sas enterprise miner overview
TRANSCRIPT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
BUILDING THE DATAndashDRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
AGENDA
Why is Big Analytics a topic today
The role of Big Analytics in the data driven enterprise
Challenges that organizations face
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHY IS BIG ANALYTICS A TOPIC TODAY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Infinite Volume and Variety of Data
Disruptive Technology
UnrivaledProcessing Power
New Problem-solving Mindset
WHERErsquoS THE BUZZ
COMING FROM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Data
Driven
Enterprise
Disruptive Technology
UnrivaledProcessing Power
New Problem-solving Mindset
Infinite Volume and Variety of Data
WHERErsquoS THE BUZZ
COMING FROM
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
DIGITAL DATA IS THE OIL OF THE DIGITAL ECONOMY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHATrsquoS TRENDING INTERNET OF THINGS
40 terabyteshour
1 gigabytesecond
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
330PM
THE ANALYTICS
OF THINGS
CONNECTED EVERYTHING AND
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE ROLE OF BIG ANALYTICS IN THE DATA
DRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
AGENDA
Why is Big Analytics a topic today
The role of Big Analytics in the data driven enterprise
Challenges that organizations face
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHY IS BIG ANALYTICS A TOPIC TODAY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Infinite Volume and Variety of Data
Disruptive Technology
UnrivaledProcessing Power
New Problem-solving Mindset
WHERErsquoS THE BUZZ
COMING FROM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Data
Driven
Enterprise
Disruptive Technology
UnrivaledProcessing Power
New Problem-solving Mindset
Infinite Volume and Variety of Data
WHERErsquoS THE BUZZ
COMING FROM
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
DIGITAL DATA IS THE OIL OF THE DIGITAL ECONOMY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHATrsquoS TRENDING INTERNET OF THINGS
40 terabyteshour
1 gigabytesecond
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
330PM
THE ANALYTICS
OF THINGS
CONNECTED EVERYTHING AND
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE ROLE OF BIG ANALYTICS IN THE DATA
DRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHY IS BIG ANALYTICS A TOPIC TODAY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Infinite Volume and Variety of Data
Disruptive Technology
UnrivaledProcessing Power
New Problem-solving Mindset
WHERErsquoS THE BUZZ
COMING FROM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Data
Driven
Enterprise
Disruptive Technology
UnrivaledProcessing Power
New Problem-solving Mindset
Infinite Volume and Variety of Data
WHERErsquoS THE BUZZ
COMING FROM
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
DIGITAL DATA IS THE OIL OF THE DIGITAL ECONOMY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHATrsquoS TRENDING INTERNET OF THINGS
40 terabyteshour
1 gigabytesecond
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
330PM
THE ANALYTICS
OF THINGS
CONNECTED EVERYTHING AND
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE ROLE OF BIG ANALYTICS IN THE DATA
DRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Infinite Volume and Variety of Data
Disruptive Technology
UnrivaledProcessing Power
New Problem-solving Mindset
WHERErsquoS THE BUZZ
COMING FROM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Data
Driven
Enterprise
Disruptive Technology
UnrivaledProcessing Power
New Problem-solving Mindset
Infinite Volume and Variety of Data
WHERErsquoS THE BUZZ
COMING FROM
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
DIGITAL DATA IS THE OIL OF THE DIGITAL ECONOMY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHATrsquoS TRENDING INTERNET OF THINGS
40 terabyteshour
1 gigabytesecond
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
330PM
THE ANALYTICS
OF THINGS
CONNECTED EVERYTHING AND
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE ROLE OF BIG ANALYTICS IN THE DATA
DRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Data
Driven
Enterprise
Disruptive Technology
UnrivaledProcessing Power
New Problem-solving Mindset
Infinite Volume and Variety of Data
WHERErsquoS THE BUZZ
COMING FROM
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
DIGITAL DATA IS THE OIL OF THE DIGITAL ECONOMY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHATrsquoS TRENDING INTERNET OF THINGS
40 terabyteshour
1 gigabytesecond
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
330PM
THE ANALYTICS
OF THINGS
CONNECTED EVERYTHING AND
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE ROLE OF BIG ANALYTICS IN THE DATA
DRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
DIGITAL DATA IS THE OIL OF THE DIGITAL ECONOMY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHATrsquoS TRENDING INTERNET OF THINGS
40 terabyteshour
1 gigabytesecond
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
330PM
THE ANALYTICS
OF THINGS
CONNECTED EVERYTHING AND
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE ROLE OF BIG ANALYTICS IN THE DATA
DRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHATrsquoS TRENDING INTERNET OF THINGS
40 terabyteshour
1 gigabytesecond
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
330PM
THE ANALYTICS
OF THINGS
CONNECTED EVERYTHING AND
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE ROLE OF BIG ANALYTICS IN THE DATA
DRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
330PM
THE ANALYTICS
OF THINGS
CONNECTED EVERYTHING AND
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE ROLE OF BIG ANALYTICS IN THE DATA
DRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE ROLE OF BIG ANALYTICS IN THE DATA
DRIVEN ENTERPRISE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE QUIET REVOLUTION OF NUMERICAL WEATHER
PREDICTION
Source Nature Bauer etal 2015
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2015 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANTICIPATE OPPORTUNITY
WHAT DOES ANALYTICS HELP YOU DO
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TAKE ACTION
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DRIVE IMPACT
WHAT DOES ANALYTICS HELP YOU DO
ANTICIPATE OPPORTUNITY TAKE ACTION
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
VOLUME
VARIETY
VELOCITY
VALUE
TODAY THE FUTURE
DA
TA
SIZ
E
THRIVING IN THE NEW DATA ERA
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Transformational AnalyticsThe application of modern analytics to big data has disruptive power and has become an
important topic at the board table of many organizations
Fraud Detection
Complaint Analysis
Workforce demand planning
Human resource planning
Quality Analysis
Customer Segmentation
Customer Lifetime Value
Best Next Action
Personalize contextual marketing
Pricing
Campaign Optimization
Net Promoter Scores
CMO
COOCFORisk Modeling
Compliance
Demand Planning
CIOCyber Security
Network Capacity
Planning
Business Enablement
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
ANALYTICS GARTNER DEFINITION OF THE ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS PRESCRIPTIVE ANALYTICS
Real-time decision support Real-time decision automation
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SAS ANALYTICS CONTINUUM
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Forecasting
Machine
Learning
Optimization
Text
Analytics
BusinessSolutions
Forecasting
Machine
Learning
Text
Analytics
Optimization
Data Mining
Each technology works well on its own but
combining them all is the real opportunity
Data ManagementData Management
The Whole is Greater than the Sum of its Parts
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOPREDICTIVE ASSET MAINTENANCE TRANSPORTATION
BUSINESS CHALLENGE
bull Predict maintenance needs of individual trucks before failures
occur
bull Proactively service trucks at opportune time
bull Provide new service offering with high fleet SLA
SOLUTION
bull Data from 60+ sensors truck
bull Integrated data with product details warranty claims and related
data sources
bull Analytic models predict the likelihood of specific failures within
30 days with 90 accuracy
bull Better root cause insight led to higher productivity
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIO
BUSINESS CHALLENGE
bull Monitoring Electronic Submersible Pump efficiency amp well performance
for deep sea drilling rigs
bull Failure of one pump is $2Mday one day of productivity loss equates to
$20M in deferred revenue
CRITICAL COMPONENT FAILURE
AVOIDANCEOIL AND GAS COMPANY
SOLUTION
bull Over 21 million sensors generating 3 trillion rows of dataminute
monitored for potential failure (temperature vibration )
bull Failure predicted 90 days in advance
bull Reduced down time from 3 days to 6 hours
bull Savings est $3 million per failure
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CUSTOMER
SCENARIOSMART GRID STABILIZATION UTILITIES
Continuous monitoring for
patterns of interest
Detecting
Occurrence
Detection
Qualification
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WIFI
MONETIZATION
CLIENT EXAMPLE SPONSORED FREE WIFI
25
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
CHALLENGES THAT ORGANIZATIONS FACE
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
TECHNOLOGY
PEOPLE
BUSINESS
PROCESS
SUCCESSFUL
ANALYTICS
CHALLENGES ALIGNMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
PROCESS ADVANCED ANALYTICS LIFECYCLE
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
This slide is for video use only
Copyright copy 2014 SAS Insti tute Inc Al l r ights reserved
ldquoldquoExperiment is the only
means of knowledge at
our disposal E poetry
minusMax Planck
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Lo
st
Va
lue
ADVANCED
ANALYTICSREDUCE TIME TO DECISION
It is proven that analytically
empowered decision making
provides a significant uplift
Producing a new model or
adjusting an existing model
for the business often takes
too long to meets fast
changing markets
Complexity is added as many
stakeholders are involved in
the predictive analytics
process
Big data is adding to the
complexity
Automation of the process
model is needed to provide
fast repeatable and high-
quality results
Value
Time
Data
Latency
Deployment
Latency
Decision
Latency
Lost Time
Modeling
Latency
Evaluation
Latency
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXPANDING DATA REQUIRES NEW APPROACH
bull Project centric business use
bull Mainly structured and internal
selected data
THE NEW
ANALYTICS
PARADIGM
Data
Data
Data
Data
Data
Data
bull Discovery centric
business use
bull All data
relevant for
the problem to
solve
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Controller
Client
ADVANCED
ANALYTICS
SCALE YOUR ANALYTICS PLATFORM WITH YOUR DATA
AND YOUR PROBLEM COMPLEXITY
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Retention Campaigns 15 improvement
270 million price points analyzed in 2 hrs (from 30 hrs)
Increase coupon redemption rate from
10 to 25
Recalculate entire risk portfolio from 18 hours to 12 minutes
Regression analysis from
167 hours (1 week) to 84 seconds
OUR PERSPECTIVE
CASE STUDIES BIG DATA ANALYTICS DRIVES HIGH IMPACT RESULTS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Discovery Deployment
Iterative
Visual
Experiments
Fail Fast
Data Science
Interactive
New data
Innovation
Deep Learning
Governed
Robust
Automated
Regulated
Actions
Consistent
Documented
Decisions
Analytics Lifecycle
Prepare
Explore
Model
Implement
Act
Evaluate
Ask
Data
Domain Expert
Ask questions
Evaluates processes and ROI
BUSINESS
MANAGER
Data exploration
Data assessment
BUSINESS
ANALYST
Exploratory analysis
Variable generation
Variable reduction
Descriptive segmentation
Predictive modeling
Model assessment
DATA
SCIENTIST
Model deployment
Model execution
IT SYSTEMS DATA
MANAGEMENT
Decision maker
Evaluates success
BUSINESS
MANAGER
Model monitoring
Model retraining
IT SYSTEMS DATA
MANAGEMENT
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
SO WHAT IS A DATA SCIENTIST
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
MIT RESEARCH CHALLENGES - PEOPLE
Top Three Analytics Challenges
MIT Sloan Management Review
Fifth annual research to
understand the challenges and
opportunities with analytics
2719 global survey respondents
across industries
28 in-depth interviews with
executives from companies like
Coca-Cola General Mills General
Electric DBS Bank etc
bull Companies that have a talent strategy and are able to successfully combine
analytics skills with business knowledge are more likely to create a
competitive advantage with data
bull Increase in data not insights Despite more data companies are still
struggling to turn their data into insights that drive value
bull 8 in 10 respondents are seeing an increase in data but only half are seeing
an increase in insights from the data
bull Surprisingly 80 have yet to develop a strategy to build and maintain their
talent bench
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
EXTEND ORGANIZATIONAL TALENT POOL
Analytics CollaborationData Scientist Superhero
COLLABORATION
Business Analyst
Citizen Data Scientist
Data Scientist
Builds model
pipeline templates
Adapts model
pipeline templates
Uses model
pipeline templates
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WHAT IS HAPPENING TO THE STATISTICIAN
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
DATA SCIENTIST ARE EXPENSIVE AND HARD TO FIND
Source Indeedcom
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d sascom
THANK YOU
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS
bull Data latency drivers
bull Data from different sources and systems
bull Departmental silos Dependency of LOB on IT
bull Need to create ABT for modeling task
bull Move data between data repositories and analytics environment
bull Modeling latency drivers
bull Different tools for different steps in analytics workflow
bull Tools do not support experiments and iterative approach
bull Need to apply algorithms from different analytical disciplines
bull Analytics environment does not scale with size of data and
complexity of problem
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Deployment latency drivers
bull Departmental silos Dependency of LOB on IT
bull No integration between development and production
environment
bull Manual creation and validation of production scoring assets
bull Decision Latency drivers
bull Production scoring environment does not scale with size of
problem
bull Results are not provided at right time in right format
bull No buy-in for business on use of analytical results in business
process
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
LATENCY DRIVERS (CONTINUED)
bull Evaluation Latency drivers
bull Failure to automatically capture actual outcomes and
feedback into the analytic loop
bull No standard workflow for addressing model decay
bull No standardized KPIs to measure model performance
bull No standardized thresholds for actions to refresh models
bull No automated model refitting where appropriate
PREDICTIVE
ANALYTICS
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
WORKFLOW MANAGEMENT (METADATA)
bull Open Source Analytics can leverage SAS enterprise capabilities
bull Data Access and Preparation
bull Data Dictionary
bull Lineage
bull Resource Management
bull Deployment
bull Model Assessment
DATA MANAGEMENT MODEL DEVELOPMENT MODEL DEPLOYMENT
SAS and Open Source Analytics
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
THE REALITY OF
MANAGING BIG
DATA
THE SITUATION TODAY
BUSINESS
PROBLEM
BUSINESS
DECISION
2080
Preparing
to
solve the problem
Solving
the
problem
Copyr i g ht copy 2014 SAS Ins t i tu t e Inc A l l r ights reser ve d
FLIPPING THE
SCRIPTHOW CAN YOU CHANGE THE EQUATION
BUSINESS
PROBLEM
BUSINESS
DECISION
20 80
Preparing
to solve the
problem
Solving
the
problem