Adding the Power of Foresight to Business Intelligence

1
Adding the Power of Foresight   to Business
Intelligence

• A Discussion of Predictive Analytics
• Tim Daciuk
• Director, Worldwide Demo Resources
• SPSS Inc.
• Alan Payne
• Manager, Member Research and Development,
• Navy Federal Credit Union

2
Agenda

• About SPSS
• From Business Intelligence to Predictive
  Analytics
• What is Predictive Analytics
• Why does it matter
• Summary

3
Company Fundamentals Company

• Leadership
• Market leader in Predictive Analytics
• 40 years in business
• 250,000 customers
• 1,200 employees
• Global infrastructure (51 of revenue from
  outside NA)
• 300M in annual (2007) revenue
• NASDAQ SPSS

4
A Worldwide Brand
5
Company Fundamentals Customers

• Customers in more than 100 countries
• All 50 U.S. state governments
• 90 of nations top universities
• 95 of the Fortune 1000 companies
• 10 largest pharmaceutical companies
• 85 of top consumer packaged goods companies
• 10 largest market research firms

6
The Data Analysis Timeline
Viability of Predictive Analytics
Business Maturity
The Data Warehouse Years
The Data Harvesting Years
The Data Anticipation Years
Inflexion point
(The age of Predictive Analytics)
(The quest for the predictive enterprise)
(The search for order in the house of data!)
Business Advantage
Start of Market Traction
Time

Phase 2
Phase 3
Phase 4
Phase 5
Phase 1
1994
2000
2003
2007
2010
1997
Dominant Market Demand
Database DW vendors
E-opportunists ERP vendors
Middleware BI vendors
Predictive Analytics BPM-centric vendors
Event Anticipation, Process Simulation KM
vendors
7
The Next Phase of Business Reporting
High
Prediction What Might Happen?
Predictive Analytics
Monitoring Whats Happening Now?
Query, Reporting Search tools
Complexity
Analysis Why did it Happen?
OLAP Visualization tools
Reporting What Happened?
Query, Reporting Search tools
Business Value
Low
High
First Quarter 2007, TDWI Best Practices
Report Predictive Analytics, Extending the Value
of Your Data Warehouse Investment
8
Predictive Analytics (PA) Defined

• Data driven approach to problem solving
• Focused on Business Objectives
• Leverages organizational data
• Uncovers patterns using predictive and
  descriptive techniques
• Uses results to help improve organizational
  performance

9
What Does PA Actually Do?

• Predictive Analytics uses existing data to
• Predict
• Category membership
• Numeric Value
• Group
• Cluster (group) things together based on their
  characteristics
• Associate
• Find events that occur together, or in a sequence
• Find outliers
• Identify cases that dont follow expected behavior

10
Platform for Predictive Analytics
11
Common Applications of PA

• Customer Analytics
• Identify and market to profitable
  customers/prospects
• Identify high value customers for acquisition and
  cross-sell
• Predict likelihood of defection
• Student lifecycle management
• Donor and alumni development
• Fraud and Risk Reduction
• Identify risks
• Identify fraud or suspicious activity
• Process Improvement
• Uncover the factors that lead to product failures

12
Predictive Analytics What It Isnt

• Not a product, a particular piece of software, or
  a given algorithm
• Not a model, segmentation scheme or business
  rules
• Not an end product in and of itself
• Not an SQL query, an OLAP hub, or a BI Dashboard
• Not statistics per se

13
Differences Between BI and PA

• BI supplies the core facts of an organization
• What?
• Reporting tables
• Core business metrics
• Factual reporting
• KPIs

• PA delivers the reasons or drivers of those facts
• Why and How?
• Predictive associations
• Optimized models
• Causal reporting
• KPPs Key Performance Predictors

14
Predictive Analytics Text Added to Data
15
The Predictive Advantage

• Theres analyticsand analytics
• Core Analytics / BI
• Predictive Analytics

16
IDC - Independent Financial Impact Studies

• The median ROI for the projects that
  incorporated predictive technologies was 145,
  compared with a median ROI of 89 for those
  projects that did not.
• Source IDC, Predictive Analytics and ROI
  Lessons from IDCs Financial Impact Study

17
Why Predictive Analysis is Critical
Beforeanalytics
Afteranalytics 21 18 12 10 60
Banner ad click through rates 0.3 Mail
response rates 0.5 Merchandising response
rates 0.2 Conversion rates (post-response)
0.9 Buyer repeat rates 2.0

• - Performance of analytics targeted to certain
  consumers cross-industry and channel, research
  from Forrester, Jupiter, Amazon.com and Ovum (DM
  Review, Feb 11, 2003)

18
Data Heart of the Predictive Enterprise
Customer Contact Channels
Text data Up to 40better predictions

• Attitudinal data
• - Opinions
• Preferences
• Needs
• Desires

• Interaction data
• - Offers
• Results
• Context
• Clickstreams
• Notes

Web data Up to 20better predictions
Website Email Phone Mail Branch ATM Agent Mobile
Attitudes Up to 30better predictions
Marketing Attitudinal Interaction Web Call-center
Operational
Customer View

• Behavioral data
• - Orders
• - Transactions
• Payment history
• Usage history

• Descriptive data
• Attributes
• Characteristics
• Self-declared info
• (Geo)demographics

19
Where PA Fits In Your Organization
Analyze data to provide insight and predict the
future
Recommend the mostappropriate actionto take
Store detailed data on customers, events, etc.
Operational processes and systems
20
Keys to Success

• Define the right strategic objective
• Get the right resources
• Plan, develop and implement the solution
• Socialize the results throughout the organization

21
A Predictive Analytics SuccessNavy Federal
Credit Union

• Alan Payne
• Manager, Member Research and Development
• Navy Federal Credit Union

22
Outline

• Business needs
• Selection process with SPSS
• What we have
• Case looking at satisfaction
• Going forward

23
Business Needs

• 1st Data Analysis
• Utilize a tool to help make sense of the data we
  have and are gathering (exploratory analysis and
  reporting)
• 2nd Prediction
• Begin to utilize the tool and data to begin
  generating predictive analytics (regression,
  trees, clusters etc.)
• 3rd Action
• Initiate predictive analytics in a production
  environment (response, segmentation, and
  predictive models)

24
Business Needs

• 1st Data Analysis
• Utilize a tool to help make sense of the data we
  have and are gathering (exploratory analysis and
  reporting)
• 2nd Prediction
• Begin to utilize the tool and data to begin
  generating predictive analytics (regression,
  trees, clusters etc.)
• 3rd Action
• Initiate predictive analytics in a production
  environment (response, segmentation, and
  predictive models)

25
Selection Process with SPSS

• 1st Software
• I needed a tool that could be used out of the box
• 2nd Company
• I required a company that would respond to me
• 3rd Usage
• I had to have a product that staff at various
  levels of knowledge could use immediately
• 4th Scalability
• The product had to scale as we move from
  understanding data to implemented predictive
  analytics
• 5th Service
• World-class training

26
What We Have
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
27
What We Have
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
28
What We Have
Clementine
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
29
What We Have
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
30
A Case Study Looking at Satisfaction
31
A Case Study Looking at Satisfaction
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
32
A Case Study Looking at Satisfaction
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
33
A Case Study Looking at Satisfaction
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
34
A Case Study Looking at Satisfaction
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
35
A Case Study Looking at Satisfaction
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
36
A Case Study Looking at Satisfaction
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
37
Going Forward

• Tools are in place
• Able to expand our predictive analytics
• Infrastructure is in place to begin pushing
  models
• Staffing development is in place (entry to
  expert)

Clementine
Text Analysis/Complex Samples
SPSS Statistics Regression/Trees/other models
SPSS Statistics - Tables
38
Results

• Board approved significant internal investments
• Going from 100 branches to 160 in three years
• 2 new call centers (Winchester, VA and Pensacola,
  FL)
• Over 1,500 new employees
• Statistically significant improvements in access
  to Navy Federal by members
• Statistically raised member satisfaction
• 2 SD above top banks (ACSI)
• Developed custom products and services
• Increase in holdings
• 25 billion in Oct. 2005 to almost 36 billion
  today

39
Other Examples of the Success of Predictive
Analytics
40
Natexis Assurance Increases Productivity

• Background
• Insurance division of the French Groupe Banque
  Populaire
• Business goals
• Increase effectiveness of marketing and account
  managers in branches
• Solution
• Implemented Predictive Marketing to target high
  value customers in marketing campaigns, and
  generate targeted leads for advisors

• 46 cost reduction on investment product campaign
• 55 more policies sold
• 109 more revenue, since customers made higher
  investments
• Bottom line1.6 Million Euro additional revenue
  on a single campaign

41
Fortis Bank Increases Conversion Rates

• Background
• International financial service provider in
  banking and insurance. Among the twenty biggest
  financial institutions in Europe.
• Market capitalization EUR 39 billion
• Business goals
• Build success in implementing targeted direct
  marketing campaigns
• while reducing the total campaign circulation
  and rise in conversion
• Solution
• Create better target group selections and more
  relevant offers that fit in better with the
  wishes and expectations of individual customers.

Results

• Modelling gives the marketing team the
  possibility of predicting the effectiveness of
  campaigns in advance
• Reduction in the total campaign volumes by 20
• Increase in conversion of 50 to 75

42
Richmond Police Department

• Background
• Established 1807 - one of the first law
  enforcement agencies in US
• 12 policing sectors, serving 200,000
• Business goals
• Proactively reduce crime by using data to predict
  staff likely hot spots
• Present officers with real-time data displayed in
  geographic maps
• Reduce staffing costs
• Solution
• Merge and analyze data resources (weather,
  events)
• Build model to characterize and predict criminal
  activity, incl. locales/times
• Display results in an interactive GIS

Results

• Proven model identifies actionable crime patterns
• Reduced crime
• Reduced staffing costs
• E.g. New Years Eve (2003/4)
• 49 reduction in random gunfire incidents
• 246 increase in weapons seized

43
Questions
44
What Have We Covered?

• Predictive Analytics Defined
• What is the difference between BI and predictive
  analytics
• Data Analysis Timeline
• Where is the evolutionary place of predictive
  analytics
• The Methodology for Predictive Analytics
• How to bring a PA mindset to business problems
• How to Move from Research to Action
• How to apply PA insight to proactive initiatives

45
What Do We Know?

• Predictive analytics does not replace BI
• Predictive analytics leverages data investments
  already in place
• Predictive analytics unlocks the hidden potential
  residing in data stores
• Navy Federal has shown the power of predictive
  analytics put into action
• http//www.cutimes.com/article.php?article40946

46
Contact Information

• Visit www.spss.com for general information
• Michael Doane mdoane_at_spss.com