This webcast discussed advanced analytics in practice. It began with setting the context for analytics by defining key terms like analytics, business intelligence (BI), and discussing trends like democratizing analytics. It then covered how information management infrastructure is critical for analytics. Advanced analytics strategies were discussed, including predictive analytics, data mining, and examples of their use. Best practices emphasized using templates and pre-configured systems to make analytics more affordable and focusing modeling on business outcomes. The webcast concluded with a question and answer session.
8. Poll Question 1
Q: We’re most interested in advanced analytics for:
•Analyzing customer data
•Finding new product opportunities
•Understanding our risk
•Predicting revenue
•Analyzing supply chains
•Other
9. Smarter Business Intelligence
Advanced Analytics In Practice
David Stodder
Contributing Editor, TechWeb
Perceptive Information Strategies
dstodder@gmail.com
10. Today’s Webcast Discussion
Agility: business context for analytics
Business intelligence and data warehouse key
trends and directions for analytics
Analytics, information management and data
warehousing
Making analytics affordable: templates and pre-
configured systems
Analytics: predictive analytics and data mining
Best practices and conclusion
12/7/2010 Copyright (c) David Stodder
11. To Be Agile, Be Aware: The Basis of
Competition Today
•Informed and proactive: Can anticipate and
react in a coordinated fashion
• Competing in real time: The strategic and
tactical edge only lasts so long
• Need to realize value from every customer,
partner
and process
• Optimize to reduce
latency, cut costs and
improve performance
12/7/2010 Copyright (c) David Stodder
12. Agility and Awareness: Information
Systems Support – or Thwart?
Danger that traditional information systems can
thwart agility
Standard, siloed reports can deliver incomplete
and inaccurate views of a multi-channel world
Known problems, known solutions: What about
the unknown?
Smart devices, smarter customers: Keeping pace
with rising intelligence
12/7/2010 Copyright (c) David Stodder
13. Business Intelligence and Data
Warehouse: Success Factors
How good is the information?
- Accuracy and quality; need it now, but need it right
- Comprehensiveness; all relevant sources included? In search
of the single view
What can I do with the information?
- Timely data is good, but do users understand it?
- Think dynamically: Continuous people and process
improvement
How can I profit from (or protect myself with) this
information?
- Operational intelligence is about business innovation
- Risk and regulatory compliance are major drivers
12/7/2010 Copyright (c) David Stodder
14. Data Scarcity: Not the Problem
Many organizations already swimming in an
abundance of data
Focus on gaining higher value from data
Data is already big: “Big Data” focus on
dynamic behavior and velocity of information
(data at rest, in motion)
Info management
challenge: integrating
access to internal and
external data sources
12/7/2010 Copyright (c) David Stodder
15. Business Analytics: Sharpening Focus
On Desired Business Outcome
• When outcome, and course are unknown: Use
information to iterate toward clarity
• Optimization: Use information to ensure no
steps are wasted
• Monitor and measure progress using BI and
performance
management
• Example: customer
loyalty tracking
12/7/2010 Copyright (c) David Stodder
16. Business Intelligence: Can it Take Us
Where We Want to Go?
BI systems primarily provide
quantitative data and tools to
manipulate it for analysis and
to support decision-making
Goal is to deliver
comprehensive views of
business states and directions
Broadening out from
traditional base of analysts
and power users working
with limited data and updating
12/7/2010 Copyright (c) David Stodder
17. Business Intelligence Expectations
Most Important BI Features (3) Reason for Utilizing BI (Top 5)
90%
80%
80%
70%
70%
60%
60%
50% 50%
40% 40%
30%
30%
20%
20%
10%
10%
0%
0%
Fast data Ability to collect Ability to predict
exploration, query and analyze customer behavior,
and analysis operational data in risk or business
capabilities real time outcomes
- InformationWeek Analytics BI and Information Management Survey, September 2010
12/7/2010 Copyright (c) David Stodder
18. Business Intelligence Technology:
Pushing Past Limitations
Historical data is vital, but can limit perspective
and “actionability” of data
Focus on exceptions, not “data dumps”
User requirements always change: Self-service
necessary to free users – and IT – of long, often
unsatisfying development
Search can’t be a stranger: How most people
find information
Collaboration: Decisions made by teams, not
individuals; embed in applications, services
12/7/2010 Copyright (c) David Stodder
19. “Smarter” BI: Collaborative,
Current, Easier to Use and Trusted
• Dashboards display BI/perf. mgmt info, quickly and
easily understood; drill down for anomalies
• Performance management starts to “template”
information around KPIs and metrics
• Collaborative potential: Integrating BI and
collaboration (e.g., IBM
Cognos 10 and Lotus
Connections)
• “Real time” – meaning
what’s important now
(could be real-time data)
IBM Cognos 10 dashboard example
12/7/2010 Copyright (c) David Stodder
20. But Can BI Do Analytics?
Getting beyond reporting: Spreadsheets still most
widely used tool for analysis
Financial analysis #1 reason for utilizing BI: Critical to
expand beyond accounting to support strategic and
operational analysis (including activity-based costing)
Implementation goals for BI (IWK Survey):
◦ Monitor/share metrics: 72%
◦ Analyze customer data to increase sales: 56%
◦ Analyze customer data to retain customers: 53%
12/7/2010 Copyright (c) David Stodder
21. Polling Question #2
Q. What is your top priority with business
intelligence?
- Giving users self-service capabilities for
visualization and drill-down analysis
- Accessing real-time data
- Using BI to improve data quality and
consistency
- Enabling performance management KPIs and
metrics
- Financial reporting and analysis
12/7/2010 Copyright (c) David Stodder
22. Smarter BI: Nowhere Without Info
Management Infrastructure
The “Hercules” of Business
Intelligence
Taking advantage of hardware
advances (virtualization, very
large memory, new chip
designs)
Information integration
Enterprise data warehouse to
support BI; rules to facilitate
information governance
(security, HIPAA, etc.)
Single view of the truth: Data
quality, profiling, discovery
12/7/2010 Copyright (c) David Stodder
23. Relevant, Timely and Reliable Data:
Challenges Remain
12/7/2010 Copyright (c) David Stodder
24. From Traditional to Next-
Generation Data Warehousing
• Serving small, internal user Real-time analytics to improve
communities customer, partner service;
• Built around extraction, real-time event alerting
transformation and loading ETL, MDM and federated
(ETL) information integration
• Historical analysis and Support for performance
reporting management KPIs and
• Batch loading at off hours metrics, dashboards and
• Different systems for simple scorecards
& complex queries (e.g., ODS Continuous updating
and DW) Deployment of appliances in-
• Info delivery not memory analytic apps and
synchronized with processes pre-configured systems
Little external data “Cloud” data services
12/7/2010 Copyright (c) David Stodder
25. Critical Trends in Information
Management: Templates
Rapid development: Using pre-built data
warehouse models, often specific to industry or
application
Models and templates to improve consistency
of implementation
Example: IBM Delivery Accelerators: e.g., retail-
specific template, dashboard and workbench
accelerators, data models, development tools,
processes, predictive modeling
12/7/2010 Copyright (c) David Stodder
26. Critical IM Trend: “Semantic”
Integration Managed Centrally
Relieve BI/analytics tools and users of having to
define data types; reduce “what is a customer?”
chaos
Develop coordinated, accurate and stable
business definitions and semantic meaning:
Master data management
Managing ETL processes more effectively to
reduce cost and delay
Improving data quality: BI fails without it!
12/7/2010 Copyright (c) David Stodder
27. Critical IM Trend: Pre-Configured
Systems and Appliances
“Complex queries” – analytics – the most
frequent reason organizations purchasing
appliances and specialized databases (e.g.,
column-oriented)
Pre-configured to speed deployment
Tight integration
Specialized for analytics
Scalability
IBM Power7 Systems
12/7/2010 Copyright (c) David Stodder
28. In-Memory, In-Database Analytics:
Feeding the Need for Speed
In-memory analytics: Bringing more power and
flexibility to the user’s workstation
In-database analytics: Using the database
system to power analytics; e.g., SAS relationship
with IBM
Real-time “trickle” data feeds and analytics;
“ELT” processing
Embedding BI/analytics with processes
12/7/2010 Copyright (c) David Stodder
29. Polling Question #3
What is your top priority for information
management to support BI/analytics?
- Deploying BI/DW appliances and pre-configured
systems
- Enabling information integration layer (including
ETL, ELT, MDM) to support BI/analytics
- Taking advantage of better hardware (e.g.,
virtualization, blades, faster chips)
- Moving data warehouse systems to the cloud
(public or private infrastructure as a service)
12/7/2010 Copyright (c) David Stodder
30. Analytics: Improving Outcomes
“Simple” Analytics “Advanced” Analytics
- BI “what-if” queries - Optimization
- Accounting-oriented - Activity-based costing
financial analysis modeling and analysis
- Performance - Time series analysis
management metrics and forecasting
- Online analytical - Predictive analytics
processing (OLAP) - Whatever those
- Stuff nontechnical Ph.D.’s are doing
users can do
12/7/2010 Copyright (c) David Stodder
31. Advanced Analytics: Topping
InformationWeek BI “Wish List”
3.8 on scale of 5 (“extremely interested”); one
third rated it 5
Proactive objectives: anticipate demand to
adjust pricing, manufacturing forecasts and
supply chain planning
Know what customers want before they ask for
it – or go to a competitor
Fraud example: Isolate the bad so that good
claims are processed faster
12/7/2010 Copyright (c) David Stodder
32. Analytics: Case Example
Infinity Property and Casualty: auto insurance
for drivers who represent higher than normal
risks and pay higher rates for comparable
coverage
Objective of speeding claims process and
improve efficiency, while cutting fraud and
improving customer satisfaction
“Right-tracking”: claims profiled up front and
sent to appropriate specialists based on claim
characteristics
12/7/2010 Copyright (c) David Stodder
33. Infinity Property & Casualty
Example, Continued
Predictive traits in claims modeled using IBM SPSS
Able to address concerns beyond just fraud
Six months to develop models and rules to
integrate predictions into Infinity’s claim system
Benefits: Used to take 40 days for claims to reach
specialists; now takes 48 hours
Success rate in proving fraud now 87%; company
able to discontinue using third-party firm to
handle collections ($12 mill/yr)
12/7/2010 Copyright (c) David Stodder
34. Predictive Analytics Objectives
Data mining: discovery of previously undetected
patterns and relationships in data
Predictive analytics: applying historical patterns
to predict future outcomes
Statistics (e.g., regression); AI (e.g., neural nets);
hybrid (e.g., decision trees); optimization (e.g.,
Monte Carlo simulation)
Acknowledgements to Eric Siegel, Prediction Impact
12/7/2010 Copyright (c) David Stodder
35. Data Mining: CRISP-DM Cycle
Source: www.crisp-dm.org
12/7/2010 Copyright (c) David Stodder
36. Customer Analytics: Priority Use of
Predictive Analytics
How to increase margin, not just sheer number
of customers?
What are the most effective metrics and
indicators of customer attrition and acquisition?
Predictors are linked directly to business
strategy – to desired business outcomes
Development of incentives programs aimed at
the right customers
12/7/2010 Copyright (c) David Stodder
37. Analytics: Challenges
People and politics: Will they trust the results,
or go with the gut?
Model development – combining predictors –
can be slow, trial-and-error process; models
must be kept up to date
Structured data only half the story: Adding text
analytics and mining to apply quantitative and
linguistic analysis to words and sentiment
12/7/2010 Copyright (c) David Stodder
38. Proactive and Focused on Business
Outcomes: BI and Analytics Together
Anticipate the future, plan how to act with
consistency rather than case-by-case
BI: visualization, alerting and “simple” analytics –
backed by more advanced analytics – to make
information actionable
Linking performance management to analytics
Spreadsheets: Either making them more useful,
or replacing them with better tools
12/7/2010 Copyright (c) David Stodder
39. Making Analytics Affordable: Key
Trends to Watch
Labor and expertise are huge costs: Using
templates for analytics and information
management
Industry models for rapid development
Pre-configured appliances and scalable systems
enable organizations to reduce time and cost
Data services: Analytics in the cloud
Embedding analytics in applications and
processes
12/7/2010 Copyright (c) David Stodder
40. Best Practices for Analytics
Start with your BI platform: Where are users
bumping up against limits for understanding
“unknowns”?
The information management layer is critical:
Analytics thrives on lots of data from multiple
sources for correlation and pattern analysis
Don’t delay: Predictive analytics becoming more
mainstream; key to competitiveness
12/7/2010 Copyright (c) David Stodder
41. Best Practices: Going Forward
Focus modeling on desired business outcomes
Be patient with model development, and be
prepared to update models continuously;
evaluate industry models
Customer analytics and fraud detection: where
depth of experience is greatest
Social network analysis and “Big Data”: valuable
external sources, though in the realm of the
“experts” for analytics use
12/7/2010 Copyright (c) David Stodder
42. Questions and Answers
Chris Murphy
Editor, InformationWeek
David Stodder
Analyst, Research and Writer
Perceptive Information Strategies