2017 predictive analytics symposium - soa€¦ · 2017 predictive analytics symposium . ... blind...

2017 Predictive Analytics Symposium

Session 24, General Insurance Applications of PA

Moderator: Stuart Klugman, FSA, CERA, Ph.D.

Presenter:

Peter Wu, ASA, FCAS, MAA

SOA Antitrust Compliance Guidelines SOA Presentation Disclaimer

https://www.soa.org/legal/antitrust-disclaimer/https://www.soa.org/legal/presentation-disclaimer/

General Insurance Applications of

Predictive Analytics: Past,

Current, and Future

Deloitte Consulting LLP

2017 SOA Predictive Analytics SymposiumChicago, September 2017Peter Wu, FCAS, ASA, MAAA, CSPAManaging Director

- 2 - Copyright 2017 Deloitte Development LLC. All rights reserved.

Theme

Data analytics in the U.S. Property and Casualty insurance industry is HOT!

Why?

Because P&C insurance is a zero sum game, and data and analytics will create adverse selection for the companies who are not doing it!


A Success Story - Credit Score Revolution

80%

85%

90%

95%

100%

105%

110%

115%

-5%

0%

5%

10%

15%

20%

25%

30%

35%

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Com

bine

d R

atio

Gro

wth

Rat

e

Year

Guess Which Company is It?

Industry Growth RateProgressive Growth RateIndustry Combined RatioProgressive Combined Ratio


The Evolution of P&C Insurance Data AnalyticsAdvanced analytics along with leveraging large amount of internal and external data have become mainstream over the last 20 years in several industries within financial services. Property and Casualty insurance has been one of the leading industries in the integration of advanced data analytics into core operations.

Credit Scoring An early bellwether of the disruptive power of data in insurance.

Predictive modeling transformation of the P&C industry and actuarial profession.

Analytics-powered key operations such as underwriting, claim triage, and marketing

From predictive modeling to broad based analytics and big data

Increasingly and granular applications on every aspect of insurance operations and customer service.

A core strategic capability Actuaries and data scientists

1990 2000s Today and Future


Credit Score Revolution

Introduced in late 1980 and early 1990s First important factor identified over the past 2

decades Strongly correlated with P&C insurance loss Composite multivariate score vs. raw credit

information Viewed at first as a secret weapon Quiet, confidential, controversial, black box, etc

1997 NAIC/Tillinghast Study of 9 Companies' DataLoss Ratio Relativity of the Best and Worst 20% of Credit Score

Co1 Co2 Co3 Co4 Co5 Co6 Co7 Co8 Co9 AvgBest 20% -38% -29% -19% -15% -14% -34% -22% -22% -36% -25%Worst 20% 48% 20% 32% 30% 46% 59% 20% 22% 95% 41%

Early believers and users have gained significant competitive advantage!

Credit Score for Auto Insurance Application

Sheet1

Exhit 1 Tillinghast - NAIC Credit Score Study [4]

Company 1Company 2Company 3

Scores & Loss Ratio Relativity SummaryScores & Loss Ratio Relativity SummaryScores & Loss Ratio Relativity Summary

ScoreMidpointEarnedLoss RatioScoreMidpointEarnedLoss RatioScoreMidpointEarnedLoss Ratio

IntervalPremiumRelativityIntervalPremiumRelativityIntervalPremiumRelativity

813 or more850.010.2%0.657-0.380840 or more854.010.0%0.607-0.290826 or more845.010.0%0.723-0.187

768-812790.09.9%0.584823-839831.010.0%0.813803-826814.510.0%0.903

732-767749.511.0%0.692806-822814.010.0%0.626782-803792.510.0%0.895

701-731716.010.9%0.683789-805797.010.0%1.342759-782770.510.0%0.795

675-700687.510.4%1.184771-788779.510.0%1.059737-759748.010.0%1.073

651-674662.59.8%0.793748-770759.010.0%1.019710-737723.510.0%0.941

626-650638.09.9%1.332721-747734.010.0%1.322680-710695.010.0%0.912

601-625613.010.0%1.280686-720703.010.0%0.810640-680660.010.0%1.115

560-600580.09.4%1.2140.483635-685660.010.0%0.9860.202583-640611.510.0%1.2210.321

559 or less525.08.6%1.752635 or less592.09.9%1.417583 or less535.010.0%1.421





832 or more859.010.0%0.672-0.151845 or more857.010.0%0.800-0.141810 and up837.519.7%0.656-0.344

803-832817.510.0%1.027830-845837.510.0%0.919765-809777.020.1%0.795

767-803785.010.0%0.823814-830822.010.0%0.740715-764739.520.8%0.911

739-767753.010.0%1.036798-814806.010.0%0.733645-714679.520.2%1.066

720-739729.510.0%0.775779-798788.510.0%0.855Below 645600.019.2%1.5930.593

691-720705.510.0%1.000757-779768.010.0%0.889

668-691679.510.0%1.041730-757743.510.0%0.993

637-668652.510.0%1.023695-730712.510.0%1.143

602-637619.510.0%1.2510.301643-695669.010.0%1.3000.464

602 or less571.010.0%1.350643 or less600.010.0%1.628





750 and up795.021.3%0.783-0.217755 or more775.08.9%0.767-0.218780 and up815.016.8%0.637-0.363

685-749717.025.8%0.900732-754743.09.3%0.798745-779762.013.7%0.715

630-684657.019.6%1.083714-731722.59.6%0.859710-744727.013.9%0.734

560-629594.519.3%1.150698-713705.59.9%0.969670-709689.515.0%0.807

Below 560520.013.9%1.2000.200682-697689.510.3%0.922635-669652.012.1%0.909

666-681673.59.7%0.978590-634612.011.2%1.241

647-665656.010.5%1.070530-589559.59.8%1.3570.945

625-646635.510.2%1.107Below 530495.07.5%2.533

592-624608.010.7%1.1220.223

591 or less562.010.8%1.324

Sheet2

1997 NAIC/Tillinghast Study of 9 Companies' Data

Loss Ratio Relativity of the Best and Worst 20% of Credit Score

Co1Co2Co3Co4Co5Co6Co7Co8Co9Avg

Best 20%-38%-29%-19%-15%-14%-34%-22%-22%-36%-25%

Worst 20%48%20%32%30%46%59%20%22%95%41%

Sheet3


Sample Equation: .4591 - 0.053 * (Account Years) + 0.037 * (Number of Late Payments) + 0.026 * (Law Suits) - 0.075 * (Credit Limits) + 0.025 * (Number of Collections) - 0.038 * .

500 900

Credit Score RevolutionTr

ansf

orm

to F

inal

Sc

ale

Pred

icted

Loss

Rati

o

135%125%

110%115%

100%90%80%

70%

140%

55%50%

57%61%

64%

80%

74%

86%

120%

90%

OverallLoss Ratioof 68%

Better than Average Accounts

Average Accounts

Below Average Accounts

Decile

1 8 9 105432 6 7

Pred

icted

Loss

Rati

o

135%125%

110%115%

100%90%80%

70%

140%

55%50%

57%61%

64%

80%

74%

86%

120%

90%



Average Accounts


Decile

1 8 9 105432 6 7

Payment pattern information, account history, bankruptcies/liens, collections, inquiries, bad debt/defaults

Formula scoring or rule-based scoring Industry scores vs. company proprietary scores

Credit Score- A composite score that usually contains 10 to 40 credit characteristics


Credit Score Revolution

Large scale multivariate scoring using external data sources, a classic example of advanced data analytics applications

A significant behavior economic characteristic translated into a powerful Auto insurance class plan factor

Brilliant marketing approach for credit score:Benefits/ROI are measurable and lift curve can be translated into bottom-line benefitBlind test and independent validation can be done to verify the benefit

Pred

icted

Los

s Rat

io

135%125%

110%115%

100%90%80%

70%

140%

55%50%

57%61%

64%

80%

74%

86%

120%

90%



Average Accounts


Decile

1 8 9 105432 6 7

Pred

icted

Los

s Rat

io

135%125%

110%115%

100%90%80%

70%

140%

55%50%

57%61%

64%

80%

74%

86%

120%

90%



Average Accounts


Decile

1 8 9 105432 6 7

Why is credit score so successful?


The Evolution of P&C Data AnalyticsAdvanced analytics along with leveraging large amount of internal and external data have become mainstream over the last 20 years in several industries within financial services. Property and Casualty insurance has been one of the leading industries in the integration of advanced data analytics into core operations.









3rd Party Data

Marketing and Sales

Claims Data

Weather

CustomerData

PolicyInformation

Coverage Information

AgencyInformation

BillingData

Claims DataLossesFrequencyTiming / PattersLoss Control DataFraud / Lawsuit

Agency InformationRetentionRecruitingProfitabilityAdjusted Premium RatioNew Business VolumeContinuing Education

3rd Party Data

Business CreditPersonal CreditCrime StatisticsTraffic Patterns / StatsCLUE / MVRCredit BureausReal EstateGeographic/Geo-codingDemographicPsychographicBureau Data SourcesConsumer / LifestyleEnhanced Census

Campaign, PromotionCust Response ScoresCust Segmentation

Data Analytics Transformed P&C Insurance Industry Insurance industry contains large amount of data and is ideal for DM and PM

applications. Information age provides a wealth of external and 3rd party data sources

Billing / Payment HistAccepted ApplicationsRejected Applications

Billing Data

Marketing Data

Coverage and Policy Data

Policy InformationApplication InformationProduct CoverageInsureds Information


Data Analytics Transformed P&C Insurance Industry

Adoption of a wide range of new and powerful modeling and data exploration techniques- GLM, Neural Networks, Decision Trees, Clustering Analysis, MARS, etc.

1

X1

X3

X2

Z1

Z2

Y

1a11

a12

a21

a31

a321

b1

b2

a22

a01

a02

b0

FREQ1_F_RPT 0.500

TerminalNode 2

Class = 1Class Cases %

0 2508 57.41 1859 42.6

N = 4367

LIAB_ONLY 0.500

TerminalNode 3

Class = 0Class Cases %

0 7591 96.51 279 3.5

N = 7870

NUM_VEH 4.500

Node 4NUM_VEH

N = 20844

Node 1NUM_VEHN = 57203

0 20 40 60 80 100

-2-1

01

2

x

yy = 0.29 + 0.02*x

0 20 40 60 80 100

-2-1

01

2

x

y

y = 0.29 + 0.02*x - 0.086*max(0,x-35

0 20 40 60 80 100

-2-1

01

2

x

y

y = 0.29 + 0.02*x - 0.086*max(0,x-35) + 0.084*max(0,x-65)


DATA SOURCES

`

Business Rules Engine

External Data

Internal Data

Applicants Data

Agency Data

MODELING PROCESS

SCORE FOR EACHPOLICY

DRIVE BUSINESSDECISIONS

Data Aggregation+

Data Cleaning

Evaluate and Create Variables

Model Development

You learn why

Scoring Engine

`


..etc


85% of our new products are automated with predictive modeling, which enhances underwriting consistency and makes it easy for our commercial agents to do business with Safeco

SafecoFully launched Customized Pricing, our predictive pricing model that automatically provides the most appropriate price for a new small business submission

Achieved overall written premium growth of 4%.

Hartford

Commercial lines NPW grew 3% for Q2, driven by $83M in new commercial business, up 13%compared to Q2 2006.

The dramatic improvement is a direct result of our multidisciplinary WC improvement strategy and predictive modeling

Selective

The quality of our property, WC, auto, and specialty mix is continuing to improve as we use predictive analytics

Agents are giving us preferred shelf space vs. weaker competitors

Hanover

Increased ease of use through faster decisions, streamlined processing, and expanded account rounding

Reduced system quote to issue time through dynamic questions, improved agent interface, and automated UW.

Travelers


Strong performance in our core Property & Casualty Operations via portfolio optimization, data-driven underwriting, cross-sell, claim excellence, and catastrophe exposure management

CNA

P&C Insurers Public Statements on the Benefits of Predictive Modeling


The Evolution of P&C Insurance Data AnalyticsAdvanced analytics along with leveraging large amount of internal and external data have become mainstream over the last 20 years in several industries within financial services. Property and Casualty insurance has been one of the leading industries in the integration of advanced data analytics into core operations.








- 14 - Copyright 2017 Deloitte Development LLC. All rights reserved.Copyright 2015 Deloitte Development LLC. Proprietary and Confidential. All rights reserved.

Today and Future: Many Disruption ForcesThe P&C industry continues to experience disruption due to various driving forces:

Technology

Big Data and Analytics

Sharing Economics

Exponential Disruption

AI and Robotic Automation

Process

Digital and Mobile Reality

- 15 - Copyright 2017 Deloitte Development LLC. All rights reserved.Copyright 2015 Deloitte Development LLC. Proprietary and Confidential. All rights reserved.

Perception: Analytics evolving LINEARLY. Reality: Occurring EXPONENTIALLY

Indu

stry

Impa

ct O

ppor

tuni

ties

N

ew In

nova

tions

1995 2010 2025

ActuarialModels

InsightEconomy

2020+

Digital Enterprise

Internet of Things

Analytics asa Disruptor

2014-2018+

Machine Learning / AI

Crowd-sourcingAnalyticsApplied

2013-2016

Big Data

AnalyticsAware

2009-2013

CloudComputing

Data ScientistsAnalyticsas R&D silo

1995 - 2009SocialMedia

Smart Phones

Today and Future: ..and Accelerating at an Exponential Rate


Today and Future Technology Reshaping Every Industry

Mobile App

Drivers DNA

Night Driving

Speeding

Fatigue

Traffic Condition

Telematics


Today and Future: More Access to Data for Insurance Companies

http://www.data.gov/

http://www.data.gov/


Today and Future Big Data, More Data, More Modeling


Today and Future: A Core Strategic Capability

Customer

Workforce Finance

Supply Chain

Sector-Specific

Pricing and UW

segmentation analytics

Safety and Loss

Control analytics

Performance metrics

identification, design and

benchmarking

Customer analytics

Property and

Casualty Analytics

Claim Modeling

FraudDetection

Salvage Rate Modeling

Loss Control Modeling

Warranty Modeling

Underwriting Modeling

Pricing Modeling Account

Modeling Premium Audit

and Leakage Modeling

Demand Modeling Cross Sale and

Upsale Modeling Marketing Mix

Modeling Customer

Segmentation and Retention Modeling

Lifestyle Base Analytics

Customer Lifetime Value Analytics

Workforce Analytics

Workers Safety analytics

Business Information Analytics

Competitive Analysis

Increasingly Data Analytics Applications on Every Aspect of Insurance Operations and Customer Service


Lessons Learned from the P&C Analytics Journey

What Analytics IS NOT

A Black Box approach by quants only

Replacement for people

A shining complex math equation

A single variable magic bullet

Actuarial and/or systems projects

One of the many projects

Model complexity drives results

Short live hype

POC

What Analytics IS

An approach with transparency and communications

Tools and capability for underwriters and actuaries

A wide range of internal and external data is gold

Relationship among many variables is power

Business and strategic initiatives

Senior management support and all hands on deck

Implementation drives results

Stay for a long time and will get even bigger and better

Mature and proven to be impactful


For Life Insurance Industry: Growing Applications for Data Analytics

Customer

Workforce Finance

Supply Chain

Sector-Specific

Pricing and UW

segmentation analytics

Safety and Loss

Control analytics

Performance metrics

identification, design and

benchmarking

Customer analytics

Life Insurance Analytics

Fraud Detection: Application fraud

for non-smoker discounts

Fabricated death claims discount

Fake applications for agent rewards

Application Triage Analytics: Using external data

to identify preferred customers for fast track underwriting

Significant time and cost saving and more accurately placing policyholders risk status

Group Life Mortality Assessment: Large scale modeling

on group life insurance mortality experience

Goal is to develop more accurate pricing tool on exposure rating in addition to experience rating

Identify additional pricing factors beyond traditional factorsInforce

Management Analytics: Retention models

to identify insureds with potential high risk of lapse

Identify cross sale, up-sale opportunity to increase business and retention of existing policyholders.

Growth in Life Insurance Applications for Data Analytics


Learning and innovation go hand in hand. The arrogance of success is to think that what you did yesterday will be sufficient for tomorrow-William Pollard

Fruits for Thoughts


Q&A

Cover pageWu

2017 predictive analytics symposium - soa€¦ · 2017 predictive analytics symposium . ... blind...

Documents