Академический Документы
Профессиональный Документы
Культура Документы
Detailed View
CRM Analytics:
Data Mining
(Release BW 3.2)
Content
Introduction
Predict with:
Decision Trees
Scoring
Clustering
ABC Classification
Business Scenarios
Conclusion
Data mining not only provides insights by analyzing past data, but it is
also capable of predicting future trends and behaviors
Changes in Technology:
Better informed
More demanding
Changes in Competition:
Evolution of strategies:
Mass marketing vs. One-to-One marketing
Increased competition
Fast-paced environment
Decision Trees
Scoring
(Regression)
Customers
Scoring (Regression)
td
ic
re
P
Exploratory Data New Data
Customer Age Income Household Size Probability of Buying Customer Age Income Household Size Probability of Buying
Sandy Jones 30 low 1 Yes Rob Eagle 45 high 2 Yes
Peter Maier 18 high 2 No Eva Smith 30 high 3 No
Sandra Müller 45 medium 2 Yes John White 20 medium 4 Yes
Susanne Wolf 70 low 1 Yes Tom Wolf 22 high 2 Yes
Oliver Paul 34 medium 3 No Oliver Mann 56 low 1 No
... ... ... ... ... ... ... ... ... ...
Clusterin
g
A B C
ABC
Classification
Association
Analysis
Partner
Target
Group
Intelligence Connector*
Export data mining models definition rather than just their results
BW CRM
Introduction
Predict with:
Decision Trees
Scoring
Clustering
ABC Classification
Business Scenarios
Conclusion
Benefits
Break down your customer base into segments that share the
same behavior
New Developments
… … … … … …
Historical Historical
1. Train 2. Evaluate
Customer … … … … Buying
Behavior
Historical
Kenichi … … … … No
Ohmae
Michael … … … … Yes
Porter New
… … … … … Yes
3. Predict
Predicted
Accountant Consultant
Leaf Nodes
Age Rule:
If Age is less than 35 and
>= 35 <35
Income is greater than $5000 and
Occupation is consultant, then
Buy Income
Accountant Consultant
?
What are the most important factors I can use to predict
my customers’ buying behavior?
Most Important
1. Age
2. Income
3. Occupation
Least Important 4. Purchase Frequency
Train
Evaluate
Step 1. Train
Decision Trees Predict
Step 2. Test
on Historical Identify
and Evaluate Step 3. Apply Model
Data
Model on New Data Step 4. Identify
Customers with High
Fraud Probability
Application
Transfer Act
Choose
Choose
Settings
Settings
View
View Train
Train
Training
Training Model
Model
Results
Results
Root
Root
Node
Node
Class
Class
Distributio
Distributio
nn
Decisio
Decisio
nn Node Test
Test
Node
Criteria
Criteria
Leaf
Leaf
Node
Node
Rule
Rule
Leading
Leading to
to
Leaf Node
Leaf Node
Preview
Preview
Pane
Pane
SAP AG 2003, Predict, Ramine Eskandari / 25
Decision Trees – Evaluation Phase
Predicted
Predicted
Class
Class Class
Class in
in
Evaluation
Evaluation
Data
Data
Evaluatio
Evaluatio
nn
Results
Results
Error
Error
Rate
Rate
Single
Single
Record
Record
Prediction
Prediction
Creation
Creation of
of
CRM Target
CRM Target
Group
Group
Generate
Generate PMML
PMML
with
with Model
Model
Definition
Definition
Export
Export PMML
PMML to
to
Flat File
Flat File
Benefits
Weighted Score
Tables
Linear
Regression
Nonlinear
Regression
… … … … … …
3. City
Weights
4. Gender
5. Age
6. Family Size
-
Weighted Score
Tables
1 Dimension
Overall
City + Revenue +…+ Income =
Score
Values Weights
Values Weights Values* Weights 2nd Level
New York City 50 0 0 $0-2000 20 Weights
San Diego 20 $15 000 100 $2001-5000 40
Boston 80 $5001-8000 70
*Assume Linear
Los Angeles 10 Relationship for Values $8000+ 100
Between 0-15 000
… …
Linear Regression
y = ax + b
Nonlinear Regression
y = f(x)
Customer … … … … Revenue
Peter … … … … $ 10 000
Drucker
… … … … … … New
1. Train
& Evaluate
Customer … … … … Revenue
Historical
Kenichi … … … … $ 7800
Ohmae
Michael … … … … $ 6000
Porter New
… … … … … …
2. Predict
Predicted
Select
Weight
Step 1. Select
Dimensions Score
Step 2. Assign
Affecting
Weights to Step 3. Score your Identify
Customer Value
each Customers Base with
Dimension Created Model Step 4. Sort
Customers Based on
their Value
Transfer Act
Application
Step 5. Transfer Step 6. Send Special
Results to CRM Offers to Most
Valuable Customers
Score
Score
Distribution
Distribution
Weighting
Weighting
Scheme
Scheme
Quality
Quality
Measures
Measures
Benefits
N
New Developments E
W
Share data mining models between different systems using
PMML
Clustering helps the user explore vast amounts of data and organize
it in a systematic way
Age
High
Income
Low High
+ +
Clustering
Income Revenue
+ Age
SAP AG 2003, Predict, Ramine Eskandari / 45
Clustering - Definition
Revenue 30-40s,
… City $40 000
West Coast,
…
Age
…
Income
Select
Select Type
Type of
of
Transformation
Transformation
View
View Data
Data
Distribution
Distribution
Choose
Choose
Weights
Weights For
For
Input Fields
Input Fields
Create
Create CRM
CRM
Select
Select Target
Target
Output PMML
PMML Group
Output Group
Type Export
Export
Type
Cluster
Cluster
Size
Size
Influence
Influence
Distribution
Distribution
Within
Within
Clusters
Clusters
SAP AG 2003, Predict, Ramine Eskandari / 50
Clustering – Attribute Charts
Attribute
Attribute
Charts
Charts
Screen
Screen
Cumulated
Cumulated
Percentages
Percentages
Total
Total
Percentages
Percentages
Attribute
Attribute
Value
Value Index
Index
Quality
Quality
Measures
Measures
Number Of
Records
Family Size
+
Value Index
+ +
+ 1 -2
2 +2
50 + + 3 0
4 0
+
1 2 3 4
Family Size
Distribution:
Mainly Married Couples
Overall Are In Cluster 1
Cluster 1
Benefits
New Developments
N
E
W
Association
Analysis
Customer
Transactions
Lift: 80/40=2
Probability: ,4 X2 Probability: ,8
Importance of Support
Select
Execute
Step 1. Select
Transactional Evaluate
Step 2. Run
Data from
Association Step 3. Analyze and Select
Customers
Analysis Model Evaluate Results
Step 4. Select Rules
with High Confidence
and Support
Settings
Settings
Define
Define aa Training
Training
Source
Source for the
for the
Data
Data
Define
Define Filters
Filters
for
for the
the Data
Data
Output
Output
Export
Export Rules
Rules
to
to CRM
CRM
Depending
Depending
Leading
Leading Item
Item Item
Item
Benefits
% of Revenue
100
“A” Class:
20% of customers
80
generate 40% of revenue
“B” Class:
60 Next 30% of customers
generate 40% of revenue
40
“C” Class:
Last 50% of customers
generate 20% of revenue
20 A B C
% of Customers
20 40 50 60 80 100
Region
California California
Customer Revenue
… … A
… …
… …
B
… … C
Massachusetts Massachusetts
Customer Revenue
… …
… … A B C
… …
… …
Nevada Nevada
Customer Revenue
… …
… …
… …
… …
… Single Model …
SAP AG 2003 , Predict, Ramine Eskandari / 68
ABC Classification – Gold Customers
How Much Profit are my Most Valuable Discovery
Customers Generating for my Business?
Select
Define
Define Model
Model
Parameters
Parameters
Choose
Choose Field
Field
Type
Type
Choose
Choose
Classification
Classification
Method
Method
Define
Define
Intervals
Intervals
Choose
Choose Value
Value
for
for Grouping
Grouping
Attribute
Attribute
Gain
Gain Chart
Chart
Introduction
Predict with:
Decision Trees
Scoring
Clustering
ABC Classification
Business Scenarios
Conclusion
A B C
At Risk B
A
Loyal
C
Association
Analysis
Transfer Cross-Sell
Knowledge Rules
Marketing Campaign
Association
Weighted Score Tables ABC Classification Analysis
A B C
…+…+…=… B
Cross-Sell
…+…+…=… A Rules
…+…+…=…
C
Transfer
Knowledge
Marketing Campaign
A B C
Tr
ai
n
A
B
C
Pr
ed Decision Trees
i
ct Tr
Transfer ai
ne
Knowledge Prospects d
“A”
Marketing Campaign
Introduction
Predict with:
Decision Trees
Scoring
Clustering
ABC Classification
Business Scenarios
Conclusion
Future Developments
Direct contact:
CRM Product Management:
Ramine.Eskandari@sap.com
No part of this publication may be reproduced or transmitted in any form or for any purpose without the express
permission of SAP AG. The information contained herein may be changed without prior notice.
Some software products marketed by SAP AG and its distributors contain proprietary software components of
other software vendors.
Microsoft®, WINDOWS®, NT®, EXCEL®, Word®, PowerPoint® and SQL Server® are registered trademarks of
Microsoft Corporation.
IBM®, DB2®, DB2 Universal Database, OS/2®, Parallel Sysplex®, MVS/ESA, AIX®, S/390®, AS/400®, OS/390®,
OS/400®, iSeries, pSeries, xSeries, zSeries, z/OS, AFP, Intelligent Miner, WebSphere®, Netfinity®, Tivoli®,
Informix and Informix® Dynamic ServerTM are trademarks of IBM Corporation in USA and/or other countries.
ORACLE® is a registered trademark of ORACLE Corporation.
UNIX®, X/Open®, OSF/1®, and Motif® are registered trademarks of the Open Group.
Citrix®, the Citrix logo, ICA®, Program Neighborhood®, MetaFrame®, WinFrame®, VideoFrame®, MultiWin® and
other Citrix product names referenced herein are trademarks of Citrix Systems, Inc.
HTML, DHTML, XML, XHTML are trademarks or registered trademarks of W3C®, World Wide Web Consortium,
Massachusetts Institute of Technology.
JAVA® is a registered trademark of Sun Microsystems, Inc.
JAVASCRIPT® is a registered trademark of Sun Microsystems, Inc., used under license for technology invented
and implemented by Netscape.
MarketSet and Enterprise Buyer are jointly owned trademarks of SAP AG and Commerce One.
SAP, SAP Logo, R/2, R/3, mySAP, mySAP.com and other SAP products and services mentioned herein as well as
their respective logos are trademarks or registered trademarks of SAP AG in Germany and in several other
countries all over the world. All other product and service names mentioned are trademarks of their respective
companies.