mySAP CRM Detailed View

CRM Analytics: Data Mining
(Release BW 3.2)

Content

 Introduction  Predict with:
    Decision Trees Scoring Clustering ABC Classification

 Business Scenarios  Conclusion

© SAP AG 2003, Predict, Ramine Eskandari / 2

CRM Analytics Architecture

© SAP AG 2003, Predict, Ramine Eskandari / 3

Analytical Methods

© SAP AG 2003, Predict, Ramine Eskandari / 4

Ramine Eskandari / 5 . but it is also capable of predicting future trends and behaviors  Data mining allows organizations to make the critical jump from retrospective analysis to prospective decision-making © SAP AG 2003.What is Data Mining? Data mining helps you analyze and understand customer behavior:  Data mining is an analytical approach that looks for hidden data patterns and relationships in large databases  Data mining not only provides insights by analyzing past data. Predict.

Ramine Eskandari / 6 . Predict.Why is Data Mining Important Today? Changes in Technology:  Increased usage of the Internet  Appearance of data warehouses  Increase in computing power  Better modeling approaches © SAP AG 2003.

Why is Data Mining Important Today? Changes in Customer Behavior:  Better informed  More demanding  Increased willingness to switch to competitors  Evolution of needs: more complex. harder to satisfy © SAP AG 2003. Predict. Ramine Eskandari / 7 .

Why is Data Mining Important Today?
Changes in Competition:
 Evolution of strategies:

Mass marketing vs. One-to-One marketing

 Increased competition  Fast-paced environment  Emergence of niche players

© SAP AG 2003, Predict, Ramine Eskandari / 8

Data Mining Methods
Supervised Learning (Predictive):

 Decision Trees

 Scoring (Regression)

© SAP AG 2003, Predict, Ramine Eskandari / 9

Data Mining Methods
Customers

Train

Clustering (Prediction)

Decision Trees

Scoring (Regression)

Exploratory Data
Customer Sandy Jones Peter Maier Sandra Müller Susanne Wolf Oliver Paul ... Age 30 18 45 70 34 ... Income low high medium low medium ... Household Size 1 2 2 1 3 ... Probability of Buying Yes No Yes Yes No ...

New Data
Customer Rob Eagle Eva Smith John White Tom Wolf Oliver Mann ... Age 45 30 20 22 56 ... Income high high medium high low ... Household Size 2 3 4 2 1 ...

t c i d e r P
Probability of Buying Yes No Yes Yes No ...

© SAP AG 2003, Predict, Ramine Eskandari / 10

Data Mining Methods
Unsupervised Learning (Informative):

 Clusterin g  ABC Classification  Association Analysis  Scoring (Weighted Score Tables)

A

B

C

© SAP AG 2003, Predict, Ramine Eskandari / 11

Using and Storing Data Mining Results SAP EP (Enterprise Portal) SAP BW ODS Objects InfoCubes Master Data Hierarchies SAP CRM Business Partner Target Group Online Prediction Data Mining PMML PM ML Intelligence Connector* 3rd Party Tool *Used for Real-Time Analytics © SAP AG 2003. Ramine Eskandari / 12 . Predict.

 Export data mining models definition rather than just their results  Use data mining models defined with third-party tools  Share data mining models easily between different applications and systems  Enable operational processes to use data mining models obtained from multiple sources © SAP AG 2003.PMML Predictive Model Markup Language (PMML) is an XML markup language that provides a standard way to define and share data mining models. Ramine Eskandari / 13 . Predict.Data Mining .

Define Model Step 3.Data Mining – Process Overview Step 1. Transfer Results BW CRM Step 2. Create Query Step 4. Ramine Eskandari / 14 . Run Model © SAP AG 2003. Predict.

Ramine Eskandari / 15 .Content  Introduction  Predict with:     Decision Trees Scoring Clustering ABC Classification  Business Scenarios  Conclusion © SAP AG 2003. Predict.

Predict. Ramine Eskandari / 16 .Decision Trees Benefits  Identify the different factors driving the behavior of your customers  Break down your customer base into segments that share the same behavior New Developments  Share data mining models between different systems using PMML N E W © SAP AG 2003.

Predict. Ramine Eskandari / 17 .Definition A decision tree is a classification approach that uses the values of input variables to predict the value of a categorical variable.  Decision trees allow the user to:     Generate readily-understandable rules with which to predict customer behavior Evaluate the different values of possible outcomes and the probability to reach them Produce a graphical representation of how different factors affect an outcome Create a segmentation scheme based on the decision tree results © SAP AG 2003.Decision Trees .

Decision Trees – Process Overview Customer Income Age 48 22 36 … 34 63 … Occupation Accountant Consultant Accountant … Consultant Accountant … … … … … … … … … Buying Behavior Yes No Yes … ? ? ? New Customers (New Data) Rosa Kanter $ 10 000 Peter Drucker Tom Peters … Kenichi Ohmae Michael Porter … $ 3000 $ 8000 … $ 6500 $ 2000 … Existing Customers (Historical Data) © SAP AG 2003. Predict. Ramine Eskandari / 18 .

Predict. Predict Predicted © SAP AG 2003. Train 2. Ramine Eskandari / 19 . Evaluate Customer … … … … Buying Behavior No Yes Historical Kenichi Ohmae Michael Porter … … … … … … … … … … … … … New Yes 3.Decision Trees – Process Overview Historical Historical New New 92% Accuracy 1.

Definition Age >= 35 <35 Root Node Test  A tree is “grown” from top to bottom  A record sets out from the root node  A test determines which node the record should go to next  All records end up in a leaf node Buy <=$5000 Income >$5000 Decision Node Won‘t Buy Occupation Leaf Nodes Accountant Consultant Buy Won‘t Buy © SAP AG 2003.Decision Trees . Predict. Ramine Eskandari / 20 .

Predict.Decision Trees .Definition Rule: If Age is less than 35 and Income is greater than $5000 and Occupation is consultant. Ramine Eskandari / 21 . then  The customer will buy the product Age >= 35 <35 Buy <=$5000 Income >$5000 Won‘t Buy Occupation Accountant Consultant Won‘t Buy Buy © SAP AG 2003.

2.Decision Trees – Process Overview What are the most important factors I can use to predict my customers’ buying behavior? Purchase Frequency? Income? City? Age? Occupation? Gender? Family Size? Level of Education? Number of Complaints? ? Most Important Least Important 1. 3. Predict. 4. Age Income Occupation Purchase Frequency © SAP AG 2003. Ramine Eskandari / 22 .

Predict. Test and Evaluate Model Predict Discovery Identify Step 3. Identify Customers with High Fraud Probability Transfer Step 5. Transfer Results to CRM Act Step 6. Ramine Eskandari / 23 .Decision Trees – Fraud Detection How Can I Reduce Customer Fraud? Train Step 1. Train Decision Trees on Historical Data Evaluate Step 2. Adapt Payment Option to Type of Customer Application © SAP AG 2003. Apply Model on New Data Step 4.

Decision Trees . Predict.Settings View View Training Training Results Results Train Train Model Model Choose Choose Settings Settings © SAP AG 2003. Ramine Eskandari / 24 .

Ramine Eskandari / 25 .Decision Trees – Training Phase Root Root Node Node Class Class Distributio Distributio n n Decisio Decisio n Node n Node Test Test Criteria Criteria Leaf Leaf Node Node Rule Rule Leading to Leading to Leaf Node Leaf Node Preview Preview Pane Pane © SAP AG 2003. Predict.

Ramine Eskandari / 26 . Predict.Decision Trees – Evaluation Phase Predicted Predicted Class Class Class in Class in Evaluation Evaluation Data Data Evaluatio Evaluatio n n Results Results Error Error Rate Rate © SAP AG 2003.

Predict.Decision Trees – Prediction Phase View Results View Results with a with a Hierarchy Hierarchy Prediction Prediction Results Results Single Single Record Record Prediction Prediction Creation of Creation of CRM Target CRM Target Group Group © SAP AG 2003. Ramine Eskandari / 27 .

Predict.Decision Trees – Prediction Phase Decision Decision Tree Results Tree Results Using Using Hierarchy Hierarchy © SAP AG 2003. Ramine Eskandari / 28 .

Predict. Ramine Eskandari / 29 .Decision Trees – PMML Export Generate PMML Generate PMML with Model with Model Definition Definition Export PMML to Export PMML to Flat File Flat File © SAP AG 2003.

Ramine Eskandari / 30 . Predict.Decision Trees – Online Prediction Decision Decision Tree Web Tree Web Item Item © SAP AG 2003.

Predict.Scoring Benefits  Create a single score for your customers by taking into account different dimensions  Use regression to predict future customer behavior N E W New Developments  Additional statistical measures have been added  Scoring output visualization has been enhanced  Online prediction feature for single records has been added © SAP AG 2003. Ramine Eskandari / 31 .

Ramine Eskandari / 32 . Predict.Scoring .Definition Three different techniques:  Weighted Score Tables  Linear Regression  Nonlinear Regression © SAP AG 2003.

Ramine Eskandari / 33 .Scoring . Predict.Weighted Score Tables Customer Rosa Kanter Peter Drucker Tom Peters Kenichi Ohmae Michael Porter … Income $ 10 000 $ 3000 $ 8000 $ 6500 Age 48 22 36 34 Gender Female Male Male Male City San Diego Los Angeles New York City New York City Boston … … … … … $ 2000 63 Male … … … … … … © SAP AG 2003.

Income 1.Weighted Score Tables + 2.Scoring . Gender Weights 3. Age 6. City 5. Ramine Eskandari / 34 . Predict. Family Size Weighted Score Tables 1 Dimension Customer Value Index © SAP AG 2003. Revenue 6 Dimensions 4.

Weighted Score Tables 2/10 3/10 3/10 2/10 1st Level Weights City Values New York City San Diego Boston Los Angeles … Weights 50 20 80 10 … + Values* 0 $15 000 Revenue Weights 0 100 +…+ Values $0-2000 Income Weights 20 40 70 100 = Overall Score 2nd Level Weights $2001-5000 $5001-8000 $8000+ *Assume Linear Relationship for Values Between 0-15 000 © SAP AG 2003.Scoring . Predict. Ramine Eskandari / 35 .

Predict. Ramine Eskandari / 36 .Regression Regression is a predictive technique that reveals and measures the value of a given variable (where values are continuous) in terms of other variables  Linear Regression y = ax + b  Nonlinear Regression y = f(x) © SAP AG 2003.Scoring .

Ramine Eskandari / 37 .Scoring .Regression Customer … … … … Revenue Rosa Kanter … Peter Drucker … … … … … … … … … … … … $ 5000 $ 10 000 … Historical New 1. Predict Predicted © SAP AG 2003. Train & Evaluate Customer … … … … Revenue Historical Kenichi Ohmae Michael Porter … … … … … … … … … … … … … $ 7800 $ 6000 New … 2. Predict.

Score your Customers Base with Created Model Discovery Identify Step 4. Transfer Results to CRM Act Step 6. Assign Weights to each Dimension Score Step 3.Scoring – Customer Value Score How Valuable are my Customers? Select Step 1. Select Dimensions Affecting Customer Value Weight Step 2. Predict. Send Special Offers to Most Valuable Customers Application © SAP AG 2003. Ramine Eskandari / 38 . Sort Customers Based on their Value Transfer Step 5.

Results Online Online Prediction Prediction Score Score Distribution Distribution © SAP AG 2003.Scoring . Ramine Eskandari / 39 . Predict.

Predict. Ramine Eskandari / 40 .Scoring – Weighted Score Tables Weighting Weighting Scheme Scheme © SAP AG 2003.

Scoring . Ramine Eskandari / 41 .Regression Quality Quality Measures Measures © SAP AG 2003. Predict.

Predict. Ramine Eskandari / 42 .Clustering Benefits  Segment your customer base into homogeneous clusters based on the dimensions of your choice New Developments N E W  Share data mining models between different systems using PMML  Evaluate your results and interpret them more easily by using quality measures  Use additional functions to pre-process your data © SAP AG 2003.

Clustering . Predict.Definition Clustering is a data mining technique that creates groups of records that are:  Similar to each other within a particular group  Very different across different groups  The degree of association between members is measured by all the dimensions specified in the analysis  Clustering helps the user explore vast amounts of data and organize it in a systematic way © SAP AG 2003. Ramine Eskandari / 43 .

Definition Age High Low © SAP AG 2003.Clustering . Predict. Ramine Eskandari / 44 High Income .

Definition + Family Size + Clustering + Income Revenue + © SAP AG 2003. Ramine Eskandari / 45 Age .Clustering . Predict.

2-3 Children … Age … Income © SAP AG 2003. Predict.Definition Revenue … City 30-40s. … Select for Marketing Campaign 50s. High Income East Coast.Clustering . $40 000 West Coast. Ramine Eskandari / 46 .

Analyze and Evaluate Results Step 4. Predict. Select Dimensions for Clustering Execute Step 2. Ramine Eskandari / 47 .Clustering – Marketing Campaign How Can I Improve the Efficiency of my Marketing Campaigns? Select Step 1. Run Model on Selected Data Evaluate Discovery Identify Step 3. Identify Valuable Cluster(s) Transfer Step 5. Tailor Campaigns to each Segment Application © SAP AG 2003. Transfer Results to CRM Act Step 6.

Predict.Clustering . Ramine Eskandari / 48 .Settings Select Type of Select Type of Transformation Transformation View Data View Data Distribution Distribution Choose Choose Weights For Weights For Input Fields Input Fields © SAP AG 2003.

Ramine Eskandari / 49 .Results Quality Quality Measures Measures Name Name Clusters Clusters Single Single Record Record Prediction Prediction Select Select Output Output Type Type PMML PMML Export Export Create CRM Create CRM Target Target Group Group Cluster Cluster Size Size Influence Influence © SAP AG 2003. Predict.Clustering .

Clustering – Value Distribution Value Value Distributio Distributio n Screen n Screen Distribution Distribution Within Within Clusters Clusters © SAP AG 2003. Ramine Eskandari / 50 . Predict.

Ramine Eskandari / 51 .Clustering – Attribute Charts Attribute Attribute Charts Charts Screen Screen Total Total Percentages Percentages Cumulated Cumulated Percentages Percentages © SAP AG 2003. Predict.

Predict.Clustering – Quality Measures Attribute Attribute Value Index Value Index Quality Quality Measures Measures © SAP AG 2003. Ramine Eskandari / 52 .

Ramine Eskandari / 53 . Predict.Clustering – Attribute Value Index Number Of Records Family Size + + + + + + + 1 2 3 4 Family Size Value 1 2 3 4 Index -2 +2 0 0 50 + Distribution: Overall Cluster 1 Mainly Married Couples Are In Cluster 1 © SAP AG 2003.

Clustering – Online Prediction Clustering Clustering Web Item Web Item © SAP AG 2003. Ramine Eskandari / 54 . Predict.

Association Analysis Benefits  Identify cross-selling opportunities  Optimize the bundling of your products and maximize revenue New Developments  Compound characteristics can be used in defining model N E W © SAP AG 2003. Predict. Ramine Eskandari / 55 .

etc. Predict.Definition Association Analysis uncovers the hidden patterns. correlations or casual structures among a set of items or objects. products.Association Analysis ..) Group different items by affinity Create readily-understandable rules describing customer buying behavior Organize web pages in order to optimize user accessibility © SAP AG 2003.  It allows the user to:     Understand and quantify the relationship between different items (e. Ramine Eskandari / 56 .g. click stream..  It is typically used for Market Basket Analysis (MBA).

Ramine Eskandari / 57 . Predict.Association Analysis .Definition Bundles Association Analysis Customer Transactions © SAP AG 2003.

Ramine Eskandari / 58 Leading Item Depending Item Confidence Support . he will also be likely to buy a printer 80% of the time. Predict.Definition Rule When a customer buys a desktop computer.Association Analysis . This behavior can be found in 30% of all purchases Desktop computer Printer 80% 30% © SAP AG 2003.

he will be 2 times more likely to buy a printer © SAP AG 2003. Predict. Ramine Eskandari / 59 .Definition When a customer buys a desktop computer.Association Analysis . he will also be likely to buy a printer 80% of the time A customer buys a printer 40% of the time Lift: 80/40=2 When a customer buys a desktop computer.

8 Confidence: 100% Support: 10% © SAP AG 2003. Ramine Eskandari / 60 Confidence: 80% Support: 30% .4 Importance of Support X2 Probability: . Predict.Definition Importance of Lift Probability: .Association Analysis .

Cross-Sell Products to Customers Application © SAP AG 2003. Analyze and Evaluate Results Discovery Select Step 4. Predict. Select Rules with High Confidence and Support Transfer Step 5. Run Association Analysis Model Evaluate Step 3. Select Transactional Data from Customers Execute Step 2.Association Analysis – Cross-Selling Opportunities How Can I Sell More to my Customers? Select Step 1. Ramine Eskandari / 61 . Transfer Results to CRM Act Step 6.

Settings Settings Settings Define a Training Define a Training Source for the Source for the Data Data © SAP AG 2003. Ramine Eskandari / 62 .Association Analysis . Predict.

Association Analysis . Predict.Output Settings Define Filters Define Filters for the Data for the Data Output Output © SAP AG 2003. Ramine Eskandari / 63 .

Ramine Eskandari / 64 .Association Analysis .Results Export Rules Export Rules to CRM to CRM Leading Item Leading Item Depending Depending Item Item © SAP AG 2003. Predict.

Predict.ABC Classification Benefits  Classify your customers into different segments based on their value to your business New Developments NE W  ABC Classification output visualization has been enhanced  Classification can be done with grouping variables  Compound characteristics can be used in defining models © SAP AG 2003. Ramine Eskandari / 65 .

ABC Classification . or products) based on a particular measure (such as revenue or profit)  Two different approaches for classification:   Define intervals for classification criteria Define intervals for classified object  Absolute values or cumulated percentages can be used to create these intervals © SAP AG 2003. Ramine Eskandari / 66 . employees.Definition ABC Classification is used to classify objects (such as customers. Predict.

Ramine Eskandari / 67 . Predict.ABC Classification – Gain Chart % of Revenue 100 80  “A” Class: 20% of customers generate 40% of revenue  “B” Class: Next 30% of customers generate 40% of revenue  “C” Class: Last 50% of customers generate 20% of revenue 60 40 20 A B C % of Customers 20 40 50 60 80 100 © SAP AG 2003.

Ramine Eskandari / 68 Single Model … . Predict.ABC Classification – Example Region California Customer … … … … Revenue … … … … California A B C Massachusetts A B C Massachusetts Customer … … … … Revenue … … … … Nevada Customer … … … … Revenue … … … … Nevada … © SAP AG 2003 .

Analyze and Evaluate Results Transfer Step 5. Define Intervals for A. Transfer Results to CRM Act Step 6. B. Predict. Select Transactional Data from Customers Define Step 2. Improve Customer Service for “A” Customers Application © SAP AG 2003 .ABC Classification – Gold Customers How Much Profit are my Most Valuable Discovery Customers Generating for my Business? Select Step 1. Run ABC Classification Model Evaluate Step 4. and C Classes Execute Step 3. Ramine Eskandari / 69 .

ABC Classification – Settings Define Model Define Model Parameters Parameters Choose Field Choose Field Type Type © SAP AG 2003 . Ramine Eskandari / 70 . Predict.

Predict.ABC Classification – Settings Choose Choose Classification Classification Method Method Define Define Intervals Intervals © SAP AG 2003. Ramine Eskandari / 71 .

Predict. Ramine Eskandari / 72 .ABC Classification – Results Choose Value Choose Value for Grouping for Grouping Attribute Attribute © SAP AG 2003.

Predict. Ramine Eskandari / 73 .ABC Classification – Results Gain Chart Gain Chart © SAP AG 2003.

ABC Classification – Web Item ABC Web ABC Web Item Item © SAP AG 2003. Ramine Eskandari / 74 . Predict.

Content  Introduction  Predict with:     Decision Trees Scoring Clustering ABC Classification  Business Scenarios  Conclusion © SAP AG 2003. Ramine Eskandari / 75 . Predict.

Predict. Ramine Eskandari / 76 .1) Customer Retention Decision Trees ABC Classification Clustering A B C At Risk Loyal B A C Goal: Reduce Customer Churn Association Analysis Transfer Knowledge Cross-Sell Rules Marketing Campaign © SAP AG 2003.

Predict.2) Customer Conversion Weighted Score Tables ABC Classification Association Analysis A B C …+…+…=… …+…+…=… …+…+…=… B A C Cross-Sell Rules Goal: Convert B to A Customers Transfer Knowledge Marketing Campaign © SAP AG 2003. Ramine Eskandari / 77 .

Ramine Eskandari / 78 .3) Customer Acquisition Clustering ABC Classification A B C Decision Trees Tr ai n A B C Goal: Acquire New “A” Customers Decision Trees Tr ct ai ne Prospects d “A” Transfer Knowledge Pr ed i Marketing Campaign © SAP AG 2003. Predict.

Content  Introduction  Predict with:     Decision Trees Scoring Clustering ABC Classification  Business Scenarios  Conclusion © SAP AG 2003. Ramine Eskandari / 79 . Predict.

Predict. Ramine Eskandari / 80 .Future Plans Future Developments  Data Mining methods integrated in Analysis Process Designer  Ever closer integration of analytical tools with CRM operational processes  Additional transformations in the Analysis Process Designer  Enhanced visualization for data mining methods  And much more… © SAP AG 2003.

com © SAP AG 2003. Ramine Eskandari / 81 .Further Information  CRM Training Course CR900 and TACRM6 (Partner)  mySAP CRM Analytics SAPNET Alias: CRM-Analytics  Direct contact: CRM Product Management: Ramine.Eskandari@sap. Predict.

MVS/ESA. mySAP. Tivoli®. Massachusetts Institute of Technology. SAP. Citrix®. The information contained herein may be changed without prior notice. R/2. OS/390®. the Citrix logo. Program Neighborhood®. Netfinity®.            © SAP AG 2003. AS/400®. X/Open®. z/OS. Inc. HTML. MarketSet and Enterprise Buyer are jointly owned trademarks of SAP AG and Commerce One. Ramine Eskandari / 82 . Inc. Intelligent Miner.Copyright 2003 SAP AG. AIX®. XML. NT®. All Rights Reserved  No part of this publication may be reproduced or transmitted in any form or for any purpose without the express permission of SAP AG. EXCEL®. Inc. DB2 Universal Database. WebSphere®. SAP Logo. World Wide Web Consortium. AFP. and Motif® are registered trademarks of the Open Group. DB2®. JAVASCRIPT® is a registered trademark of Sun Microsystems. xSeries. MetaFrame®. VideoFrame®. S/390®.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. IBM®. Informix and Informix® Dynamic ServerTM are trademarks of IBM Corporation in USA and/or other countries. JAVA® is a registered trademark of Sun Microsystems. R/3. OS/2®. OSF/1®. UNIX®. ORACLE® is a registered trademark of ORACLE Corporation.. pSeries. PowerPoint® and SQL Server® are registered trademarks of Microsoft Corporation. Predict. MultiWin® and other Citrix product names referenced herein are trademarks of Citrix Systems. Microsoft®. WINDOWS®. Some software products marketed by SAP AG and its distributors contain proprietary software components of other software vendors. XHTML are trademarks or registered trademarks of W3C®. OS/400®. mySAP. zSeries. used under license for technology invented and implemented by Netscape. DHTML. Word®. iSeries. ICA®. Parallel Sysplex®. All other product and service names mentioned are trademarks of their respective companies. WinFrame®.

Sign up to vote on this title
UsefulNot useful