Está en la página 1de 44

Using Predictions to

Power the Business


Wayne Eckerson
Director, TDWI Research
May 4, 2010
Sponsor
Speakers

Wayne Eckerson Rick Pechter


Director, Senior Director, Advanced
TDWI Research Analysis & Data Mining,
MicroStrategy
Two Strains of Analytics
Exploration & Analysis Prediction & Optimization

Navigate historical data Model historical data


Top Down Bottom Up
Deductive Inductive
Query tools Data mining tools

4
Definition
A set of tools and mathematical techniques that
identify patterns and relationships in large volumes
of data that can be used to describe or predict
behavior that drives business value.
Terms:
- Knowledge discovery
- Data mining
- Advanced analytics
- Predictive analytics

5
Analytical Challenges
• Culture
– Fact-based decision
making driven from top
• Skilled people
– Who know the business,
process, data, and tools
• High quality data
– Integrated, consistent,
trustworthy data
• Tools
– More than spreadsheets
and desktop databases

6
Meet Your Analysts

Sam Beth Ann Bruce


“Super User” “Business “Analytical “Business
Analyst” Modeler” Manager”

Often, one person

7
What do analysts do?
1. Collect Data 2. Integrate Data 4. Present Data

Access, tools, schema, Excel, PowerPoint


Link, merge, summarize,
definitions, normalize, fix

3. Analyze Data 5. Distribute Data

Slice, dice, sort, group, rank, Email, Thumb drives


visualize, correlate, model

8
10 Ways to Empower Analysts
10. Provide ready access to corporate data
9. Consolidate detailed data in one place
8. Integrate and normalize the data
7. Standardize the metadata
6. Clean the data
5. Deliver the data in timely fashion
4. Provide self-service to casual users
3. Give them access to analytical sandboxes
2. Make them part of the BI working committee
1. Don’t dictate what tools they can use

9
Business Analyst Tools
• Excel, Access, SQL
• Managed Excel
• BI search
• Ad hoc reporting tools
• Mashboards
• OLAP
• Visual discovery tools
• Planning tools
• Text analytics
• Analytic engines
• Data mining software
• In-database analytics Analyst Toolbox(?!)

10
High In-database
SQL ROLAP Analytic engines
analytics

Dimensional Text Analytics


analysis
Analytic
Data Volumes

modeling
Ad hoc for
power users
Data mining
Managed
Ad hoc Excel*
reporting MOLAP Dimensional
Visual analysis
discovery “What if”
BI search modeling
Planning
Mashboards
Ad hoc for Excel
casual users
Low Calculation Complexity High

(All tools are nearly equally interactive)

11
Analytical Models
• A model describes relationships among
variables
• Models either estimate or classify data values
• A model is a mathematical formula
(Variable * Weighting) + mathematical
operator
Y = W + W X + W X +….+ W X
$6,000

$5,000

$4,000

Formulas draw lines

Sales
$3,000

$2,000

$1,000

through data points $0


$0 $100 $200 $300 $400 $500 $600
Advertising

12
Classification
O O
O X O
X O
X
O O O O
X
X X X O
X
X X
O X X
X X
O O

Decision Tree Cluster


X
O
O O
O
X
O X O
O
X
X O O
O
O O X O
X
X
X
X O X
O X
X O

Logistic Regression Neural Net

13
The Importance of Preparation
• “After finding the right problem to solve,
data preparation is often the key to solving
the problem. It can easily be the difference
between success and failure, between
usable insights and incomprehensible
murk, between worthwhile predictions and
useless guesses.”

Dorian Pyle
“Data Preparation for Data Mining”

14
Applying the Model to New Data
W0 + W1 X1 + W2 X2 +….= Y (Score)
A. Download records to
score to desktop

Y
B. Convert formula to SQL
Y
Y
and run in the database Y
Y
Convert formula to C or SQL
Y

Database W0 + W1 X1 + W2 X2 +….= Y (Score) Database

Traditional Options

15
Analytics Architecture
Desktop Analytics In-Database Analytics
Desktop Desktop
Data exploration
Data preparation [Data preparation]
Data modeling [Data modeling]

Function calls
Scoring code
Data

Data preparation
Data scoring Data modeling
Data scoring

Data Warehouse Data Warehouse

16
How to Bootstrap an Analytics Practice
Launch the Practice
1. Find an analyst
2. Find a sponsor
3. Start small

Deliver Useful Results


4. Don’t get uppity
5. Make it actionable
6. Make it proactive

Sustain the Practice


7. Centralize and standardize data
8. Provide open access to data
9. Offload reporting
10. Centralize analysts

17
I. Launch the Practice

1. Find an Analyst
• Characteristics
– Inquisitive, critical thinker
– Likes to experiment & test
– Persistent Where to find them?
• Knowledge • IT report developers
– Business process • Data analysts
– Company data • MBA graduates
– Databases and systems • Social scientists
• Tools skills • Statisticians
– Excel or Access required
• Six Sigma practitioners
– OLAP and SQL preferred
• Operations research
– SPSS/SAS ideal

18
Analysts vs Non-Analysts
Analysts Non-
Analysts
Willing to push myself to reach challenging work 82% 66%
goals
Ready to put my heard and soul into my work 77% 64%
Really care about the fate of my company 74% 59%
Willing to put in a great deal of effort beyond 70% 51%
what’s expected
Get excited thinking about new ways to do my 67% 50%
job more effectively

Source: Jeanne G. Harris, Elizabeth Craig, and Henry Egan, “How to Engage and Retain
Your Analytical Talent,” Accenture Institute for High Performance, 2009, as quoted in
“Analytics at Work” by Tom Davenport, Jeanne Harris, and Robert Morison, p. 102.

19
I. Launch the Practice

2. Find a Sponsor
• Find an exec who wants to test long-held
assumptions
– “Which clients will likely default on their loan?”
– Does mailing the event brochure 12 weeks out
provide the best lift?”
• Avoid geek speak
– Don’t mention “models,” “statistics,” “analytics,” etc.
– Talk in business terms only
• Pique their curiosity!
– And get permission to proceed….

20
Analytical Leaders
• Create a fact-based decision making culture
• Recruit and nurture other analytical leaders
• Prioritize and fund analytics projects
• Set a hands-on example
• Know the limits of analytics

Adapted from “Analytics at Work” by Tom


Davenport, Jeanne Harris, and Robert Morison

21
I. Launch the Practice

3. Start Small
• Use existing people and tools
– Free up an analyst for several weeks
– Use data mining extensions in your BI tool or
open source data mining tools
• Select the right project (or prototype)
– Small scope with low risk
– But interesting enough to be valuable
– Start with clean, known data sources
• Build on your successes

22
Advanced Analysis & Data Mining
using MicroStrategy

Rick Pechter
Senior Director, Technology Program Management
Advanced Analysis & Data Mining
April 2010

23
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Background: 1st ROLAP Engine
MicroStrategy Overview 1st BI Server
1st BI Web Interface
• Founded in 1989 1st ‘Windows-on-the-Web’ BI
• Largest independent public BI vendor (NASDAQ: MSTR)
• Selected by Forbes as one of the 200 Best Small Companies
• Over 1 million business users at over 3,000 organizations
• Direct operations in 41 cities in 23 countries across the world
• Over 1,800 employees worldwide; 20% are dedicated to R&D
• Over 50 patents pending or issued for MicroStrategy software

Atlanta Denver Seattle Buenos Aires Barcelona Johannesburg Munich Vienna Melbourne
Boston Los Angeles Tampa Mexico City Brussels Lisbon Paris Warsaw Seoul
Charlotte Montreal Toronto Monterrey Cologne London Rome Zurich Singapore
Chicago New York Washington Sao Paulo Copenhagen Madrid Stockholm Sydney
Dallas San Francisco Frankfurt Milan Utrecht Tokyo

24
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Microstrategy Enables Growth Of Analytics

25
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Why Integrate Advanced Analysis into BI?

 “50-75% of the effort is getting to a good dataset”


 Following BI Best Practices puts you well down this path
 This the way to “Empower Analysts” a la Wayne
 Ready Access, Standardized, Clean, Timely, Self-Serve, Open
 Advanced Analysis doesn’t have to be intimidating
 While details are sophisticated, concepts can be intuitive
 “Black boxes” are ok – consider credit risk (FICO) scores
 MicroStrategy has compelling analytical features
 For Statisticians: People who demand analytical breadth & depth
 For “Analytical Amateurs”: People who love data & getting results

Now is the time to integrate


Advanced Analytics & BI
26
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
The Platform Is The Key To Supporting Data Mining

Dynamic Grid Intelligence Cube Document Alerts &


Reports Reports Reports Notifications Report Data
Generates Set
Data Set for Model Building
Grids & OLAP Report Distribution Data Mining
Graphs Services Services Services Model Building

User-based Metadata Objects and Security Layer

Persist the Model in Metadata Y=B0+B1X1+B2X2


As a Predictive Metric

• Usable in other objects, such as filters,


Global Multi-dimensional Model Windows custom groups, prompts, thresholds, even
Unix
other metrics
Analytical Engine Linux
• All the security, usability and other features
ROLAP Engine automatically apply
• Deployable to any interface, in all formats
including reports, graphs, dashboards,
Data Warehouses & Data Marts emails, all supported platforms
27
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Analytic Deployment Challenges

• How do you get the analytic into the hands of the


person who will be using it?
• How do you ensure the data used to score the
model is consistent with the data used to build
that model?
• Is the data the same?
• Is consistency automatically enforced?
• Models evolve over time – how can the model be
updated without causing a lot of re-work?

Your BI platform must overcome


these challenges
28
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Powerful Predictive Analyses For Everyday Use
Typical Predictive Analyses Powerful Predictive Analyses
Based on Regression Techniques Based on Data Mining Techniques

Revenue Forecast Area


Achieve Revenue
Stay in Budget
DETERMINE Respond
WHO IS Purchase
LIKELY TO … Defraud
Be Profitable
On Time

Multi-variate Linear & Logistic


Regression,
Decision Trees, Clustering

Note: Linear, Exponential & Seasonal Forecasts

29
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Multiple, Customized Forecasts from One Model

30
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
TDWI Best Practices Award for Predictive Analytics

MicroStrategy customer Corporate Express won


a TDWI 2007 Best Practices in Data Warehousing
Award in the category of Predictive Analytics. TDWI’s
Best Practices Awards program is designed to identify
and honor companies that have demonstrated excellence
in developing, deploying, and maintaining BI and DW
applications.

31
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Multiple, Customized Forecasts from One Model

32
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Multiple, Customized Forecasts from One Model

33
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Dashboard Analyzing Regional Performance

34
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Scorecard combining descriptive & predictive

Predicition

Probability
Sales
Info

Demographic
Usage Info
Info

35
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Scorecard combining descriptive & predictive

Revenue Risk = Churn Propensity


x
Remaining
Contract
Value

36
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Scorecard combining descriptive & predictive

37
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Scorecard combining descriptive & predictive

38
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Dynamic Dashboards with Predictive Content

39
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Example:
Dynamic Dashboards with Predictive Content

40
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Questions?

But
Askof
Yes
42
course
Maybe
Again

41
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Questions?

Feel free to contact me


Rick Pechter
rpechter@microstrategy.com

42
CONFIDENTIAL
The Information Contained In This Presentation Is Confidential And Proprietary To MicroStrategy. The Recipient Of This Document Agrees That They Will Not Disclose Its Contents To
Any Third Party Or Otherwise Use This Presentation For Any Purpose Other Than An Evaluation Of MicroStrategy's Business Or Its Offerings. Reproduction or Distribution Is
Prohibited.
Questions??

43
Contact Information
• If you have further questions or comments:

Wayne Eckerson, TDWI


weckerson@tdwi.org

Rick Pechter, MicroStrategy


rpechter@microstrategy.com

44

También podría gustarte