3. Bottom-line questions in CRM
Who are my most profitable customers?
Who are my repeat website visitors?
Who are my loyal customers?
Who is likely increase purchase?
What clients are likely defect to my rivals?
Will my customer respond to the direct mail
solicitation?
4. Data mining as part of CRM strategy
Business that knows it’s customers best will serve
them best
Best spent marketing dollar is the one that retains the
existing customer
Business forecasting is essential
A fast food chain doing hourly demand projection for its
outlets
Objective and quantifiable insight into customer
profiling data
Which of my high-profit customers are most likely to leave?
Be proactive to retain them.
Which of my low-profit customers are least likely to leave?
Raise their price and make them more profitable.
5. Data mining is NOT magic
Data mining can not ingest noisy data
Data mining can not use ready-to-use business
strategies based on analysis of raw data without
intelligent interpretation
Garbage-in garbage-out
The information produced by data mining apps require
human review
Real data mining is methodology with technology
support
“Hype” data mining is mythology with marketing
support
6. Data mining techniques in CRM life
cycle stage
CRM stage Activities Data Mining technique
Discovering Lead generation Customer acquisition profiling
Web data mining for prospects
Targeting market
Reaching Marketing programs Customer acquisition profiling
Selling Contact selling Customer acquisition profiling
Online shopping
Scenario notification
Customer-centric selling
Satisfying Product performance Customer retention profiling
Service performance Scenario notification
customer service Customer centric selling
Inquiry routing
Retaining Customer retention Customer retention profiling
Scenario notification
Individual customer profiles
7. Data mining methodologies
CRISP-DM (Cross-Industry
Standard Process) methodology
SEMMA (Sample, Explore,
Modify, Model, Assess)
methodology
Other Common approaches
Different tools have different way of
doing the typical data mining task
Data gathered -> conditioned & analyzed
-> descriptive models -> predictive
models
8. Data mining methods
Classification & regression
Association & sequencing
Association rules (Market Basket Analysis)
Sequential analysis
Clustering
Link analysis
Visualization
Regression
Rule induction
9. The mathematics in Data mining
Feature space (Euclidian space)
Probability distribution
Standard deviation and z-score
Feature space computation
Clusters
Numeric coding
Creating Ground truth
Synthesis of features
10. Data mining techniques
Neural Network
Problem solving with
Neural network
Training and validating
Decision Trees
Predictive model
Based on classification
11. Case Study – Loan Risk Analysis
Problem definition
Mortgage company ACME financial has to predict and
analyze the risk associated with the Applicants before
approving the loan
Data Collection
Loan application data
Credit history and score data
Data preparation (cleansing)
Clean and categorize data
Building the model
Data mining with decision trees
12. Data Collection
Ap.ID Name Address Income Company Date Hired
1 John Cook San Jose, CA $105,000.00 IBM 03/15/1999
2 Willie Chun Freemont, CA $92,000.00 Cisco 06/19/1998
3 Robbert Gillman Phoenix, AZ $28,000.00 City County 08/23/1990
4 Sam Wong Phoenix, AZ $27,000.00 Racing Co. 06/30/1995
5 Jill will Las Vegas, NV $35,000.00 Undertakers, NULL
INC.
6 Rob Chung New York, NY $75,000.00 Monsters, INC. 12/14/2000
7 Amit Khare Sunnyvale CA $91,000.00 Mysql 04/01/1997
Applicant ID Company Balance
1 LTC Mortgage $400,000.00
1 Visa $15,000.00
2 Bank of America $150,000.00
2 ACME Financial $60,000.00
3 Toon Depot $45,000.00
3 Toon Bank $125,000.00
4 Master Card $54,000.00
5 Financial Aid $60,000.00
6 Toyota Credit $44,000.00
7 Financial Aid $23,000.00
13. Data Preparation
Applicant ID Debt Level Income Level Job > 5 Years
1 High High No
2 High High Yes
3 High Low Yes
4 Low Low Yes
5 Low Low No
6 Low High No
7 Low High Yes
15. Conclusion
The fusion of BI and CRM is creating new
opportunities as well as challenges
Increasingly sophisticated consumers are creating
hyper-competition for businesses
Low brand loyalty among new breed of consumers
highlight the importance of customer centric data
mining
BI and CRM together provides 360-degree view of
customer data
16. References
Books
Data Mining Explained: A Manager's Guide to Customer-Centric
Business Intelligence by Rhonda Delmater and Jr., Monte Hancock
Web Articles and tutorials
An Independent Study in Data Mining http://dataml.net/datamining/
The Data Warehousing Information Center
http://www.dwinfocenter.org
Test Drive Data Mining - SQL Server 2005 tutorial
http://msevents.microsoft.com/CUI/WebCastEventDetails.aspx?
EventID=1032291442&EventCategory=3&culture=en-
US&CountryCode=US