Methods of Analyzing Data in Data Warehouse

Data Mining methods for Customer Relationaship Management

Abstract-Data warehousing and data mining, applied to Customer Relationship Management (CRM), are a relatively fresh sector for organizations considering their immense advantage. Through data mining, organizations can identify key customers, predict future behaviors and take proactive and knowledge driven decisions. In this article, we discuss few data mining techniques for various CRM activities.

Keywords- Customer relationship management (CRM); Data mining

The crucial element of any successfull business is value creation for a customer. CRM basically tries to pull in, maintain and manage customers. Business intelligence analyses and interprets large amounts of customer data to provide organizations with practical information which can be used to devise strategies for improvement and growth. The ease of data collection in todays world coupled with the low cost of maintaining a data warehouse has increased the accessibility of huge customer data. Data mining has become a backbone for CRM activities. Few years back, data analysis was linked with expensive computing and complicated logic which only mathematicians could understand but this has changed now due to the availability of user friendly desktop tools. Data mining has become a backbone for CRM activities.

Data mining and data analysis methods when used properly help to better all CRM phases. In order to retain important customers and to continue to provide the best customer satisfaction, availability of timely and actionable information is very critical. Such information plays a key role in facilitating smart organizational decisions which enables in creating better value for the customer. For maintaining a successful CRM strategy, investments in the understanding of data mining techniques is crucial.  This paper would try and mention some of the data mining techniques for optimization of CRM activities and also to find which technique is important.

Pulling in new customers, minimizing defection by important customers and enhancing the experience of existing customers are the key aspects which an effective CRM business strategy assists with. .This article mentions some of the data mining techniques for optimizing CRM activities.

A. Content analysis for identifying the right customer

The method of deciding on the units of analysis based on the direction of the study is known as content analysis. The concept  of ‘computable’ and ‘specifiable’, categories are defined which help in grouping the processed units of data, quantify and analyze the samples. Content analysis is often used on various advertisements styles, in email communications and on social media. The figure below shows information flow from data collection to useable knowledge.

Figure: Knowledge discovery process. [2]

First step applies pre-established rules to select data and categorize them into relevant groups. Next step is to clean up and reorder data by disposing off unnecessary information, to establish record-keeping formats and for the purpose of maintaining the integrity and consistency of data which helps in the construction a data platform. The organised data is refined further by grouping into related subjects using data transformation methods. Models and interconnections are developed after analysis which in turn help with decision support. This data when applied to a companies unstructured data is used to find the core (right) customer which in turn helps in planning efficient business strategies and providing appropriate customer service to different customers.

B. Mining Customer behavioural changes

In an ever changing business environment, trying to analyse customer behaviour is very helpful. Customer, transaction and product databases can be used for change mining. Below figure shows the flow of change mining.

Fig. 2. Flowchart for mining changes in customer behavior. [3]

Generally some usefull information is concealed in the huge amount of raw data and this needs to be extracted using data transformation. Customer and transaction databases can be used to analyse customer behaviour by using data integration and transformation. As per the figure above, customer, product and transaction databases are used to analyse the customer behavioural variables: Recency, Frequency, Monetary (RFM). Recency represents most recent transaction time, frequency is the number of purchases during a certain period and monetary is  the standard amount of expenditure. The attributes, frequency and monetary, are used to segment the customer into different categories namely: Uncertain, Frequent, Spender, Best. The Recency variable can be combined with the above study to develop a target market.

Association rules are used for mining customer behaviour by analyzing the relationship of products purchased by a customer in retail stores. A classic application of association rules is the market basket analysis where products bought by a customer during a visit to the supermarket are analysed. This can be used to identify corelations between product purchase and customer profile represented by demographic variables. Customer behavioural data are most effective for generating predictive data which optimizes the CRM.

The study of investigating change in customer behaviour is called change mining. Change mining looks at changing customer behaviour to develop some pointers which can be used to mathematically quantify the change in beahviour.  The output of all the above is analysed data which can be used to support efficient marketing.

C. Learn and Usage model for Customer retention

Loosing current customers to a rival company is termed as customer churn.  Finding better methods for customer retention is very crucial as acquiring new customers proves to be very expensive.This is two phase model: Learning and Usage. The learning phase constructs a churn model which tests and predicts the probablity of defection for a certain customer based on historic data. A policy model is also constructed which clusters the churners by grouping them based on notable attributes. These are then used for creating proper policies for individual groups. During the usage phase the churn model predicts if a certain customer would defect and when there are strong chances of defection, the policy model comes up with relevant policies to retain them. This method not only predicts churning but also helps in reducing it. The following figure shows the learner mode architecture:

Fig3: Architecture for learning mode [5]

During the learning mode the churn model learner is constructed using histiorical data of individual subscriber like loyalty history, deactivation data, payment history, usage patterns, etc. The churn model can be represented as a decision tree which is used to comment on the likelihood of a specific customer defecting based on their previous data. The policy model constructor is used to build retention strategies for potential churners. The policy model builds retention strategies in two steps, first step is to consult learned churn model to recognize the attributes which have strong relation to the churning and based on these attributes the churners are classified into diferent groups and labelled as per the most significant attribute. Second step is to analyse the significance of these attributes and reccomends policies for retaining the group of churners. In the usage phase the churner model is consulted to predict the chances of customer churn based on the customer data. If the churn probablity is more than 60% then it is considered that the customer has high chances of churning and so policy model proposes a policy response based on the attribute of the group to which the customer belongs for the purpose of retention.

This journal reviews the literature concerned with data mining techniques and its applications in CRM. Considering the current competition in the market all organizations suffer from a lot of customer churn which results in huge losses as the cost for acquiring new customers is ten times more than that of retaining existing customers. Hence a lot of research has been made on customer retention. This is also evident from the number of reasearch papers submitted between the years 2000 – 2006 [1]. The area of customer retention definetly seems to be crucial and requires further research. Data mining methodologies help in providing better understanding of raw data and hence would always be one of the major areas to be researched upon in the futire.

References

1. E.W.T Ngai, Li Xiu and D.C.K Chau, “Application of data mining techniques in customer relationship management” Expert Systems with Applications, vol. 36, Issue 2, part 2, March 2009, pp. 2592 – 2602.
2. C.W Chang, C.T Lin, L.Q Wang, “Mining the text information to optimizing the customer relationship management“, Expert Systems with Applications, Vol. 36, Issue 2, Part 1, March 2009, pp. 1433-1443.
3. Mu Chen Chen, Ai Lin Chiu, Hsu Hwa Chang, “Mining changes in customer behavior in retail marketing“, Expert Systems with Applications, vol. 28 (2005), pp. 773-781
4. W. Buckinx, D.V.D. Poel, “Customer base analysis: Partial defection of behaviorally-loyal clients in a non-contractual FMCG retail setting“, European Journal of Operational Research, 164 (2005), pp. 252-268.
5. Bong-Horng Chu, Ming Shian Tsai, Cheng Seen Ho, “Towards a hybrid data mining model for customer retention“, Knowledge-Based systems, Volume 20, Issue 8, December 2007, pp. 703-718
6. Y.L. Chen, K. Tang, R.J. Shen, Y.H. Hu, “Market basket analysis in a multiple store environment“, Decision Support Systems, 40 (2005), pp. 339-354.
7. YangSeog Kim, W. Nick Street, “An intelligent system for customer targeting: a data mining approach“, Decision Support Systems, Volume 37, Issue 2, May 2004, pp. 215-228.