship Management. This is due to the fact data mining is a closed system function that is dependent on the information it has to work with as a data series (Ngai et al, 2009). As a result of this foundation, the above provides insight as to what was uncovered and focused upon in this study, the completeness of the Customer Relationship Management data gathering resources. It also includes the analytical approach used in setting the parameters for the data mining function that determines the accuracy and value of this function. As a result, there is a mutual dependency where data mining’s contribution is a direct function of the information available to it under CRM. 1.2 Literature Review Foundations As a means to analyze and critically review the contribution of data mining to Customer Relationship Management, a number of factors were equated to formulate the research objectives. The analysis of data mining and its contribution to as well as a working relationship with Customer Relationship Management is the foundation of this review of literature, each segment (CRM and data mining) will be explored. This means looking into them from the standpoint of theoretical constructs as well as practical application. These will be formulated under five main Research Objectives as represented by:   1 A critical review concerning the contribution of data mining to Customer Relationship Management. 2 An identification of the critical success factors represented by the integration of data mining and CRM as found in relevant literature. 3 An assessment of the impact of the preceding areas on and in the success of Customer Relationship Management. 4 A look at and identification of current CRM and data mining practice. 5 Recommendations as represented by a framework form for successfully integrating data mining into Customer Relationship Management.   1.3 A critical review concerning the contribution of data mining to CRM. Customer Relationship Management is, in essence, a data capture tool that in and of itself has no value unless an organization has taken the time, effort and analysis to understand what it can do concerning increasing its completeness with respect to data collection areas (Berson and Smith, 2002). Company management, along with its marketing, sales, forecasting and other related divisions need to understand the data needed to be gathered. This includes the collection points for such information as a means to make the Customer Relationship Management database as complete and relevant as possible (Berson and Smith, 2002). Winer (2001) concurs with this view and states in order to reap benefits, all points of customer contact need to gather information and correlate it back to the individual database records to build on its value content as well as worth for future use. He adds that correlation methods can use different measures to tie data together as represented by telephone, credit cards, addresses, zip codes, and other data sets (Winer, 2001).   Without the needed, relevant and encompassing database collection and correlation measures, no amount of data mining will be useful in yielding effective, timely or useful information as the foundation would not be there. Data mining is the means by which information, discoveries, and useful data are pulled from a Customer Relationship Management database (Cabena et al, 2007). Rygielski et al (2002) make the assertion that CRM and data mining, in terms of their use as predictive sources concerning customer behaviour, is limited by its lack of widespread usage. There is difficulty in agreeing with this view as the usefulness of an application, technique, statistical tool or data method is not dependent on wide scale usage. Instead, it relies on the comprehensiveness of its users in understanding the data needed to be collected and how to access (mine) it (Xu et al, 2002). This view is further supported by (Ngai et al, 2009) who state data mining is a set of processes as well as enabling systems geared to support the business strategy of an enterprise. They add it is constructed with the understanding it is long term and entails building a profitable set of relationships with customer as well as the business (Ngai et al, 2009).   As the above reveals, the effectiveness and efficiency of data mining have to do with the thought, planning, and extent of the database it has access to (Rygielski et al, 2002). In addition, it is pointed out the skill set of the people doing the analysis, along with their familiarity with varied statistical tools, systems, and other disciplines is highly important in extracting data that has value and usefulness (Rygielski et al, 2002).       1.4 Integration of data mining and CRM and an assessment of its impact The end of the prior section left off with the importance and significance of the skill set of the individuals conducting the data mining process as being key to obtaining valuable and useful information. An example of this is found in the four dimensions of CRM as represented by customer identification, attraction, retention and development (Ngai et al, 2009). These are made more meaningful under data mining techniques that exploit the value of the database via association, classification, clustering, forecasting, regressions, sequence discovery and visualization (Ngai et al, 2009).   As a means to understand the manner these seven areas contribute to providing a means to extract and or explore Customer Relationship Management data Ngai et al (2002) provide the following illustrative representation:   Figure 1 – Seven Factors of Data Mining CRM Extraction and Analysis (Ngai et al, 2002)   As shown, these seven factors tie into the four CRM cornerstones represented by customer identification, attraction, retention and development (Ngai et al, 2009). 1.5 Impact of Data Mining on CRM and Current Practices The preceding understanding concerning the four Customer Relationship Management cornerstones and seven data mining factors provided an overview of how these two areas interact with each other. As a means to understand the manner the seven data mining facets draw data from CRM, an analysis of each of the areas provides the foundation for understanding the above. 1.5.1 Association In terms of data mining, there is what is termed as the association rules that are an important aspect (Larose, 2005). Under data mining, association rules represent if / then statements that are utilised to aid in the uncovering sets of relationships between data that can seemingly appear to be unrelated under a relational database or information repository (Larose, 2005). An example of this in the application is demonstrated by a consumer who shops for eggs is also has a high potential to include milk and or bread in that same purchase (Temple University, 2004). With regard to association rules, it has two parts or components (Bastide et al, 2000). These represent an antecedent (represented by if), and a consequent (represented by then) (Bastide et al, 2000).   The antecedent represents an item that is found within the data, with a consequent representing an item found in combination with the antecedent (Bastide et al, 2000). Association rules are created in data through analysing it (the data) for what is termed as the frequency of if / then patterns and or representations. It then utilises criteria support as well as confidence as the manner to uncover and identify the relationships that are most important or frequent (Zaki, 2000). With regard to the above, support represents an indication concerning the frequency that the items appear within the database. Confidence indicates the number of occurrences or times the if/then representations that have been uncovered are found to be true (Zaki, 2000).   The usefulness of data mining association rules is that they provide a means to analyze as well as predict the behaviour of customers (Lin et al, 2002). Its use is an important component in shopping basket analysis, catalog design, the layout of stores as well as clustering products to heighten sales and consumer convenience (Lin et al, 2002). The following represent theoretical applications used to extract information correlations depending upon the end results desired. 1.5.2 Classification In terms of data mining, classification is a technique that is used for predicting group membership in terms of data instances (Glusman et al, 2000). Classification predicts class label categories and also aids in classifying the data through constructing a model (Glusman et al, 2000). The preceding model is based upon the training set as well as values (termed as class labels) as the classifying attributes (Ishibuchi and Yamamoto, 2004). Classification segments the records of customers into distinct categories that are termed as classes (Ishibuchi and Yamamoto, 2004). Its purpose is to group items using certain key characteristics that have a number of by which can be accomplished (Tan et al, 2006).   The goal of classification is to provide an overview of the problem posed by categories (classes) (Colet, 2008). Some of the mathematical formulas used in classification approaches entail regression analysis, and decision trees (Colet, 2008). In utilising a regression approach, it is used to divide a classification into regions that thus aids in the prediction of classes (Colet, 2008). Distance represents another method used in a classification where items are placed into the class they are closest to (Hall et al, 2009). The above is used to determine the distance between a class and an item with classes represented by a central value (centroid). The representative point is termed as a medoid, and the formulation includes individual points plotted as a result of the above (Hall et al, 2009).   In utilising a decision tree Kohavi and Quinlan (2002) advise they are portioning based where the search spaces are divided into rectangular regions. As one of the more popular of the data mining measures, decision trees represent one of the more used methods for learning from examples that are feature-based (Rastogi and Shim, 2000). It, decision trees, represent a means of classification determined by a scheme that generates a tree along with a set of rules that are a model of different classes from a particular dataset (Rastogi and Shim, 2000). Hans and Kamber (2001) tell us a decision tree is a flow chart like structure where its internal nodes represent a test on an attribute. In this instance, each branch of the tree denotes an outcome of the test. They further explain the leaf nodes signify the class or classes of distributions (Hans and Kamber, 2001).   In terms of strengths, decision trees represent a way to generate rules that are understandable, along with being able to handle categorical as well as numerical attributes (Srivastava et al, 2002). In addition to the above, decision trees provide clear indications of the fields that are the most important for classification and or prediction (Srivastava et al, 2002). In terms of weaknesses, decision trees in some forms can only handle binary valued target classifications (Qin and Lawry, 2005). Others are capable of assigning records to a number of classes that are arbitrary (Qin and Lawry, 2005). The issue with the above is the potential for the decision trees to be error-prone when the numerical set of training examples is small (Qin and Lawry, 2005). The above is even more pronounced in instances where a tree has many levels and consequently a number of branches per node (Qin and Lawry, 2005). With regard to data cost, a decision tree represents an approach that is termed as being computationally expensive (Qin and Lawry, 2005). The reason for this is that at each node the individual candidate field splits need to be examined prior to locating the best example (Qin and Lawry, 2005). 1.5.3 Clustering Clustering is one of the oldest forms of prediction utilised in data mining (Strehl and Ghosh, 2002). Simply put, it represents having like records grouped together, hence the term cluster. It, clustering, represents a means for the end users to obtain a high-level view of what is occurring in a database (Strehl and Ghosh, 2002). Clustering is used in some instances to accomplish what is termed as mean segmentation (Strehl and Ghosh, 2002). This is useful in marketing as a way to visualize various customer and other activities in their behaviour patterns and purchasing (Strehl and Ghosh, 2002).   Clusters can be constructed using a number of different variables, which includes age, income, occupation, ethnic background, zip codes and other factors (Crespo and Weber, 2005). Various tags (names) are assigned to the cluster group categories, based on company preferences, and then run against the database to pull selected datasets (Crespo and Weber, 2005). By running the database against the tagging parameters, the data sets are grabbed thus offering a view of what is in the database and its value in terms of information possibilities (Crespo and Weber, 2005).   1.5.4 Forecasting Forecasting is a function that can be performed using data modeling when the database has up to date information or context that fits the parameters of the forecast (Bose and Mahapatra, 2001). Using customer data on past purchases, dates, amounts and other factors, forecasts concerning future tendencies based on the analysis parameters used or desired can yield useful information. Examples represent using last year’s Christmas sales to forecast the probability of spending patterns this year by day, amount and general categories or specific categories as a means to aid in organizing promotions and events.   1.5.5 Regression In terms of regression, it is a method of prediction that is based upon either an assumed and or known numerical value (Bal et al, 2004). The value (output) is a factor of what are termed as a series of recursive portioning that at each step contains one more numerical values as well as another grouping of dependent variables that branch out into separate pairs (Bal et al, 2004). A regression tree begins with either one or multiple precursor variables and ends or terminates with a final output variable (Bal et al, 2004). Under this method (regression), the dependent variables can be either continuous or they can be discrete (Bal et al, 2004).   The main difference between a regression tree and a classification tree is the dependent variable (Prasad et al, 2006). Under a regression tree, the dependent variables are numerically dependent, while under a classification tree they are categorical (Prasad et al, 2006). In a regression tree, the values are either discrete as well as ordered, or they can be indiscrete (Prasad et al, 2006). In terms of a classification tree, it differs as it has a set number or amount of unordered values (Prasad et al, 2006).   1.5.6 Sequence discovery As brought forth a number of times in this investigation concerning data mining, varied hidden and or not so obvious facts, patterns and or associations reside in a database. Sequence mining represents looking for or uncovering patterns that reside in a database through running various combinations that lead to their discovery (Fayyad et al, 1996). Through running differing information or series patterns to search for hidden associations, a company can extract more value and use from its database (Fayyad et al, 1996). This also allows or permits it the opportunity to hone it to increase marketing effectiveness and sales results through discovered sequences and patterns that are identified (Fayyad et al, 1996).   Other areas of discovery that can occur are the detection of fraud activity, improvement in manufacturing quality through product failure detections parameters and a host of other areas (Fayyad et al, 1996). Sequence discovery can be planned or unplanned, but it usually occurs under the former as a company runs varied types of data request patterns to ascertain if there are patterns hidden in the data file (Fayyad et al, 1996).   1.5.7 Visualisation As the word implies, data mining visualization represents the study of data under a visual context. The purpose behind data visualization is to communicate effectively as well as clearly via graphical means (Keim, 2002). The above can be through an appealing representation illustration or raw data display, however, the second method is more useful in revealing what is contained in a database run and the associations inherent in a data computation (Keim, 2002). Soukup and Davidson (2002) elaborate on this by stating visualizing can and does provide benefits through revealing complex information in a manner that allows it to be conceptualised differently due to seeing how patterns and or associations link together. 1.6 Recommendations represented by a framework for successfully integrating data mining into CRM The exploration into CRM and data mining uncovered these two components are a process as opposed to a format or defined framework. This is due to CRM representing a customised framework for gathering and collecting data for input, and data mining representing data extract and discovery based on utilisations having value to individual corporate areas that are defined or desired. The research into CRM and data mining has uncovered the importance as well as the significance of the intertwined relationship that exists between them. Becker et al (2009) in a study representing Customer Relationship Management indicated many companies that implemented CRM have reported unsatisfactory levels of improvement. The preceding sweeping analysis represents only part of the story as Becker et al (2009) identified two important base issues to be considered and understood with regard to the implementation of Customer Relationship Management. Payne and Prow (2005, p. 168) clarify this by stating CRM is “… a cross-functional integration of processes, people, operations and marketing capabilities that is enabled through information, technology, and applications”.   The above indicates the success or effectiveness of Customer Relationship Management and data mining is a product of the preparation, planning, and depth of analysis as opposed to any inherent failure or inability present in the process itself. This view has been brought forth throughout this examination. Through examples, it has consistently pointed to the importance of having prior effective planning for data input as well as modes of examination. These are the foundations of extracting information that is and has been pointed to concerning the importance of having prior planning and purposes for data input. This also includes experienced and expert individuals to mine the data. The recommendations concerning a framework to successfully integrate data mining into CRM starts with management having a clear understanding of what it seeks to accomplish, and then devising systems and collection points to input and gather data from every contact point possible. The key to successful data utilisation is having an extensive base of information from which to draw upon.   The more data available in the Customer Relationship Management database the better the analysis as well as comparative factors that result from this foundation. As indicated by Becker et al (2009), the implementation and or use of CRM requires a commitment on the part of the organization in order to be successful and yield results. It also takes experienced personnel in data mining to extract information and look for relationships and patterns as explained herein. In fact, the keys to success in Customer Relationship Management and data mining is not a mystery, but rather a commitment and understanding from the outset to fully develop data inflows and expert data mining personnel as the means to realize the potentials of the process.     References Bal, J.,, Yates, A., Crippen, G.m Fischer, V., et al (2004) Use of Classification Regression Tree in Predicting Oral Absorption in Humans. Journal of Chemical Information and Modeling. 44(6) Base, I., Mahapatra, R. (2001) Business datamining — a machine learning perspective. Information and Management. 39(3). Bastide, Y., Pasquier, N., Taouil, R., et al (2000) Mining Minimal Non-redundant Association Rules Using Frequent Closed Item sets. Computational Logic. 10)8) Becker, J., Greve, G., Albers, S. (2009) The impact of technological and organizational implementation of CRM on customer acquisition, maintenance, and retention. International Journal of Research in Marketing. 26(3) Berson, A., Smith, S. (2002) Building Data Mining Applications for CRM. New York, N.Y., United States. McGraw Hill Publications Cabena, P., Hadjnian, P., Stadler, R. et al (2007) Discovering Data Mining: From Concept to Implementation. New York, N.Y., United States. Prentice Hall Colet, E. (2008) Clustering and Classification: Data Mining Approaches. Accessed on 5 August 2012 from http://www.taborcommunications.com/dsstar/00/0704/101861.html Crespo, F., Weber, R. (2005) A methodology for dynamic datamining based on fuzzy clustering. Fuzzy Sets and Systems. 159(2) Fayyad, U., Piatetsky-Shapiro, G., Smyth, P. (1996) From Data Mining to Knowledge Discovery in Databases. AI Magazine. Spring p. 3 Glusman, G., Bahar, A., Sharon, D. et al (2000) The olfactory receptor gene super family: data mining, classification, and nomenclature. Mammakian Genome. 11(11) Hall, M., Frank, E., Holmes, G. (2009) The WEKA data mining software: an update. ACM SIGKDD Explorations Newsletter. 11(1) Hans, J., Kamber, M. (2001) Data mining concepts and techniques. New York, N.Y., United States. Morgan Kaufman Publishers. Ishibuchi, H., Yamamoto, T. (2004) Fuzzy rule selection by multi-objective genetic local search algorithms and rule evaluation measures. Fuzzy Sets and Systems. 141(1) Keim, D. (2002) Information visualization and visual data mining. Visualization and Computer Graphics. January / March Kohavi, R., Quinlan, J. (2002) Data mining tasks and methods: Classification: decision-tree discovery. New York, N.Y., United States. Handbook of data mining and knowledge discovery Larose, D. (2005) Discovering Knowledge in Data: An Introduction to Data Mining. London, UK. John H. Wiley and Sons. Lin, W. Alvarez, S., Ruitz, C. (2002) Efficient Adaptive-Support Association Rule Mining for Recommender Systems. Data Mining and Knowledge Discovery. 6(1). Ngai, E., Xiu, L., Chau, D. (2009) Application of data mining techniques in customer relationship management: A literature review and classification. Expert Systems with Application. 36(2) Payne, A., Prow, P. (2005) A strategic framework for Customer Relationship Management. Journal of Marketing. 69(4). p. 168 Prasad, A., Iverson, L., Liaw, A. (2006) Newer Classification and Regression Tree Techniques: Bagging and Random Forests for Ecological Prediction. Ecosystems. 9(2) Qin, Z., Lawry, J. (2005) Decision tree learning with fuzzy labels. Information Sciences. 172(2). Rastogi, R., Shim, K. (2000) PUBLIC: A Decision Tree Classifier that Integrates Building and Pruning. Data Mining and Knowledge Discovery. 4(4). Rygielski, C., Wang, J., Yen, D. (2002) Data mining techniques for customer relationship management. Technology in Society. 24 Schmitt (2008), Customer Experience Management: A Revolutionary Approach to Connecting with Your Customers. New York, N.Y. John H. Wiley & Sons. P. 18 Soukup, T., Davidson, I. (2002) Visual Data Mining: Techniques and Tools for Data Visualization and Mining. New York, N.Y., United States. John H. Wiley & Sons. Srivastava, A., Han, E., Kumar, V., Singh, V. (2002) Parallel Formulations of Decision-Tree Classification Algorithms. High Performance Data Mining. 3(1) Strehl, A., Ghosh, J. (2002) Relationship-based clustering and cluster ensembles for high-dimensional data mining. Austin, Texas, United States. University of Texas at Austin Tan, P., Steinbach, M., Kumar, V. (2006) Introduction to data mining. New York, N.Y., United States. Addison Wesley Publishers Temple University (2004) Data warehousing, filtering, mining. Accessed on 3 August 2012 from http://www.scribd.com/doc/78046368/1-Association-Rule-Mining Williams, D. (2008) Services Marketing: Integrating customer Focus Across the Firm. Maidenhead, UK. McGraw Hill Winer (2001) A Framework for Customer Relationship Management. California Management Review. 43(4) Xu, Y., Yen, D., Lin, B., Chou, D. (2002) Adopting customer relationship management technology. Industrial Management & Data Systems. 102(8) Yu, L. (2001) Successful Customer Relationship Management. MIT Sloan Management Review. 42(4) Zaki, M. (2000) Generating non-redundant association rules. New York, N.Y., United States. Proceedings of the sixth ACM SIGKDD international conference on Knowledge discovery and data mining Zhang, S., Zhang, C., Yang, Q. (2003) Data preparation for data mining. Applied Artificial Intelligence. 17(6). Read More