elisabete silva sogrupo - sascommunity · 1 elisabete silva, henrique nicola sogrupo - si cgd group...
TRANSCRIPT
2
Agenda
The requirement...The requirement...
Data Mining as part of Business IntelligenceData Mining as part of Business Intelligence
Data Mining projects at CGD BankData Mining projects at CGD Bank
Next / current projects…Next / current projects…
ConclusionsConclusions
3
The requirement...The requirement...
Don Tapscott & Art Caston, em “Paradigm Shift”
“A pressing reality of the new global environment is theemergence of a new era of competition.
Competition is arising not only from traditional adversariesin traditional markets, or from new entrants to a specific
industry or economic sector, but also from disintegration of barriers to previously insulated and protected markets.”
4
The requirement...The requirement...
Some of the main goals for CGD are:
to be “Customer Oriented”predict Customer needssolve Customer problems... not just “answer their questions”
To satisfy these goals it’s necessary:
to improve Knowledge about Customers (using Data Mining)to invest in Data Quality Projects (to have correct information)
The highest priorities for CGD Group are:
Increase Profitability (Customers, Transactions, Products,...)Decrease Risk (Customers, Credit products,...)
5
Business Needs: Improving Information Business Needs: Improving Information Quality of CGD CustomersQuality of CGD Customers
One of the objectives of DMK credit cards area is to increase the number of its customers. The objective was to identify the customer segment at which marketing and sales actions were to be particularly targeted.
An analysis of the profile of credit card customers may help to refine the definition of customers at whom a campaign should be targeted and the most suitable approach to be made – reducing campaign costs and increasing its efficiency.
6
Communications platform
DB2
LANWindows 95
NT Server
TCP/IPSNA
Data Marts
OLAP
Data Mining
Ad-Hoc Queries,Reporting
• Temporary Data• WEB Server• Data Mining Server• DB2 Connect Server• ...
Data Mining as a part of Business IntelligenceData Mining as a part of Business Intelligence
Information System
DB2
Data Marts
External Data
DW
OS390 Servers
Operational Systems
7
«Mastering Data Mining»«The art and science of customer relationship management»
Michael J. A. BerryGordon S. Linoff
“Data Mining is the process of exploration and analysis, by automatic or semiautomatic means, of large quantities ofdata in order to discover meaningful patterns and rules.”
Data Mining Data Mining -- What is it?What is it?
8
Data Mining Data Mining -- What is it?What is it?
Data Mining is the process used to select, exploit and model large volumes of data with a view to identifying previously unknown consistent patterns or systematic relationships between variables, with a view to generating major value to an organisation’s business oractivities.
9
Data Mining Data Mining -- What is it?What is it?Name Age Scoring Own
Credit CardJohnMaryJuliaPamelaMichaelTomGeorgeCindy
6586455535826542
1227293632474349
YNNYYNNY
Y
Y
YYN
N N
N
10 20 30 40 50Scoring
20
40
60
80
Age
Scoring10 20 30 40 50
20
40
60
80
Age
Y
Y
Y
Y
N
N
N
N
Scoring10 20 30 40 50
20
40
60
80
Y
Y
Y
Y
N
N
N
N
Age
10
Data Mining Data Mining -- What is it?What is it?
10 20 30 40 5020
40
60
80
Y
Y
Y
Y
N
N
N
N
Scoring
Age
Scoring >=20
Age < 60 Age >= 60
Scoring<30
Scoring >=30
Scoring <20
YN Y N
Scoring10 20 30 40 50
20
40
60
80
Y
Y
Y
Y
N
N
N
N
Age
10 20 30 40 5020
40
60
80
Y
Y
Y
Y
N
N
N
N
Age
Scoring
Decision Tree
11
Initial Work Initial Work -- Extraction and Analysis of Data Extraction and Analysis of Data QualityQuality
Definition of data set to be analysed;
Extraction of data required for the defined analyses from legacy systems (operational systems);
Loading of extracted data in SAS®;
Definition of rules for the preparation of data and new files;
Analysis and validation of data.
12
Work Performed Using SAS Enterprise Work Performed Using SAS Enterprise MinerMinerTMTM
Training on use of tool;
Installation of tool on NT server and team PC workstations;
Study of profile of CGD customers: differentiating characteristics between credit card and non credit card customers
Creation of a credit card acquisition propensity model (standard model);
Creation of a “classic” (i.e. standard) credit card acquisition propensity model (Classic model);
Creation of a gold credit card acquisition propensity model (Gold model);
Creation of a gold credit card acquisition propensity model for classic credit card customers (upselling model);
Creation of customers segments.
13
Several Tool OutputsSeveral Tool Outputs
AGE on 30 June 1999
Credit Card Indicator
14
Cluster 5Cluster 5
Cluster 1Cluster 1
Cluster 0Cluster 0
Cluster 3Cluster 3
Cluster 6Cluster 6Cluster 2Cluster 2Cluster 4Cluster 4
InactiveInactive Traditional SaversTraditional Savers
Active InvestorsActive Investors
TopTop--toptopHigh SpendersHigh Spenders
BorrowersBorrowersDebit Card UsersDebit Card Users
Cluster 0 – Inactive (medium %)
low asset levels - 57% of customers have less than PTE 100,000 in investments;
only 4% have residential mortgage loans and only 2% have taken out personal loans;
customers with weak ties to CGD, having little more than a current account.
Cluster 1 – Traditional savers (medium %)
39% of customers are retired and 18% are middle age (50-65);
few of these customers (3% and 2% respectively) use debit and credit cards;
99% of these customers have more than PTE 200,000 in Deposits and 37% have more than PTE 2,000,000;
they do not use Credit products;they prefer term deposits – more traditional customers
– 76% have more than one term deposit account;4% are classified as preferred customers.
Cluster 2 – Active Investors (low %)
around 66% of customers are over 50;98% have deposits amounts more than PTE
2,000,000;these customers do not use credit but have
savings accounts, some financial products ranging from the most traditional to the most sophisticated.
Cluster 3 – Top-top (very low%)
45% of customers are middle aged (50-65);94% have deposits amounts more than PTE
2,000,000;83% of these customers have more than one current
account and 22% have overdrafts;these customers are investors - 68% have
securities portfolio, 61% have one or more term deposit accounts and 56% have savings accounts;
86% of these customers have taken out a residential mortgage loan;
18% are classified as preferred customers;these are sophisticated, investing customers who
make use of a wide range of the bank’s products.
Cluster 4 – Debit Cards Users (low%)82% of these customers have investments of less
than PTE 200,000 although they have savings accounts;
they are transactional customers – 11% areCaixaDirecta subscribers, 79% have debit cards and 34% have credit cards.
Cluster 5 – High Spenders (low %)most of these customers are aged between 25-65 and
belong to the working population (82%);these customers have high credit levels and do not
make investments - 6% have recorded payment defaults;
these are customers with a high level of credit product propensity – 91% have taken out residential mortgage loans, 25% have taken out personal loans and 24% have overdrafts.
Cluster 6 – Borrowers (low %)these are customers with loan product propensity -
76% have taken out residential mortgage loans, 31% have taken out personal loans and 26% have overdrafts;
these customers are moderate in their use of credit- the payment default rate is not particularly high.
Results by SegmentResults by Segment
15
Cluster 5Cluster 5
Cluster 1Cluster 1
Cluster 0Cluster 0
Cluster 1Cluster 1
Cluster 0Cluster 0
Cluster 3Cluster 3
Cluster 6Cluster 6Cluster 2Cluster 2Cluster 4Cluster 4
InactiveInactive Traditional SaversTraditional Savers
Active InvestorsActive Investors
TopTop--toptopHigh SpendersHigh Spenders
BorrowersBorrowersDebit Card UsersDebit Card Users
Results by SegmentResults by Segment
16
Marketing Campaign
Results by SegmentResults by Segment
17
Extraction from DW corporate customers’ data
Add external data to improve information about these customers
Apply Data Mining techniques to discover and understand customers’ profile: “What’s the profile of a customer that has purchase a corporate product?”
Creation of a Global Industry Offer acquisition propensity model
Credit customers propensity model
Deposits customers propensity model
Potential customers propensity model
Work done in a Corporate Customers’ Project
18
SAS Enterprise Miner screen shot:
Work done in a Corporate Customers’ Project
19
Have improved knowledge about corporate customers;
Understand the profile of some corporate customers; the characteristics of customers that have bought the Industry Offer Global product;
Understand the characteristics of customer segments according to the products that they have: current accounts, credits, insurance, ...
Now, CGD Group:
Work done in a Corporate Customers’ Project
20
SAS modules at CGD
SAS ACCESS to DB2
SAS BASE SAS EIS
SAS Enterprise MinerSAS Enterprise Miner
SAS CONNECTSAS CONNECT
NT ServerLANs
Windows 95 e NT
TCP/IPSNA
DB2
21
Conclusions
The quality of information obtained on CGD customers was highly relevant to CGD’s Marketing Department;
It was shown that the use of a data mining tool - SAS Enterprise MinerTM - made it possible to obtain fresh information of vital business interest;
Much of the project’s success was, in part, based on the multi-disciplinary (IT and business) composition of the project team.
The rightinformationto the rightpersonat the right
Therightdecisionat the
time right time
22
Data Mining with the Electronic Channels Department;
The 4th iteration of Data Warehouse is about Risk and Channel Analysis...
The 5th iteration of Data Warehouse is about Profitability...
Goal: a consistent approach to customer servicing, independent of channel utilised.
Next / current Projects...Next / current Projects...
23
Next / current Projects...Next / current Projects...
Click ThroughClick From
Clickstream Analysis
www.cgd.pt
caixadirecta.cgd.ptwww.fidelidade.pt
24
Next / current Projects...Next / current Projects...
Customer Churn (Attrition)
Customer Retention
Customer Acquisition
Customer Profitability
Targeted MarketingCross Selling
Market-basket analysis
Upgrading or Upselling
Fraud Detection
Channel Management
Risk AnalysisAttributing of Credit Limits
Web Mining
25
Data Marts