NATIONAL COLLEGE OF IRELAND

Application of Data Mining Techniques in

Customer Relationship Management



Literature Review











1

Table of content
Abstract..1
Introduction...2
1. Integration of Data Mining in CRM framework .3
2. Classification Method ...4
2.1. Customer Identification................................4
2.2. Customer Attraction....4
2.3. Customer Retention........5
2.4. Customer Development .............5
3. Data Mining Application Domains............5
3.1.1. Data Mining for Retail...............6
3.1.2. Data Mining for Banking...............7
3.1.3. Data Mining for Telecommunications.........7
3.2. The Data Mining Methodologies............8
4. Data Mining Techniques for CRM( Segmentation) ........9
4.1. Association Rule Learning............10
4.2. K-Nearest neighbour .............10
4.3. Classification ..................10
4.4. Forecasting ................10
4.5. Clustering................11
4.6. Linear/Logistic Regression Analysis...........12
4.7. Visualisation................12
5. Data Mining Tools.................12
6. Conclusion...................................13
7. References.................. 13




2




Abstract
For the last two decades Customer Relationship Management (CRM) had been
gained a huge popularity in business area and Customer Retention strategy
became an increasingly emergent issue within CRM.
The purpose of this paper is to capture the major concepts, the most relevant
techniques and highlight the elements of Data Mining process in order to improve
Customer Relationship Management.
In order to achieve the objective of this paper, eighteen articles and research studies
about the Application of Data Mining techniques to Customer Relationship
Management were identified and reviewed. Afterwards, eleven articles and books
were selected, reviewed and classified.
This analysis provides a guideline for a future research and a help in gaining
knowledge in application Data Mining techniques to CRM.


Introduction
Customer Relationship Management (CRM) is considered basically the most
important talking point in business world, which is developing a new business culture
within it and makes customers the most important asset within an organisation.
Therefore CRM is considered a strategy which integrates sales; service and
marketing that facilitate operating procedure in order to understand customers from
different perspectives and to build manage and strength loyal customer relationships.
Nowadays CRM is customer-centric focused on customer insight with its scope
personalised which treats customers as particular entity. Customers are understood
and grouped considered their distinguished needs, behaviours and preferences.



3

1. Integration of Data Mining in CRM
Since few decades ago, Data Mining technology is developing in areas such as
statistics, IT, artificial intelligence and machine learning; nowadays already in data
integration, business intelligence or business environment.
Using Data Mining tools we are able to use raw data (real data) in order to construct
a business model. With data Mining process that uses data analysis and modelling
techniques we are able to discover patterns and relationships in datasets that can
help in making predictions or to build a plan on which group of customers you need
to focus by giving individual offers , or to identify good or bad customers (
customers which will stay or who will leave). This type of CRM application that use
DM is called Analytics CRM.
In this case Data Mining is considered an important part of CRM, which makes
companies to focus towards customer oriented methods.
Three general categories can be gained from the process orientation aspect:
Discovery process in which by searching into database we can find hidden
patterns without stating a hypothesis about what these patterns can be, or having a
predetermined idea.
Predictive Modelling process of using patterns discovered in datasets and use
them in order to predict future.
Risk Analysis and Fraud detection (Forensic Analysis) - process in which
extracted patterns are used to find abnormal data elements.
In order to build an effective CRM system using Data Mining, the following steps has
to be followed:
Define Business Question
Creating Marketing Database
Explore Data
Prepare Data for Modelling
Build the model
Assess Model
Set-up Model and assess Results




4

2. Classification Method in CRM dimensions using Data Mining
Data Mining in CRM is helping organisations for a better discrimination and a
more effectively in allocating resources to the most profitable group of customers
through the cycle of customer identification, customer attraction and customer
retention and customer development. The following dimensions are the base in
CRM efforts that has to be considered in order to gain customer insight.

1. Customer Identification
Customer identification which in other words means customer acquisition too, is the
step that involves targeting the customers who are most profitable to the company or
population who are most likely to become customers. This dimension involves target
customer analysis and customer segmentation in order gain profitable customer
segments through customer analysis (subdividing of the entire customer base into
smaller customer groups with similar elements).
2. Customer Attraction
After the segments of potential customers are identified, organisations are able to
direct their resource into attracting the targeted group of customers.
Marketing is usually the direct method and promotion process in customer attraction.
3. Customer retention
For veteran companies, customer retention is the most important concern for
Customer Relationship Management. The key element in customer retention is
customer satisfaction, or, in other words the correlation between customer
expectation and their perception of being satisfied.
Strategy elements for successful customer retention are various, but the most
common nowadays are One-to-one or Personalised Marketing Campaigns which
include:
Analysis, detection and prediction of changes in customers behaviour or
lifestyle
Customer profiling
Loyalty programs
Personalised offers/ discounts

4. Customer Development
5

This consists from the idea of expanding customer transactions, transactions value
and customers personal advantages. An important element in customer
development is customers basket analysis which gives the opportunity for
businesses to create personalised offers, to reveal customer behaviour or to predict
customers social status and financial situation.




















Fig.1. Classification framework for Data Mining Techniques in Customer
Relationship Management


Customer
Customer
Relationship
Managemen
t
Customer
Development
Customer Lifetime Value
Up/Cross Selling
Market Basket Analysis
Customer
Identification
Target Customer Analysis
Customer Segmentation
Customer Retention
One-to-one Marketing
Complaints Management
Customer
Attraction
Direct Marketing
Classification
Clustering
Forecasting
Regression

Visualisation
Association
6

5. CRM Model

A simple CRM Model consists from a set of the following elements, as per
Fig. 2.























Fig.2. Customer Relationship Management Model

3. Data Mining Application Domains
CREATE A DATABASE
ANALYSE
CUSTOMER
SELECTION
CUSTOMER
TARGETING
PRIVACY ISSUES
MARKETING
RELATIONSHIP

METRICS

7

At the moment many different industries such as finance, banking, healthcare,
insurance, database marketing, publicity, sales, etc., are using Data Mining tools for
CRM. Banks are using Data Mining for customer segmentation, credit approval,
predicting payment or detecting illegalities.
Data Mining in Retail can be used in order to gain information about sales directions,
customer tradition of buying and preferences, etc.



1. Data Mining for Retail
Credit cards, Loyalty cards or Point-of-Sales systems can provide retailers with a
wide range of records about every shopping transaction. All this data stored in
databases is very valuable and enable retailers to understand better about
customers preferences and segment customers in different groups.
The following applications include:
Basket Analysis this technique allows retailers to discovers which items
customer tend to buy together and their shopping behaviour in order to build
more efficiently sales or marketing strategies.
Customer segmentation is the process which transforms a large market into
smaller groups of customers with a similar buying behaviour, social status or
financial possibilities.
Sales forecasting is using time-based data in order to make stocking
decisions. For example if customers buys a products today, when they are
more likely to buy a complementary one later?
Database marketing It is similar to basket analysis, but here retailers are
able to develop customer profiles with certain behaviours in order to focus on
effective promotions.
Merchandise planning retailers are able to improve merchandise planning
by examining patterns in stores with similar demographic characteristics or to
change merchandising plan considering customers behaviour ( Beer and
Nappies example).

2. Data Mining for Banking
Nowadays in Banking Sector we have customers with a wide range of options that
they can choose in order to do their business. In this case every bank realises the
importance of long-lasting relationship with their customers in order to increase their
8

profits and gain success. In banking sector CRM focuses on two the most important
things: customer acquisition and customer retention. To note that banks are directing
their resources in order to retain the most profitable customers, to find and
implement the most profitable solution.
In order to improve Customer Retention in Banking sector, three steps has to be
followed:
1. Measurement of existing customer retention the first step in improving loyalty
and profitability analysis by segment.
2. Identification of root causes of defection and related key service issues
3. Development of corrective action to improve retention.
The following diagram explains data mining techniques and algorithms used to the
banking sector. Decision Tree is applicable for customer retention and for credit
approval process, as well as Logistic Regression. Clustering model using EM can be
used to detect frauds and prevent risks.



Fig.3. Data Mining Techniques and Algorithms in banking sector
Decision tree models are used to predict problems and solve classification. In order
to avoid building very expanded and large decision trees, the best tree has to be
found or use a stopping rule for deciding when the sub tree was created, or identify
roles at the node of splitting data.
The most common data mining methods for customer profiling are:
Clustering( descriptive)
9

Classification ( predictive) and Regression (predictive)
Association rule discovery (descriptive) and sequential pattern discovery
(predictive).
From marketing aspect, banks also can analyse the past trends in order to determine
the present demand and forecast the customer behaviour of various products and
services. Data mining techniques are able to identify profitable customers from non-
profitable ones. Another major area of development in banking is Cross selling, for
example banks make an attractive offer to their customer by making them to buy an
additional product or service.

3. Data Mining for Telecommunications
Telecommunication industry was the first industry that used Data Mining techniques.
Also, telecommunications also store the most valuable and qualitative data in their
usually huge databases.
Companies from this industry possess data about telephone calls, call records,
information about customers and billing information.
The main issue aimed by Telecommunication industry is to retain existing customers
and attracting new ones.
Knowledge that can be discovered in this industry includes:
Call record detail analysis if customer segments with the same patterns are
identified, then companies can develop more attractive marketing strategies
and promotions.
Customer segmentation the same as in other industries, in
telecommunication customer segmentation plays a very important rule. The
differences which may make this task difficult can be relevance and quality of
data, as was as the ability to use intuition or over-segmentation.
Customer loyalty - some of the customers tend to switch providers quite
often, so telecom companies use Data Mining Tools in order to identify loyal
customers and target their spending on those which are more profitable. (i.e.
offer discounts, or personalised offers)
Types of Telecommunication Data:
Call-Detail Data
Average call duration
Call to/ from different area code
Call period/ time
Network Data
Configuration of equipment
10

The support network
Customer Data
Personal details
Telephone type
Payment history, etc.
These methodologies and be applied to other industries in which CRM is also the
key to success and competitiveness.


4. Data Mining Techniques for CRM
Data Mining Techniques can be classified as follows:
1. Association Rule Learning
This method is well-known for discovering interesting relations between two
or more variables in databases. In retail, specifically in supermarkets
association rules can discover regularities between products in transactional
data recorded by POS systems. But, nowadays this rule is applied in areas
such Web usage mining, production, fraud detection, etc.
For an association modelling we use statistical tools and predefined
algorithms.

2. K-Nearest neighbour
In pattern recognition, the k-Nearest Neighbours method used for classification and
regression. In both cases it can be useful to weight the contributions of the
neighbours, so that the nearer neighbours contribute more to the average than the
more distant ones.
3. Classification
In DM classification is the most common learning model which helps in building a
model to help predict future customer behaviour by classifying data from datasets
into predefined categorical classes. An example to this technique might be used to
identify loan applicants and classify them into low, medium or high credit risk.
4. Forecasting
This type of method involves in general predicting outcome based on previous data
patterns. Outcome is generally used in predicting response of various campaigns or
even sales. In retails data mining is used for forecasting an effective inventory
system in which, based on previous sale history, retailers can predict their monthly
sales and manage orders accordingly.
11

This Data mining technique is also used to forecast bankruptcy or other forms of
default.

5. Clustering
Similar to classification model, clustering maps data items into one of many
categorical classes (segments, groups or clusters). Unlike classification in which
groups are predefined, in clustering classes are determined by data characteristics.
In other words, clusters are defined naturally in groups of data based on similarity
metrics or density models. For example, a clustering model can be defined by a
group of customers which tend to buy the same group of products.
6. Linear Regression Analysis
This is a very important statistical technique that predicts the probability of the
targeted categories and to map each data to a real value proto provide prediction
value. This technique uses regression curve fitting, prediction(forecasting), model
relationships and testing hypothesis about relationships between variables.
7. Visualisation
Visualisation is the technique that refers to the presentation of the data in a manner
that any user can view and understand complex patterns. There numerous
techniques that can be used to represent, describe and visualise complex data.

5. Data Mining Tools
Nowadays there are a wide range of available data mining tools in the market that
solve even very sophisticated problems. In general these tools are devided in three
different categories, and they are as follows:
General purpose tools
SAS Enterprise Miner
IBM Intelligent Miner
IBM SPSS Modeller
XLMMiner
CART & MARS
Oracle Darwin
Integrated DSS/OLAP/DM tools
Cognos Scenario
Business Objects
Application-Specific tools
KDI (Knowledge Discovery One)
Unica Detect
12

Unica Leaders
Unica Predictive Insight

The integrated DSS/OLAP/DM tool work with real and compelling business
requirements and has a single multifunctional decision-support tool that can provide
management reporting, analytical processing and DM capabilities within the same
framework.
Various technologies differ from each other in terms of effectiveness and actually
how easy is to use it. Data scientists have the ability to collect and explore data, but
business managers determine how to exploit this data.


6. Conclusion
Nowadays, there are many concerns and issues in regards to data privacy and many
organisations have to assure that consumers privacy is respected and that there are
legal policies within the organisation that will establish and maintain an effective
consumer relationship.
As CRM solutions focus more on analysing customers information in order to gain
financial benefits, there is a lack of privacy. Therefore, consumers are starting to
concern about this issue and the process of managing customer relationships is in
continuous increase and becomes even more necessary.


7. References
1. Farooqi, R. and Raza, K. 2011. A Comprehensive Study of CRM through Data
Mining Techniques. Department of Commerce & Business Studies. [report] New
Delhi, India: Apeejai School of Management.
2. Improving Customer Relationship Management Using Data Mining. 2012.
International Journal of Machine Learning and Computing, 2 (6), pp. 874-877.
3. Gupta, G. and Aggarwal, H. 2012. Improving Customer Relationship Management
Using Data Mining. International Journal of Machine Learning and Computing, Vol.
2. [report] New Delhi, India: pp. 874-877
4. E.W.T., N. and Xiu, L. 2009. Application of data mining techniques in customer
relationship management. Expert Systems with Applications, 2 (32).



13


5. Rygielskia, C., Wangb, J. and C. Yen, D. 2002. Data mining techniques for
customer relationship management. Technology in Society, 4 (2), pp. 483-502.
6. Berry, M. J. A. 2011. Data mining techniques. Chichester: Wiley-Interscience.
7. Tsiptsis, K. and Chorianopoulos, A. 2009. Data mining techniques in CRM.
Chichester, West Sussex, U.K.: Wiley.
8. Migu, V. L. 2012. Analytical Customer Relationship Management in Retailing
Supported by Data Mining Techniques. 9. Universidado de Porto.