Predictive Analytics and Price Optimization
Michael E. Angelina, ACAS, MAAA, CERA
Executive Director,
Academy of Risk Management & Insurance
Erivan K. Haub School of Business
Saint Joseph's University
Agenda
• Background
• Predictive Analytics defined
– IBM View, other definitions
• Insurance Industry Acceptance and Uses
• Demographics
• Price Optimization – Issues
Data Analytics - Background
• 2003 Yankees versus Red Sox, Game 7
– Pedro has the Yankees on the ropes;
– Boston manager, Grady Little decides to stay with his starter in the 8th inning
– Managerial decision based on instinct, Pedro’s reputation, and his season
• Season Stats:
–
–
–
–
14-4 Won – loss record; 2.22 ERA; .586 OPS;
29 Games Started; 186 innings; (<7 innings per start)
Only pitched into the 8th inning 5 times all season
Typically when he had 5 days of rest
• Lets mine the data a little more;
– OPS of .586 for season; in 4 starts against the Yankees OPS was .718
– OPS is on base plus slugging percentage:
Inning
1–5
6
7
OBP
.267
.295
.364
Slugging
.280
.395
.471
OPS
.534
.691
.835
Data Analytics (IBM view)
• IBM survey of 1,700 CEOs and public sector leaders identified technology
change as the most critical external factor impacting organizations.
• Three principal types of analytics solutions:
– Descriptive –what happened?
• provides information on past events (standard reporting, drill down/queries)
• Utilizes reports, dashboards, business intelligence
– Predictive –what could happen?
• provides answers for decisions (anticipate)
– Predictive modelling – what will happen next
– Forecasting – what if these trends continue
– Prescriptive – what should we do?
• explores a set of possibilities and suggests actions - optimization
• Factors uncertainty and recommends approaches to mitigate risks;
• AIG has a Science Officer to lead this global initiative
• Ace, Chubb, Travelers, and XL continue to advance analytics.
Predictive Analytics
• Not new to the industry
– Certain companies were inquisitive
• State Farm in the mid-70s; Progressive yesterday and today; Zenith in WC
• Catastrophe modeling in the 90s
• What has changed
– Computing power continues to increase exponentially
– Availability and accessibility of data (internal, personal, and external)
• Widespread acceptance in the business community
– Demographic changes; Consumer changes
– Innovate or Perish – Case Studies
• Insurance Industry Acceptance
–
–
–
–
Underwriting for personal lines and small commercial
Risk Management (Reinsurers, direct property writers)
Claims : personal and commercial lines
Distribution – personal lines and small commercial
Case Study - Yellow Pages
• In 2006 a one-inch ad in Manhattan, NY, cost $2,500
•Full-page size ad cost $92,000
•In 2011 the rough average price of a yearly ad decreased to $17,000
[1]
[1]
•According to an MSN study 70% of people do not open the Yellow Pages
•Seattle in 2010 allowed its residents to opt-out of receiving the Yellow
Pages
•2011 the 9th U.S. Circuit of Appeals sided with Yellow Pages
•By that time 79,000 Seattle residents had opted-out
[2]
[2]
[2]
•Failed to go digital fast enough
6
[1]
[2]
Case Study - BLOCKBUSTER
•Decade ago ruled the movie rental business
•25,000 Employees
•8,000 Stores
•6,000 Public DVD rental machines
[3]
[3]
[3]
[3]
•2005 company was valued at $8B
[3]
•Early 2000s Blockbuster decided not to purchase Netflix
•At the time Netflix was valued at $50M
•Current Netflix market cap is $20.8B
[4]
[4]
•Did not identify emerging technology
•Filed for bankruptcy in 2010
[4]
Image Source:
7
[4]
Analytics – Personal Lines
• Credit Scoring – controversial but high predictive value
• Telematics (Results of Deloitte Study)
–
–
–
–
25% favor; 25% opposed; 50% depends on the amount of the discount
Income level not a differentiator
Gender is not a significant differentiator
Age is a significant variable
• Younger drivers do not expect a large discount
• Two-thirds of 21-19 year olds are willing to try telematics versus 44% of over 60 year olds
• 35% yes (21-29) versus 15% yes (over 60)
• Genie is out of the bottle
– Personal lines – vehicle monitoring (bifurcated market: users and non-users)
– Commercial lines – commercial auto: taxi devices
– Behavioral shift – heightened loss control due to monitoring
Pause for a moment and reflect
Visualizing the Generations
Baby Boomers
Generation X
9
Generation Y
Purchasing Influences
[9]
10
Understanding Generation X
• Grew up in a time of technological advancement
–
–
–
–
–
Likely to research and purchase online
Values honesty and transparency
Desires fast turnarounds
Seeks tailored products and experience
In 2013 75% of Generation X banked digitally [18]
Graph Source: [18]
11
[17]
Increased use of digital banking is transitioning to
insurance purchasing habits
Smart Mobile Devices in Insurance
[9]
12
Deloitte Study on small business owners
Deloitte Small Business Study
• Surveyed 750 small business insurance buyers with <25 employees if
they would buy directly from insurers:
[23]
13
Deloitte Cont.
[23]
14
Price Optimization
• Systematic and statistical method to help an insurer estimate a
rating plan factoring in a competitive environment
• Informs an insurer’s judgment when setting rates by producing
suggested competitive adjustments to the actuarial indicated loss
costs
• Utilizes a variety of applied mathematical techniques (linear, nonlinear, integer programming) to analyze insurer’s data and other
considerations
• Enables exhaustive search across thousands of pricing alternatives
in multiple scenarios to assist insurers in comparative rate analysis
– Improves efficiency of rate setting process;
– Enables companies to more accurately predict the outcome of
their rate decisions
Ratemaking Process – Step Back
• Regulatory Requirement – rates must be adequate, not excessive,
or unfairly discriminatory
• Process (per EPIC Consulting)
– Actuaries determine expected losses, expenses, and profit loading
– Management makes adjustments to reflect business considerations,
marketing, underwriting, and competitive conditions
– Regulators permit insurers to reflect judgment and competitive
environment in rates
– Rate Filer (Insurer) must ensure that filed rates are adequate, not
excessive, or unfairly discriminatory
– Actuaries can opine that the filed rates meet statutory standard if
reasonably close to actuarial estimate (eg reserving)
Price Optimization - Proponents
• Compare price optimization to traditional rating approach
– Traditional approach: Base rate (loss cost) x adjustment factors
• Adjustment factors based on age, gender, territory, make and model year
– Price Optimization: Base rate 9loss cost) x adjustments
• Adjustments based on price optimization methodology
• All companies consider customer response in pricing either
underwriting criteria or marketing considerations
– Price optimization is just more scientific (statistics versus judgment/market)
• Loss Costs remain the foundation of the rate setting process
– Price optimization factors typically are designed to stay within constraints
imposed by confidence interval of cost estimates
• Personal lines is a very competitive market as evidence by
advertising spend by large insurers
– Competition has decreased the size of the assigned risk markets
Price Optimization - Issues
• Price Optimization has generated much controversy from Consumer
Federation of America and some regulators
• Relies on an analysis of the elasticity of demand of customers to
raise prices above the cost-based estimate on some segments of
the policyholders who are known to be less likely to change insurers
when price increases are below a certain threshold
– Great inertia in the personal lines market (people tend not to shop
much), as evidenced by recent survey
• 24% have never shopped for auto insurance (27% HO)
• 34% rarely shop for auto insurance (33% HO)
• 27% shopped within every other year for auto insurance (20% HO)
– Price Optimization tries to find these policyholders!
Price Optimization - Questions
• How does price optimization fit within the actuarial profession
– Cost-based resides with actuaries;
– Where does the demand and competitive analysis reside?
– Should actuaries be involved in price optimization at all ?
• Is price optimization ratemaking or NOT ratemaking?
– Actuarial code of conduct (precept 1?)
• Is price optimization in compliance with:
– Statement of principles on ratemaking
– Actuarial standards of practice
– Actuarial practice note (ratemaking practice note does not exist!)
• Should the actuary consider outcomes other than cost when
making rates?