Transcript MasterCard Worldwide Presentation - University of Missouri

Susan Meyer, MasterCard Network Products
April 22, 2013
Business Intelligence & Analytics
Career Paths and Industry Trends
©2013 MasterCard.
Proprietary and Confidential
Agenda
• Introduction
• The BI Job Market
• Analytical Processes
• Data Science Roles
• Analytics Products & Services
• Analytical Platforms
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 2
Data Crunchers In Demand
Data Scientist
Deemed “the sexiest job of the
21st century” by Harvard
Business Review, data scientists
bridge the gap between the skills
of a statistician, a computer
scientist and an MBA.
Salaries vary from $110,000 to
$140,000.
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 3
Data Mining Job Prospects
• Gartner says worldwide IT spending will increase 3.8
percent in 2013 to reach $3.7 trillion, and that
excitement for big data is leading the way.
• By 2015, 4.4 million jobs will be created to support big
data.
• Over 90-percent of the NCSU Class of 2013 have
received one or more offers of employment, and over
80-percent have accepted new positions. The average
base salary reached an all-time high of $96,900, an
increase of nearly 9% over the Class of 2012.
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 4
A Brief Overview of Data Mining
Innovation
Business Question
Technologies
Data Collection
(1960’s)
“What was total revenue in
the past 5 years?”
Mainframe computers,
tape backup
Data Access
(1980’s)
“What were unit sales in
New England last March?”
RDBMS, SQL, ODBC
Data Warehousing
(1990’s)
““What were unit sales in
New England last March?
Drill down to Boston”
OLAP, multidimensional databases,
data warehouses
Data Mining
(Today)
“What’s likely to happen to
Advanced algorithms,
Boston sales next month and massively parallel
why?”
databases, Big Data
http://www.thearling.com/
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 5
Business Intelligence & Data Mining Services
Descriptive
Predictive
• Dashboards
• Predictive analytics
• Process mining
• Prescriptive analytics
• Text mining
• Realtime scoring
• Business performance
• Online analytical
management
• Benchmarking
processing
• Ranking algorithms
• Authentication
These functions are highly inter-related and fall on a continuum
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 6
CRISP-DM:
Data Mining Methodology is Highly Iterative
• Up to 60% of the
work effort in a
major data mining
project is typically
related to data
preparation and
cleansing
• Be prepared for
the unexpected
when working with
real-world data
ftp://ftp.software.ibm.com/soft
ware/.../Modeler/.../CRISPDM...
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 7
Data Scientists Work in Teams
Job Categories
•Business Analyst
•Data Analyst
•Data Engineer
•Data Scientist
•Marketing
•Sales
•Statistician
Many major organizations in
St. Louis are actively using
data mining in their core line of
business
http://www.datasciencecentral.com/
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 8
Statistical Business Analyst
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 9
Statistical Programmer
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 10
Data Integration Developer
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 11
Predictive Modeler
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 12
A Typical Product Developer / Data
Scientist Role
Job Details
Facebook is seeking a Data Scientist to join our Data Science team. Individuals in this role are expected to be comfortable working
as a software engineer and a quantitative researcher. The ideal candidate will have a keen interest in the study of an online social
network, and a passion for identifying and answering questions that help us build the best products.
Responsibilities
Work closely with a product engineering team to identify and answer important product questions
Answer product questions by using appropriate statistical techniques on available data
Communicate findings to product managers and engineers
Drive the collection of new data and the refinement of existing data sources
Analyze and interpret the results of product experiments
Develop best practices for instrumentation and experimentation and communicate those to product engineering teams
Requirements
M.S. or Ph.D. in a relevant technical field, or 4+ years experience in a relevant role
Extensive experience solving analytical problems using quantitative approaches
Comfort manipulating and analyzing complex, high-volume, high-dimensionality data from varying sources
A strong passion for empirical research and for answering hard questions with data
A flexible analytic approach that allows for results at varying levels of precision
Ability to communicate complex quantitative analysis in a clear, precise, and actionable manner
Fluency with at least one scripting language such as Python or PHP
Familiarity with relational databases and SQL
Expert knowledge of an analysis tool such as R, Matlab, or SAS
Experience working with large data sets, experience working with distributed computing tools a plus (Map/Reduce, Hadoop, Hive,
etc.)
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 13
Advanced Training
• University of Missouri – St. Louis
• Northwestern University
• University of California – San Diego
• University of California – Irvine
• North Carolina State University
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 14
Case Study:
Realtime Scoring Systems
MasterCard, NAC Announce ‘Real World'
Plan to Address April 19 Chip Liability Shift
• All Maestro transactions will be blocked at
ATMs that averaged no more than one
Maestro transaction per month in 2012.
This represents approximately 80 percent
of U.S. ATMs.
• Fraud Rule Manager will block Maestro
transactions at low-activity ATMs and will
decline potentially fraudulent transactions
at the remainder of ATMs.
• A complementary Fraud Control Shield program will be rolled out
across Maestro card-issuing FIs in Europe, using similar metrics to
identify and decline potential fraudulent ATM transactions on the issuer
side.
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 16
US Maestro Cross-Border ATM Fraud is a
Localized Problem: FY 2012 Trends
Ranking of US ATM Fraud $ Losses for Maestro XR
• Fraud losses per ATM range
from zero to thousands
$80,000
Maestro XR $USD Losses
$70,000
$60,000
Only a small
number of ATMs
experience high
Maestro XR fraud
losses
$50,000
$40,000
$30,000
3.1%
$20,000
$10,000
1
360
719
1078
1437
1796
2155
2514
2873
3232
3591
3950
4309
4668
5027
5386
5745
6104
6463
6822
7181
7540
7899
$-
US ATM Count
• Only 8,242 of the 264,516 US ATMs that
processed cross-border Maestro traffic
saw any fraud (3.1% of total)
96.9%
$0 Fraud
>$0 Fraud
Source: MasterCard Fraud Mart
©2013 MasterCard.
Proprietary and Confidential
November 9, 2012
Page 17
Expert Monitoring Solutions:
Data Mining in Action
We leverage one infrastructure with layers of technology to route the transaction from
the network to the appropriate technology platform for the value-added service.
EMS Fraud
Scoring
Issuers
EMS Fraud
Scoring
Merchants
EMS
Compromise
Accounts
Web Session
Threat Index
Merchants
Fraud Rule
Manager
Transaction
Blocking
Prepaid ATM
Monitoring
Fraud Decisioning Platform
Fraud Rules Engine
(EMS iPrevent, Silver Tail, and MasterCard technology)
(IBM iLog BRMS technology)
ATM
Acquirer
Monitoring
Data Analytics
Auth IQ Platform
Intelligent, real-time transaction monitoring interface routes transactions
to the appropriate value-added service
Network
Access
Point
Acquirer
Issuer
©2013 MasterCard.
Proprietary and Confidential
Page 18
Fraud Rule Manager for ATM:
Detection Rate Performance Results
Inter-Regional ATM Performance, All Brands
73.19%
64.54%
72.01%
62.48%
3:1 TFPR
5:1 TFPR
TDR
VDR
Results for the blind test months of Nov – Dec 2012
MasterCard Model Performance Report, Mar 2013
, TDR=Transaction Detection Rate, VDR=Value Detection Rate
©2013 MasterCard.
Proprietary and Confidential
March 5, 2013
Page 19
Fraud Rule Manager for ATM:
Financial Performance Results
Genuine Blocked Txns vs. Genuine ATM Txns Retained
2%
At the 5:1 TFPR threshold:
• 72% of the $USD fraud liability
is blocked
98%
Genuine Count Retained
• 1.94% of genuine Maestro
cross-border transactions are
blocked
Genuine Count Blocked
ROI analysis has shown the fraud loss avoidance far outweighs lost
revenue on the small percentage of genuine transactions blocked
MasterCard FRM ATM Model Performance Report, Mar 2013
©2013 MasterCard.
Proprietary and Confidential
March 5, 2013
Page 20
MasterCard Fraud Data Mart
A set of platforms, software tools, and DW data dedicated to support
MasterCard’s Fraud detection and prevent efforts.
Massive
• 120 TB Netezza Platform
• 70 TB Used
• 50 TB Free
Multi-sourced
• Daily transactional inputs
• Core DW (3+ current-year
copy of Authorization/Clearing/
Debit/Chargeback/Retrieval
Requests)
• Risk (All Fraud/ADC)
• EMS (3+ years of
Supplemental data)
• AMS (Stop List)
Usage
• 65 Users
• Data Analytics, Profiling and
Predictive Modeling (SAS)
• Production Batch Processing
• 289 Million LTV Accounts
updated weekly
• Supports Business Rules
(IBM iLog)
• Daily / Weekly Batch Feeds
to Expert Monitoring
Systems
• Internal Reporting
Features
• Online Q4 2010
• 3+ year historic view
• Unique PAN Proxy/Un-proxy for
each vendor.
• IBM Pure / Netezza analytics
tools
• Transaction Life Cycle
Transaction Life Cycle*
(Match Rates)
• Clearing to Auth.
- US: 97%/ Non US: 92%
• Fraud to Auth.
- US: 97%/ Non US: 80%
• Fraud to Pin Debit - 97%.
• Fraud to Clearing - 85%
The design of the Fraud Data Mart and analytic innovations led to submission of
three pending patents and 2012 Technology Lever IT Transformation Award
©2013 MasterCard.
Proprietary and Confidential
Page 21
How Major Financial Institutions Use SAS
SAS Fraud Management - End-To-End Value Chain
•
Extraction and
Manipulation of
Data
• Data Quality
• Data preparation,
summarization
and exploration
Detection
Analytics
Data Management
•
•
Modeling
Ad Hoc Query &
Reporting
• Diagnostic
Analytics
• Optimization to
provide
alternative
scenarios
•
Continuous
Monitoring
• Alert Generation
Process
• Real-time
Decisioning
• Balance between
risk and reward
Alert Management
•
Social Network
Investigation
• Alert Disposition
• Case
Management
Integration
Case Investigation
•
Workflow & Doc
Management
• Intelligent Data
Repository
• Continuous
Analytic
Improvement
• Dashboards &
Reporting
STREAM IT - SCORE IT - STORE IT
©2013 MasterCard.
Proprietary and Confidential
Gartner Magic Quadrant for Business
Intelligence Platforms
BI Platform Decision Makers:
• IT — 38.9%
• Business user — 20.8%
• Blended business and IT
responsibilities — 40.3%
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 23
Q&A
• Your questions?
©2013 MasterCard.
Proprietary and Confidential
January 30, 2013
Page 24