analytics & insurance. serene zawaydeh
TRANSCRIPT
September 19, 2014
Healthcare Fraud – Annual Losses
USA: $70 billion to $260 billionEstimates of The National Health Care Anti-Fraud
Association and the Federal Bureau of Investigation
EU: $30 billion to $100 billion
How can we best detect, prevent and combat fraudulentactivity in health care? SAS White Paper
Big Data� Variety Structured versus Unstructured Data
� Volume
� Velocity Real Time
Value � Value
Privacy and Security of Data
The Power of Analytics
� TD Bank Commercial July 2014� Over 9 million views
� ATM Automatic Teller Machine�Turned Into Automated Thanking Machine
� Pre-selected customers � Pre-selected customers � Customized, personalized the message.� Recognized the customers � Thanked them for being faithful customers� Analyzed customers’ transactions� Recognized their needs.� Identified supported family members (Sick daughter; Kids)� Provided a gift that is customized to customers’ needs, and the supported family
members.� Tickets to Trinidad to Visit Daughter Human Touch
Disney Land for the kids
Big Data Governance &
Insurance Industry Examples
� Big Data Governance, An Emerging Imperative
� Sunil Soares MC Press 2012
Claims Analytics – Data Quality� Large health plan processed over 500 million claims per year � Each claims record consisting of 600 to 1,000 attributes � Used Predictive Analytics to determine if certain proactive care
was required
� Business Intelligence Team� Limited Effectiveness of Predictive Analytics: � Limited Effectiveness of Predictive Analytics: � Physicians used inconsistent procedure codes to submit claims� Analyzed text within claims documents� Determined candidates for disease management programs.
“Blood sugar monitoring”… Diabetes
Insurance Investigation
� Use Facebook to validate auto claims
� Claim filed for hit and run with auto insurer
� Comments on daughter’s Facebook saying � Comments on daughter’s Facebook saying daughter was responsible for the accident
� Policyholder convicted of filing a fraudulent insurance claim
Underwriting
� Should insurers be able to use social media for underwriting purposes to set rates for policies?
� Examples� Examples
� Use Facebook information of an athletic 50 year old tennis player to reduce the premium due to lower risk
� Or increase premium of a Skydiver due to higher risk?
Regulation USA
� Each state has department of insurance to regulate insurance industry within its jurisdiction.
� Each department of insurance has an important role in protecting the privacy of policyholders in that state. that state.
� US States once barred insurers from using credit scores to predict the likelihood of claims, most nowallow it, with California being an exception.
� Will regulators permit insurers to use social media to set rates?
Monitoring Social Media
� Health plan established a dedicated team to monitor mentions of company in social media
� If member or physician posted complaints about health plan on Twitter, someone at company reviewed it and responded.responded.
� Objective:
� Introduce the company’s response into public record
� Anybody who saw the complaint would see the response.
� HIPPA privacy regulations – social media comments
� Moved conversation to phone ASAP
Centralized Insurance Claims
Database European Country
Big Data and private insurance
Rate not high enoughMany insurers don’t use customer-focused or marketing-focused data warehouses
Claims losses… Fraud Investigation Claims losses… Fraud Investigation
Solitary claim for theft of luxury car from relatively low-income postal district.
However… 30 different people in the same postal district had theft claims for luxury cars with other insurers over the previous two months
Centralized Insurance Claims
Database European Country
� Insurance industry in a European Country banded together to pool claims information for fraud investigation
� Database does fuzzy matches of
� People living in the same street, with the same or similar � People living in the same street, with the same or similar birth dates, or with couple of numbers transposed, with accounts with the same bank, with similar names, and so on.
� Mostly done after claims were paid� Might be too late to recover the money
Need for Big Data Governance
Need to reassure participating insurers that - Their data is secure - Will never be disclosed to competitors- Will not be used to damage relationship with loyal clients.- Security and Confidentiality of claims analytics
- Real time or near real time access to historicalInvestigator data
- Insurers investigate the right claims upfront, versus after they have been paid
Whiplash Claims
� Minor accident but the passengers claim large compensation for neck injuries, which are very hard to prove
� Claims investigators have ability to compare claims from same area across multiple insurers where claimants have the same or similar names or postal codes and shared bank accounts
SAS Insurance Analytics
SAS White Papers
� Claims Fraud� Detect and prevent both opportunistic and professional fraud
throughout claims process
� Underwriting Fraud� Prevent premium leakage at point of sale and renewal� Prevent premium leakage at point of sale and renewal
� Rate evasion � Spot rate evasion tactics during the quote process before issuing a
policy
Analytics for Insurance
What Does Big Data Really Mean for Insurers? New Paradigms and New Analytic OpportunitiesFeaturing as an example: SAS® High-Performance Analytics An SMA Perspective. Authors: Deb Smallwood, Founder ; Mark Breading, Partner
Published Date: August, 2012. This perspective is based on SMA’s ongoing research on data and analytics in insurance.
SAS
Benefits� Deliver information that is consistent, accurate, verifiable and up-
to-date.
� SAS Insurance Analytics Architecture enables access to accurate data consistently, when and where it is needed, giving increased confidence in accuracy and timeliness of your data.confidence in accuracy and timeliness of your data.
� Complete, integrated view of all enterprise data.
� Always have access to data needed, when needed
SAS
� Predictive modeling
� Social network analysis
� Structured and unstructured data analysis
� Real time Data Cleansing
Data mining, Clustering, Neural networks, decision � Data mining, Clustering, Neural networks, decision trees
� Regression analysis
� Scorecards
� Data Dictionary covers all key insurance subject areas – e.g., customer, policy, claim, financial accounting
and reinsurance, predefined logical and physical data models Standardizes more than 5,000 insurance data elements.
Serene Zawaydeh� MBA, French – Scholarship (2006-2008)� B.Sc. Electrical Engineering (1993-1998)� CFA Level 1 (2011)
University of Toronto, School of Continuing Studies (January 2014 – Present)� Enterprise Data Analytics (Big Data) Courses: � Big Data Tools and Techniques Mining Financial, Operational, and Social Network Data (Sept - Dec 2014)� Value Proposition and Technologies of Enterprise Data Analytics� Foundations of Enterprise Data Analytics – Concepts and Controls
Leadership Essentials� Be an Effective Negotiator� Critical Thinking
� International Foundation of Employee Benefit Plans (IFEBP) and Dalhousie University E-courses (August & Sept 2014)- The Group Insurance Landscape- Canadian Course- Group Benefits Design and Administration - Canadian Course- Group Benefits Design and Administration - Canadian Course- Group Benefits Funding and Pricing - Canadian Course- Life Cycle of a Group Benefits Plan - Canadian Course
• Data Analyst, Sun Life Financial (May 2014-September 2014)
• Technology Transfer Intern – UOIT (Oshawa) (2012-Mar 2013)� Patent search and intellectual property protection; Engineering innovations; applications for funding; identifying potential licensees
� Head of Research, EquityFinancial Analysis; Equity Valuation (2008-2011)
� Stocks listed in Jordan
Telecom Market ResearchConsultant; Senior Research Analyst; Research Analyst (2003-2007)
� Middle East and North Africa (Arabic and French language skills)
� One month Scholarship to Germany – Intensive course in German language (2001)� Zentrale Mittelstufenpruefung� Zertifikat Deutsch