INNOVATION IN RISK SELECTION OF LIFE BUSINESS: SUPPORTING GROWTH... FOR NOW AND FOR THE FUTURE

Transcription

1 INNOVATION IN RISK SELECTION OF LIFE BUSINESS: SUPPORTING GROWTH... FOR NOW AND FOR THE FUTURE Bogotá, April 2015

2 Agenda 1. Risk assessment and types of underwriting 2. Use of technology in risk assessment: automated underwriting 3. Lifestyle and socioeconomic risk factors: smart underwriting 4. Innovation in risk assessment: big data and predictive underwriting 5. Conclusions 21/04/2015 2

3 RISK ASSESSMENT INTRODUCTION

4 Some considerations about risk assessment (medical underwriting) of new business Contribution of Medical Underwriting Keep claims ratio within actuarial assumptions considered in the pricing Allow to sell insurance products at affordable price, pushes limits of insurability Underwriting process Risk selection: separation of STD and substandard risks, detection of uninsurable risks Risk classification: quantifies the excess of mortality/morbidity of substandard risks Challenges of Medical Underwriting To be efficient, fast, pragmatic, affordable and accurate To increase productivity, to be adapted to distribution channels and products specifications To incorporate technology and innovation without sacrificing quality and results Types of Underwriting Traditional Automated Predictive Smart 21/04/2015 4

5 Possible Models of Underwriting Past experience, technology and innovation make possible different models of UW How? Direct Indirect Smart Variants? Based on Traditional Automated Paper info + Human Software Predictive Automated + Preferred + Predictive Analytical Data Software + analytical information Focused on Are you ill? You are well! Well being status Duration Enable 1 to 4 week 10 to 20 minutes UW manual Rulebook Immediate Algorithm 3 to 5 minutes 2º Generation rulebook + selected variables 21/04/2015 5

6 TRADITIONAL RISK ASSESSMENT

7 Processing of new life business: Traditional Model Sales Application Risk assessment Distribution Technical data Call Center Unterwriting clerk Application assessment 1 Clean case Underwriter Reflexive questions Evidence request Evidence assessment 2 3 Simple medical u/w Qualified medical u/w Reinsurer Traditional underwriting Specific case mgmt. 4 Specific case management Back office processes Enabler Application form design Workflow system Underwriting manual 21/04/2015 7

8 Traditional Risk Assessment Manual underwriting Performed by human underwriters and medical doctors Based on disclosures and evidences (medical and financial) Technology: use of Underwriting Manual in Internet, Multivariate Risk Calculators Appropriate for High sum assured Complex and sophisticated cases Disadvantages Expensive, slow, limited productivity (an issue in developing insurance markets) Low risk segmentation, difficult to adapt to different distribution channels Human error and subjectivism cannot be excluded Little use of technology and innovation Medical evidences Non medical limits required for sum insured and age and not for risk profile 21/04/2015 8

9 Medical tests used in traditional underwriting Physicians statement Blood pressure, heart rate, full physical exam ECG, Exercise-ECG, stress echocardiogram Fasting blood sugar and HbA1c Erythrocyte sedimentation rate Cholesterol, Triglycerides Liver function tests Kidney function tests Urine analysis HIV-test, Hepatitis B/C-serology Chest X-ray, Ultrasound, etc. 21/04/2015 9

10 Screening test in traditional underwriting Medical tests in underwriting are usually used as screening tests. Screening is a strategy used in a population to detect a disease in individuals without any signs or symptoms of that disease. Insurance companies believe, they improve their risk management by using screening tests. But are they right? 21/04/

11 Cost effective use of screening tests Diabetes: Pre-test probability Sum insured Cost effective Screening test Pre-test probability: high Cost of test (HbA1c): low Accuracy of test: high Cost-benefit: low sum insured Heart failure: Pre-test probability: low Cost of test (BNP): high Accuracy of test: medium Cost-benefit: high sum insured Repeat this for each target disease you want to screen for 21/04/

12 Just an example: screening test for these two persons in traditional underwriting 45 year old, slim, asymptomatic man, sum assured year old, hypertensive, obese, with occasional chest pain, sum assured Pre-Test Probability 0.3 % Pre-Test Probability 40 % An expensive stress ECG will be asked because sum assured Inefficient risk management Risk of false positive test results! No stress ECG will be asked for low sum insured! Low control of antiselection Inefficient risk management 21/04/

13 USE OF TECHNOLOGY IN RISK ASSESSMENT AUTOMATED UNDERWRITING

14 Risk selection of new business, levels of complexity Types of underwritten cases + High and/or sophisticated risks (with the reinsurer) 4 Complexity of U/W, sum assured Qualified Underwriting (substandard risks, preferred lives) Medical information, medical evidences, specialized information, etc.) Simple Underwriting (Application form with disclosures, special risks) Ad hoc additional questionnaires, etc.) 1 Clean Cases (Application form with no disclosures) /04/

15 Business Benefits of Automated Underwriting Faster acceptance decisions base on case data and disclosures More cases taken up due to faster policy issue and reduced attrition rates Can process up to 5 times more new business with no increase in customer service staff Consistent underwriting based on rules covering different risk scenarios The same results every time Underwriters concentrate on rules efficiency and high value cases Excellent customer service driven by quick decisions Creates data base for business analytics purposes Reduced costs Fewer underwriters need to process larger numbers of cases Eradicate re-work, reduces the requirement of medical evidences Faster to market with New products, New channels, New rules Improve business efficiency, drive growth.. reduce and control costs 21/04/

16 Processing of new life business auw Model on POS Sales Application Risk assessment Distribution Technical data Base questions Reflexive questions POS underwriting 1 2 Clean case Simple medical u/w Call Center Unterwriting clerk Underwriter Reflexive questions Evidence request Evidence assessment 2 3 Simple medical u/w Qualified medical u/w Reinsurer Traditional underwriting Specific case mgmt. 4 Specific case management Back office processes Enabler Application design Rules engine Rules set / rules script Workflow system Underwriting manual 21/04/

17 Fundamental requirements for success in auw State of the art software Reliability, continuous development, user friendly, flexible success stories, IT support Homemade solutions are risky and cheap solutions expensive Implementation project Involving IT stream (integration of the solution with customer s environment) Rule stream (rulebook, underwriting philosophy, parameter specification) Excellent support on rule development and rule maintenance Center of competences Utility of business analytics Business case A reliable business case is necessary In automation, the choice of software & rulebook is strategic 21/04/

18 Munich Re the 5TH Generation N E W B U S I N E S S C H A N N E L S Brokers Agents Call Centres Direct New products Flexible Graphical User Interface Interfaces INTERVIEW SERVER RULES DESIGNER BUSINESS ANALYTICS UW WORKBENCH DATA ACCESS SERVER Interfaces Policy Systems Case Management Claims Systems Agency Systems C O R E S Y S T E M S Paper 18

19 SMART RISK ASSESSMENT

20 Mix and smart approach Simplifies automated risk assessment process Generates specific base questions segmented by age groups, risk profile, risk types and distribution channel Assessment of complexity levels 1, 2 and 3 Accurate and short Reduces the requirement of medical evidences Preferred approach based on health status, lifestyle and predictive reliable criteria Considers predictive elements to determine risk profile and well being status Determine a mortality class favorable or unfavorable according to health status, lifestyle per age group and socioeconomic aspects Simplified auw + preferred + predictive = Smart underwriting 21/04/

21 Example on risk selection approach for 2 different age groups Age of applicant Who they are? What they do? years of carrier with grown children that are finishing high school, beginning college or to work. High or Senior management level. Career stabilized, financial responsibility for children & parents, Save/ Investment money for retirement. Established or declining career with children leaving the house. Reduction of financial obligations. Pensions funds. Less pressure at work; fewer hours worked; more balanced life; more time for leisure, walks, reading and friends. Remarriage. Which diseases/problems can have? Development of chronic diseases related to lifestyle like diabetes mellitus and cardiac problems, breast cancer. Depression, loss of bone density, cancer, excessive alcohol consumption. What risk they are exposed Divorce, depression, excessive alcohol consumption. Rising cost in health. Premature death or disability due to important chronic diseases reaching its peak. Specific questions Treatments, hospitalization, consume of medicines, diabetes, cancer, cardiac diseases, anxiety, depression, lipids, prostate or mammography results. Hospitalizations, diabetes, cancer, heart, arthritis, mental disorders, results of mammography or prostate tests, bone density; results of lipids test. 21/04/

22 Lifestyle 21/04/

23 Lifestyle Lifestyle: a trend in risk selection Habits and behaviours resist to change specially when changes require efforts What we learned in the childhood (eat, saving, discipline, etc.) influence habits in adulthood Current lifestyle predicts our future lifestyle Unhealthy lifestyle sooner or later will impact on mortality and morbidity The opposite is true. Current lifestyle Predicts future lifestyle Affects mortality and morbidity 21/04/

24 Wearable devices: miniaturized computers Applications of wearable: health, wellness and medicine Health apps (about 31%) target chronically ill patients to manage diseases like diabetes Health and fitness (about 28%) apps target interested consumers. Medicine apps (about 14%) to provide health care remotely Big explosion since 2009: about apps in health & fitness and medical section of apps stores. It allows the collection and analysis of data (big data) Insured persons can earn premium discounts by providing the data measured by their wearable. 21/04/

25 Wireless biometric data - Wearable's Opportunity for proactive risk management? 21/04/

26 Socioeconomic factors Socioeconomic factors are determinants of health according to WHO Educational attainment, income, relational life and housing impact on mortality and morbidity Educational level and income are correlated and are the most important socioeconomic criteria for risk assessment. In some countries like UK postal code is an important criteria as well Work environment, education system, health care system, water and sanitation, housing, agriculture and food production and unemployment determine living and working conditions 21/04/

27 Determinants of Health Determinants of health that should somehow be considered in risk management General socioeconomic, cultural and environmental conditions Social and community networks Individual lifestyle factors Living and working conditions, health care and educational services Opportunities to find a job, to study, play sport, found a family, etc. Habits (eat, drink, smoke, saving, physical activity, etc.) Personal characteristics (age, gender, constitutional factors) 21/04/

28 Social determinants of health Pathways to health and illness Social structure Material factors Work Early life, Genes, Culture Social Environment Health Behaviors Psychological Pathophysiological changes Brain Neuroendocrine and immune response Well-being, mortality, morbidity 21/04/

29 PREDICTIVE RISK ASSESSMENT

30 Digitalization and Big Data Transition from analogue to digital Data exchange and data storage Big Data and data analytics Some figures: Around 90% all data in existence were created over the past 2 years! The world produces 2,5 trillion bytes of data every day! Reliable and efficient support that ensure high quality data exchange 21/04/

31 Data Analytics Analytics can be described as the extensive use of data, statistical and quantitative analysis, explanatory and predictive models, and fact based management to drive decisions and actions. The analytics may be input for human decisions or may drive fully automated decisions. Data analytics is still inn its infancy, yet making rapid progress Source: Competing on Analytics: The New Science of Winning Thomas H. Davenport and Jeanne G Harris. 21/04/

32 Big Data: Volume Growing quantity Around 90% of all data in existence were created over the past two years Data Variety Internal / external; structured / unstructured; sensible / not sensible Velocity Quickening speed (almost real-time) Value? Increasing pool of data: an opportunity for life insurers 21/04/

33 Application of Big Data in Life Insurance Risk management Pricing Underwriting: Predictive Underwriting Claims: to prevent fraud or estimate propensity to submit claims (i.e. in health and disability) Trends detection Processes optimization Calculation of pricing Generalized Linear Models Reserving Marketing and Product Development Product development according to consumer profile and life stages Cross/up-selling and Marketing campaign Client segmentation Purchase behavior In this presentation we will focus on Predictive Underwriting 21/04/

34 A Definition for Predictive Underwriting An application of predictive modelling which: 1) uses known data to determinate pattern in underwriting outcomes of previous life insurance applications; 2) and then uses this to estimate the underwriting outcomes of a new set of target lives (without underwriting them). Allowing you to pre-underwrite life insurance applicants 21/04/

35 Motivation for Predictive Underwriting Technological progress drives changes in customer behaviour shift to new sales channels: bancassurance, direct sales, retail huge amount of data on consumers available (internet, social networks, credit card data, ) Predictive Underwriting may open new opportunities for life insurers Address new target groups via new sales channels Use available data to simplify UW process Be innovative and differentiate from competitors 21/04/

36 Principles of Predictive Underwriting 1) use data to explain the pattern in underwriting outcomes of previous protection applications: Portfolio of fully underwritten applications Customer data Link data using name, DOB and address Portfolio of fully underwritten applications Customer data Analyse the linked data using statistical models to determine how the customer data fields correlate with UW decision 21/04/

37 Principles of Predictive Underwriting 1) use data to explain the pattern in underwriting outcomes of previous protection applications: 100% Segment the lives according to their UW score and Calculate the split of UW decisions for each segment Distribution Underwriting of Score Decisions 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Postponed / Declined High Rating Low Rating Standard Segments 21/04/

38 Principles of Predictive Underwriting 2) use this to estimate the underwriting outcomes of a new set of target lives (without underwriting them): 100% 90% Portfolio of new potential customers Customer data Distribution of Underwriting Decisions Underwriting Score 80% 70% 60% 50% 40% 30% 20% Postponed / Declined High Rating Low Rating Standard 10% 0% Segments Use the model to segment the new customer linked data. And take the UW decision splits already defined as the estimated likelihoods. 21/04/

39 Principles of Predictive Underwriting In summary, predictive underwriting allows us to segment new lives according to their UW decision likelihood (without UW): UW outcome score No segmentation Decile UW outcome score Strong segmentation Decile 21/04/

40 Predictive Underwriting its applications Predictive Underwriting allows you to enhance your offering in 3 ways: UW process / product Marketing approach Tailor for each segment the: UW questions Exclusions Acceptance criteria Product features Price Or offer a simplified UW process to the better segments 21/04/

41 Predictive Underwriting its applications Predictive Underwriting allows you to enhance your offering in 3 ways: UW process / product Marketing approach Price Who to market to (could exclude less profitable segments) What marketing messages are used for each segment 21/04/

42 Predictive Underwriting its applications Predictive Underwriting allows you to enhance your offering in 3 ways: UW process / product Marketing approach Price The price can be: differentiated by segment the same for all segments (but overall reduced if not marketing to all segments) 21/04/

43 How to Control Antiselection? Essential to find suitable criteria for pre-selection comprehensive data base per market required appropriate and accurate model to be developed for individual markets Careful choice of the medical questions in simplified UW targeted to selected risks and distribution channels Choose for individual markets appropriate distribution channels Account for antiselection in pricing determine appropriate loadings Uncertainty in UW decision needs to be accounted for in higher risk premium 21/04/

44 Principles of Predictive Underwriting - The Process - 1. Develop model to predict UW decision: Consumer Data (portfolio data, internet,social networks, credit card data,...) Data analysis Model Identification of important risk factors on UW decision Predict UW decision 2. Apply model to consumer data of potential clients Consumer data of potential clients Model Predict UW decision Likelihood of being Standard Rated Declined Pre-selection Approach consumers with a high likelihood of being standard simplified UW reduce medical questions to a minimum UW decision Model based preselection allows to reduce UW effort 21/04/

45 Principles of Predictive Underwriting - Notes - The UW decision is modelled, i.e. the likelihood of being standard, rated or declined, not the actual risk (e.g. mortality, disability incidence) itself. Advantage of this approach: Consumer data only need to be linked to a traditionally underwritten portfolio, link to claims data (e.g. death, disability incidence) is not required. Implication: Approach implies that traditional UW is working fine: aim is to reproduce the traditional UW decision, not to improve it. 21/04/

46 Principles of Predictive Underwriting - Notes - Comprehensive data base required Possible data sources: Portfolio data (primary or reinsurance data) External data (e.g. by external providers) Data contents of interest Technical, Socioeconomic Life style Biopsychological Data need to be linked to underwritten (or claims) data Consideration of legal aspects important! 21/04/

47 Predictive Underwriting - conclusions - If you are considering to implement Predictive Underwriting: 1 Understand what data you have 2 and how useful it is Consider all possible sources of data: 3 Source the modelling skills In-house Commercially available Distributor-specific 4 Be able to tailor an underwriting process to specific applicants And consider costs, data protection and permissions. 5 Make your assumptions explicit (and test them) 21/04/

48 Predictive Underwriting - conclusions - If you are considering to implement Predictive Underwriting: 1 Understand what data you have 2 and how useful it is Very comprehensive data base required 3 4 Source the modelling skills Be able to tailor an underwriting process to specific applicants Do some quick initial data analyses to see what sort of segmentation you believe could be possible 5 Make your assumptions explicit (and test them) 21/04/

49 Predictive Underwriting - conclusions - If you are considering to implement Predictive Underwriting: 1 Understand what data you have 2 and how useful it is Source the modelling skills Be able to tailor an underwriting process to specific applicants Make your assumptions explicit (and test them) As well as in-house and reinsurer resource, remember that there may be expertise commercially available: Marketing firms often have this expertise Modelling consultancies 21/04/

50 Predictive Underwriting - conclusions - If you are considering to implement Predictive Underwriting: 1 Understand what data you have 2 and how useful it is 3 Source the modelling skills Having segmented the potential customer base, you need to be able to design a robust process that suits their specific characteristics 4 Be able to tailor an underwriting process to specific applicants 5 Make your assumptions explicit (and test them) 21/04/

51 Predictive Underwriting - conclusions - If you are considering to implement Predictive Underwriting: 1 Understand what data you have 2 and how useful it is Source the modelling skills Be able to tailor an underwriting process to specific applicants Make your assumptions explicit (and test them) Express the key assumptions you re making and test them by designing and executing successful experiments Try to prevent post-project neglect. 21/04/

52 Conclusions Underwriting: a multifaceted and changing task Risk assessment must be integral, efficient, fast, pragmatic, affordable and accurate The underwriting process of insurances companies should fit distribution channels & product specifications. The use of technology (internet manuals, risk calculators, automated underwriting) and innovation (predictive underwriting, smart underwriting) is a necessity to grow, differentiate and compete. Technology make possible preferred underwriting Decisions on technology and innovation are crucial for life insurance! 21/04/

53 THANK YOU VERY MUCH FOR YOUR KIND ATTENTION! Daniel Barroso