2017 Predictive Analytics Symposium

Transcription

1 2017 Predictive Analytics Symposium Session 20, Marketing and Distribution Applications of Predictive Analytics Moderator: Priyanka Srivastava Presenters: Matt Olson Xiaojie Wang, FSA, CERA SOA Antitrust Compliance Guidelines SOA Presentation Disclaimer

2 Analytical Support for Distribution in an Independent Channel August 2017 Prepared by: Matthew Olson, VP Marketing, Athene USA 1

3 About Athene USA Subsidiary of Athene Holding, Ltd. Headquartered in West Des Moines, IA Serve all 50 states Products include Fixed Indexed Annuities and Institutional Products Subsidiaries include Athene Annuity & Life Assurance Company Delaware-domiciled company, formerly known as Liberty Life Insurance Company; Athene Annuity and Life Company Iowa-domiciled company, formerly known as Aviva Life and Annuity Company; Athene Annuity & Life Assurance Company of New York New York-licensed insurance company, formerly known as Presidential Life Insurance Company. Majority of distribution is through Independent Marketing Organizations (IMOs) and their Independent Agents Currently accessing the Banks and Broker/Dealer market With growth, complexity is increasing 2

4 Athene Analytics Organized within the Marketing Department Primary focus is supporting Sales/Distribution & Marketing Retention Sales Desk / Wholesalers Marketing 3

5 The Big Picture: Goal of Athene Analytics Establishing capabilities Acquiring access to and understanding of data Bringing on appropriate analytical tools Finding the right analytical talent to staff the team Predicting behavior Demonstrate that we can accurately predict customer behavior Develop a suite of predictive models Influencing behavior Create initiatives to influence behavior using predictive models Design experiments which statistically measure effects (A/B test, Control) Impacting key business decisions Enhance decision-making at a more macro level Become more impactful on the overall profitability of the organization 4

6 Key Premise: Small Nudges To Behavior Can Result In Large Differences In Profitability Predictive analytics enables us to improve the economics of our business by targeting and influencing certain behaviors. Desirability of Behavior % Profit Dist. Expected Actual Best 20% Superior Behavior/ Economics Superior Behavior/ Economics Assumed 60% Assumed Behavior/ Economics Targeted Influence Assumed Behavior/ Economics Worst 20% Worst Behavior/ Economics Worst Behavior/ Economics 5

7 Analytical Modeling Enables Behavioral Influence Predictive modeling provides a view into the variability within each agent group. This enables the micro-targeting of the right individuals to influence their behavior. Yesterday we treated all customers the same Today we know they differ High likelihood to surrender. Send product benefit material to build confidence & stickiness. Explore informational call regarding benefits of policy Traditional block-level assumptions assign the same activation likelihood to each customer Analytical models calculate individual behavioral predictions for each customer Low likelihood to surrender. Treat as BAU no need to increase efforts. 6

8 Applications: Proactive Retention Predictive modeling enables us to accurately identify policies at risk of surrender, allowing us to tailor our messaging and tactics. Annuity Surrender Lift Curve Decile 9 & 10 is the current focus of our proactive retention activities Result: a 35% decrease in surrender vs. control group 7

9 Application: Agent Conversion Newly Contracted Agents 76% Fail to Produce a Single Policy 13% Produce only 1 Policy 11% Produce 2+ Policies Developing a Conversion Model enabled us to prioritize agents for conversion campaigns (calls & ) Result: a 17% increase in conversion vs. control group 8

10 Application: Agent Winback Using a combination of a response model and our Sales teams preferred segmentation we were able to further increase opportunities to win back producers Result: a +32% increase in conversion over a control group 9

11 Application: Predicting Net Promoter Score Accuracy of Predicting Promoters Accuracy of Predicting Detractors Incorrect 32% Correct 68% Incorrect 30% Correct 70% In addition to enabling targeting of Promoters & Detractors, we also create reasons why customers fall into these NPS categories Below is a partial example of Reason Codes for Annuity customers Primary reason for Promoter status is because they are risk adverse 2 nd reason is because the product attribute resonates well with the customer 3 rd reason is because they love their Agent Predicting these groups enabled us to proactively target dissatisfied clients/agents with messaging and offers to increase referrals and retention Result: a +18 NPS Points over control groups

12 Application: Producer Segmentation Producer Offices = Producer Segments? Segmenting our producers allows us to better tailor training, marketing, recruit, etc. Result: mixed

13 Key Distribution Partner Data Analysis From time-to-time we provide consulting services to our key distribution partners via a Data Analysis Viewed as strategic partnership and value add service Purpose of Analysis General trending and analysis Shows trajectory of business, provides historical perspective and anticipates future Demographic & economic profiling of clients Describes your client base using things like gender, age, income, net worth, home value, etc. Geospatial analysis Identifies opportunities for expansion due to underserved markets, or saturated markets Data mining for opportunities and risks Conduct a number of mini-analyses which generates further insights Total Annual Premiums Total Annual Premiums per 1MM Pop 12

14 Thanks! Contact Info: Matt Olson

15 Predictive Analytics in Marketing Applications Xiaojie Wang, FSA, CERA

16 The Protection Gap Swiss Re 2012 Study $20 Trillion Total $378k Average Deficit Gap is Growing

17 Project STING Overview In less than a decade, the US life insurance protection gap the difference between the amount of insurance carried versus needed has increased by 10 percent to more than $20 trillion. Swiss Re, along with many of our clients, are acutely aware and concerned about the gap and implications of not doing more to close it. To address this growing divide, Swiss Re's US Life & Health unit embarked on an ambitious consumer research initiative: Project STInG (Solving The Insurance Gap). This initiative began in December 2012 and involved partnering with Maddock Douglas (MD), a globally recognized innovation agency. The primary focus of the project was to gain a deeper understanding of potential consumers and to identify, develop and prioritize new ways to close the protection gap. How can we, as an industry, make insurance easier to understand, access and buy? Those were the questions we tackled throughout the rigorous process which evaluated clear insights based on consumer need, developed and tested ideas that meet those needs, and monetized innovative solutions. 3

18 Are Consumers Satis fie d With The ir Exis ting Life Insurance Coverage? Less coverage than needed Right amont of coverage More coverage than needed 0% 10% 20% 30% 40% 50% 60% How can we reach them more effectively? *Facts About Life 2017, LIMRA 4

19 Simple Illustration of Predictive Marketing Application Data. Consumer data Product data Model. Propensity to Buy/ Own Segmentation model Recommendation model Propensity to be healthy/ persistency Communication. Economic behaviour Channe l optimization Product Sold. 5

20 Demographic Financial Be havior & Life style Marketing Inte raction Product catalog Characte ristics Targeted market Policy terms Exclusions Limits Consumers Products Right Offers 6

21 Swiss Re Survey Data Objectives In-depth understanding of consumers attitudes and perceptions Prope nsity to buy life insurance Parameters minutes Questionnaire N = online interviews Target yr. old consumers HH income of $25K+ (OR) Full time students who earn less than $25K who (1) at least share re sponsibility for finance-re late d decisions (AND) own some form of insurance (Auto, Health, Home or Life) 7

22 Likelihood of Buying Life Insurance Are existing life insurance owners more likely to buy life insurance in the near future? 50% 40% 30% 20% Life Insurance Owner Non-Life Insurance Owner 10% 0% Very Likely to Buy Likely to Buy Less Likely to Buy Unlikely to Buy 8

23 Likelihood of Buying Disability Insurance Are existing life insurance owners more likely to buy Disability insurance in the near future? 60% 50% 40% 30% 20% DI Insurance Owner Non-DI Insurance Owner 10% 0% Very Likely to Buy Likely to Buy Less Likely to Buy Unlikely to Buy 9

24 Tailor the Buying Process to Reflect What Consumers Want What are the most desirable characteristics that consumers wish life insurers offered? 80% 78% 76% 74% 72% 70% 68% 66% Cost Transparency Information reconcile capability Personal information assessibility 10

25 Attitudinal Se gme ntations 11

26 Attitudinal Segmentations Among Existing Life Insurance Policyholde rs Curious & Accountable Apathetic & Conventional Confident & Controlling Carefree & Open The current life insurance ownership chart shows the result of traditional marketing on life insurance products, which focus on demographics and finances. The attitudinal distribution is quite even among different segments. 12

27 Attitudinal Segmentations Among Future Life Insurance Owne rs Curious & Accountable Apathetic & Conventional Carefree & Open Confident & Controlling The traditional marketing targets applicants by demographic factors and life triggers. It might miss the opportunity for behaviour factors. About 40% of people with high certainty of buying life insurance belongs to Curious & Accountable. 13

28 Life Insurance Purchase Likelihood Varies by Attitudinal Segment 5 Relative Difference of Purchase Likelihood Among Different Attitudinal Segmentions Curious & Accountable Apathetic & Conventional Carefree & Open Confident & Controlling Very Likely to Buy Likely to Buy Less Likely to Buy Unlikely to Buy 14

29 Data Used in Predictive Models Attitude & Behaviour Lifestyle & Interests Demographic Life events Life Insurance & Finances Marketing data Online profile 15

30 Life Insurance Current Ownership Model Among the top ten most predictive variables in this model, only one is related to attitude & behaviour and the rest are related to demographics and finance. About 70% (Accuracy) life & non-life owner insurance owners can be correctly predicted in this model This indicates that the current life insurance owner profile are driven by traditional marketing method which focus on demographics and finance. Life Insurance Owner Non-Life Insurance Owner F1 Score 70% 69% Recall 85% 59% Precision 59% 85% Accuracy 70% 70% AUC

31 Life Insurance Purchase Intention Model Nine of the top ten most predictive features belong to attitude, behaviour and lifestyle. About 7 9 % (Recall) of actual potential life insurance buyers are corrected predicted/identified. This indicates there are some untouched markets defined by attitude & behaviour. Predictive marketing model including those types of data can play an important role on new opportunities. Life Insurance Owner Non-Life Insurance Owner F1 Score 65% 71% Recall 79% 62% Precision 55% 83% Accuracy 68% 68% AUC

32 Illustration of Marketing Cost Before and After Assume the traditional marketing method targets 1,000 people without any distinction and the new marketing method only targets people who are predicted as Likely to buy. Traditional Marketing Marketing cost per person $ 10 # of people to market 1,000 Traditional cost $ 10,000 # of policies sold 415 Marketing cost per sold policy $ 24 Predictive Marketing Marketing cost per person $ 10 # of people to market 595 Traditional cost $ 5,955 # of policies sold 354 Marketing cost per sold policy $ 17 18

33 CASE STUDY L&H R&D develops a propensity to buy model for Cross sell Background: Swiss Re partnered with one strategic client to offer a comprehensive suite of services focusing on improving the end-to-end cross sell process. Overall aim of the partnership was to write more profitable business. A package of Predictive Analytics and Behavioural Economics was offered to the client. Objective: Increase sales based on a propensity to buy model on the existing consumers. Higher propensity to buy customers were selected as good leads for the product. The expertise of the Behavioural Research Unit is then applied for the launch and marketing of the product. o o o o Benefit for Direct Insurers? Write more profitable business Save on marketing expenses Improve their portfolio understanding Incre ase their customer understanding What did we learn? Other clients interested in replicating this initiative in other countries. A package including both predictive analytics and BE could be more powerful then predictive mode ls alone 19

34 This work indicates attitude & behaviour factors can play an important role on the propensity to buy life insurance. It might be another way to segment potential market other than the traditional demographic and life trigger events. 20

35 Questions? 21

36 22

37 Legal notice 2017 Swiss Re. All rights reserved. You are not permitted to create any modifications or derivative works of this presentation or to use it for commercial or other public purposes without the prior written permission of Swiss Re. The information and opinions contained in the presentation are provided as at the date of the presentation and are subject to change without notice. Although the information used was taken from reliable sources, Swiss Re does not accept any responsibility for the accuracy or comprehensiveness of the details given. All liability for the accuracy and completeness thereof or for any damage or loss resulting from the use of the information contained in this presentation is expressly excluded. Under no circumstances shall Swiss Re or its Group companies be liable for any financial or consequential loss relating to this presentation. 23