Competitive Advantage with Oracle Business Intelligence Enterprise Edition 11g/Oracle Exadata

Chris Royle (RL Polk) & William Both (Capgemini) Oracle Open World San Francisco, CA, 5 Oct 2011

Competitive Advantage w/ Oracle Business Intelligence Enterprise Edition 11g/Oracle Exadata

Insert "Title, Author, Date"

2 © 2011 Capgemini. All rights reserved.

3

Discussion Topics § Polk introduction § Automotive industry intelligence solution utilizing OBI 11g

§ Key implementation components and lessons learned

§ Addressing performance with Exadata

§ Polk’s experience with Exalytics

§ Questions

4

Automotive Data and Marketing Solutions Leader

Company Overview

•  1870—R. L. Polk & Co. founded

•  1921—Motor Statistical Operations launched

•  1965—Adds monthly statistics on motorcycles, RVs, commercial trailers, boats and business aircraft

•  1971—Overseas expansion

•  1999—CarFax Ownership

•  2000—Focus on automotive only

•  2003—PolkInsight launched

5

Depth of Automotive Data Polk manages a complex set of online vehicle data to support the automotive industry as our core business

§  2.6 Billion Transactions §  Titles §  Registrations §  Sales §  Vehicle Manufacturers §  Financial Institutions

§  500 Million Unique Vehicles §  Passenger §  Commercial

§  Over 118 Million Households, 195 Million Individuals

§  Compiled from 31 different public and self-reported data sources

§  Over 460 Variables of Consumer Data

§  Demographics §  Lifestyles §  Vehicle Intenders §  Internet Shopping

§  Over 17 Million Businesses §  35 million vehicles §  Compiled from 39 different public and

self-reported data source

6

Our Customers

§ All global OEMs (incl. emerging) § Dealers / dealer groups § Aftermarket companies § Finance and insurance companies § Advertising agencies § Media companies § Consulting organizations § Government agencies §  Investment firms § Market research firms § Web portals § Search engines

7

Our Business is Providing BI Solutions as a Service

Market Reporting

Vehicle & Consume

r Data

§  Sales and Registration §  Loyalty & Analytics §  Forecasting §  Aftermarket §  Commercial §  Network Management

"  Analyze Market Position "  Pin-Point Key Markets "  Identify Trending "  Determine Opportunities "  Illustrate Competitive Threats

PolkInsight, Polk’s flagship product, is the industry standard for answering questions about automotive market activity

8

§ Interpret new, used registration, VIO and OEM sales statistics

§ Unique customized definitions of geography and vehicle segments

§ Simple metrics sliced many ways – the vehicle was registered…..

o  By who -- Age, Income, ethnicity, …. o Where -- Do they live, was the dealer, …. o Which -- Vehicle, engine, body style, transmission, ….. o  For what purpose -- Retail, fleet, government, rental, ….

§ Infinitely variable drill path

PolkInsight Product Overview

8

9

§ Oracle 4 node BI application: Discoverer, Oracle Reports, customer query tool integrated through Oracle Portal

§ 6 node (4 dual-core CPUs, 64GB each) 10G RAC/RH4 cluster

§ 2TB warehouse – 4 shared fact tables with a combined total of 1 Billion + rows

PolkInsight Architecture – Before OBI and Exadata

10

Business Intelligence Challenges

§ Users were becoming impatient with the usability of our toolset •  Make it as easy as Excel

§ Users want to see traditionally disparate subject areas on a single dashboard – data mash-ups •  How does the data correlate?

§ Users expect ‘Google’ performance •  Why do I have to wait 30 seconds?

§ Delivery to extended devices •  I want it on my iPad

§  IT resources and activities were too involved in the analytical process •  I have to pay for that change?

Polk selected Oracle’s Exadata and Oracle Business Intelligence (OBI) to help evolve and improve PolkInsight

11

Selecting Our Business Intelligence Software Provider

Custom ?

Know what is important to your users, now and tomorrow. Do multiple PoCs – prove it works!

12

Implementation Highlights •  Clear requirements •  Manageable scope •  Support from BI Consulting

Group (now Capgemini)

PolkInsight 3.0 Portal

Exadata

OBI 11g

Oracle Discover, Reports

Oracle Map Viewer

PolkInsight 3.0

Customers

Polk Advisors

Architecture Highlights •  Highly available, clustered

environment •  Custom flex portal

OBIEE implemented and integrated in 3 months

13

PolkInsight 3.0 Portal with OBI Dashboard Integration

14

Engagement Objectives

§  PolkInsight had been a Customer Delivered solution based upon Oracle Discoverer •  Desire to Maintain similar Report Developer Interface to Discoverer Data Layout •  Not an OBIEE 10g Upgrade, rather a first time implementation •  One of the First Corporations to Implement OBIEE 11g •  Build 11g RPD Using BICG/Capgemini Best Practices

§  Utilize Existing Custom Data Warehouse (PolkInsight) •  Minimize Physical Data Model modifications

§  Create Generic Template Dashboard •  Utilizing BICG/Capgemini Best Practices for Report and Dashboard Design •  Provide Enhancements to Prior reporting capabilities

§  Train Polk Advisors in use of Answers to build Customer Specific Dashboards •  Each Advisor is responsible for Several Auto Manufacturers (Polk Customers) •  Each Customer Specific Dashboard was tailored to specific Customer needs

15

Implementation

§  Actual Design and Build is not overly difficult, rather it is complex and very meticulous work

§  Customer Involvement Earlier was Critical to Success •  Project Team 80/20 Mix •  Polk Advisors Trained and developed Customer facing dashboards (User Ownership)

§  Utilized BICG/Capgemini Best Practices for RPD Design §  Introduce MoM, QoQ and YoY reporting functionality by creating Time Series

Aliases •  MAGO •  QAGO •  YAGO

16

Polk Logical Data Model

§  Four Key Facts •  New Car Registrations •  Used Car Registrations •  Sales Figures (from each Manufacturer) •  Vehicles In Operation (VIO)

§  Principle KPI is Counts of Vehicles §  Several Common Dimensions associated with each Fact §  Limited Dimensions that are specific to only one or another Star §  Data Model Enhancement

•  Addition of a Standard W_MONTH_D table •  Modifications as needed to allow Facts to leverage the Intelligent Key

17

200 + Tables

New & Used Sales Facts 4

Category 23 Geography 38 64

Make & Model 25 21 Parameter 8 Sale Type 14

Ethnic Code 1 1 Misc 9 Total 108 100

18

Logical Data Model Example

19

Logical Data Model Example

20

Logical Data Model Example

21

Physical Layer Best Practices Use of and Naming of Aliases

Sole Metric

Time Series

22

Physical Joins

23

Logical Layer Joins

24

Logical Layer Best Practices

§  Dimensional Hierarchy for Each Logical Dimension Table §  Accurately defined Primary Keys §  Content Levels Set for each Dimension

•  Detail level for Joined •  Total level for Un-Related

25

Logical Dimension Tables §  Frequently multiple Logical Table Sources

§  Both Common as well as Customer Specific flavors or Dimension tables

§  Simple PKs

§  Minimal need for Custom Fields

26

Customer Specific Dimensional Hierarchies §  Each Logical Table has a separate

Dimensional Hierarchy •  Even if it only consists of a Total Level and

a Detail Level

§  Each Customer Specific Hierarchy could differ on Naming conventions as well as number of levels

27

Dimensional Hierarchies (Cont) §  Several Cases where Hierarchies had

multiple branches

§  Customer Specific Preferred Drill Paths

28

Logical Fact

29

Presentation Layer

§  Standardized Naming Convention •  NA •  China

§  Common, Polk Standard Subject Area

§  Customer Specific Subject Areas

30

Subject Area

§  Standard Subject Area Layout for each Customer

§  Includes both Common (Generic Polk) and Customer flavors of attributes where appropriate

§  Standardized Field Naming conventions except where Customer has indicated otherwise

31

Lessons Learned

§  Security Discussion •  Authentication (Oracle SSO) •  Authorization (LDAP - OBIEE User Groups) •  Data Level (TBD)

§  Caching Strategy •  Comprehensive approach to handle diverse customer profiles beyond initial

implementation •  Purge Strategy

§  Organizational BI Ownership •  Content •  Metadata •  Data •  H/W

32

Establishing a Performant Foundation with Exadata

Highlights: §  10x out of the box query improvements §  Less data center space §  Compression resulting in less storage and performance gains

33

§ Our Half Rack Exadata server was Delivered July 2010 •  Oracle techs were on site for 3 days •  Configured the server according to our specifications •  Left us with a fully functional system

§ Our Flagship Product (PolkInsight) was in Production by October •  Most of the time was spent testing the upgrade from 10G RAC to 11G RAC •  Our goal was to migrate the database, without trying to leverage Exadata

specific performance features (we are doing that now) •  Other than a dramatic improvement in performance, the migration was

transparent to our customers

§ By January 2011, all of our Customer Facing BI Apps were running on Exadata •  We have decommissioned quite a bit of existing hardware •  We consolidated a number of stand alone databases in the process

Our Experience with Exadata

34

End User Report Return Time

§ Out of the box improvement

§ Now optimizing to Exadata strengths •  Index removal •  Materialized Views •  Storage Indexes

35

Polk’s Experience with Oracle’s Exalytics

§ Test Pilot Program Overview •  Analysis was performed in Oracle’s lab •  Full copy of Polk’s 2TB production data set, running on an identical

Exadata Half Rack •  The Exalytics box was connected to Exadata through InfiniBand •  Test case were chosen by Polk, targeting underperforming dashboards

and drill paths

§ Performance results •  10.5x average performance gains against Exadata cache misses •  Maximum improvement of 114x on corner case drill paths •  All end user response times within 3 seconds with 87% of all interaction

with less than 2 seconds response time

36

Polk’s Feature Findings

§ Exalytics’ summary aggregate advisor •  Analyzes usage patterns •  Builds appropriate aggregate tables in memory or on Exadata •  Admin must run the analysis tool, set parameters and execute the

creation of the aggregate tables •  This is the secret sauce – more analysis required

§ Interactive visualization and data discovery •  No ‘go’ button, new data selector types •  Type ahead dropdowns •  Removal of complete screen refresh in 11.1.1.6

§ New chart types (not available in the beta release) •  Trellis •  Multi panel view and microcharts •  Wheel chart •  Animation

37

Business Value

For our customers: §  First to know §  Speed §  Most complete view of the industry §  Easy to be the expert

For Polk: §  Improve customer satisfaction §  Generate and retain business

` Enable better decisions to sell more cars, parts

and services

`

Remain the leading provider in Automotive Intelligence

38

Questions?

Insert "Title, Author, Date"

4 © 2011 Capgemini. All rights reserved.

Thank you for attending.

§  Visit the Capgemini Business Information Management booth #313

§  You could win an Apple i-Pad.

§  Visit the booth and register to WIN!!

| Sector, Alliance, Offering

More information

Insert "Title, Author, Date"

5 © 2011 Capgemini. All rights reserved.

About Capgemini

Rightshore® is a trademark belonging to Capgemini

| Sector, Alliance, Offering

More information

Insert "Title, Author, Date"

6 © 2011 Capgemini. All rights reserved.

Please contact: • Chris Royle [email protected] • William Both [email protected] • Ron Lewis [email protected]

www.capgemini.com

The information contained in this presentation is proprietary. ©2011 Capgemini. All rights reserved