Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Session 79 PD, Analytics in Action: Case Studies and Lessons on the Use of Consumer
Analytics in Insurance Sales and Marketing
Moderator: JJ Lane Carroll, FSA, MAAA
Presenters:
Sarah R. Hinchey, FSA, CERA, MAAA Denise Olivares, ChFC, CLU
Jamie Yoder
SOA Antitrust Disclaimer SOA Presentation Disclaimer
JJ Carroll, Sarah Hinchey, Denise Olivares, Jamie YoderConsumer Analytics in Insurance Sales and Marketing25 October 2016
2
SOCIETY OF ACTUARIESAntitrust Notice for Meetings
Active participation in the Society of Actuaries is an important aspect of membership. However, any Society activity that arguably could be perceived as a restraint of trade exposes the SOA and its members to antitrust risk. Accordingly, meeting participants should refrain from any discussion which may provide the basis for an inference that they agreed to take any action relating to prices, services, production, allocation of markets or any other matter having a market effect. These discussions should be avoided both at official SOA meetings and informal gatherings and activities. In addition, meeting participants should be sensitive to other matters that may raise particular antitrust concern: membership restrictions, codes of ethics or other forms of self-regulation, product standardization or certification. The following are guidelines that should be followed at all SOA meetings, informal gatherings and activities:
• DON’T discuss your own, your firm’s, or others’ prices or fees for service, or anything that might affect prices or fees, such as costs, discounts, terms of sale, or profit margins.
• DON’T stay at a meeting where any such price talk occurs.
• DON’T make public announcements or statements about your own or your firm’s prices or fees, or those of competitors, at any SOA meeting or activity.
• DON’T talk about what other entities or their members or employees plan to do in particular geographic or product markets or with particular customers.
• DON’T speak or act on behalf of the SOA or any of its committees unless specifically authorized to do so.
• DO alert SOA staff or legal counsel about any concerns regarding proposed statements to be made by the association on behalf of a committee or section.
• DO consult with your own legal counsel or the SOA before raising any matter or making any statement that you think may involve competitively sensitive information.
• DO be alert to improper activities, and don’t participate if you think something is improper.
• If you have specific questions, seek guidance from your own legal counsel or from the SOA’s Executive Director or legal counsel.
3
Presentation Disclaimer
Presentations are intended for educational purposes only and do not replace independent professional judgment. Statements of fact and opinions expressed are those of the participants individually and, unless expressly stated to the contrary, are not the opinion or position of the Society of Actuaries, its cosponsors or its committees. The Society of Actuaries does not endorse or approve, and assumes no responsibility for, the content, accuracy or completeness of the information presented. Attendees should note that the sessions are audio-recorded and may be published in various media, including print, audio and video formats without further notice.
4
5
JJ CarrollSwiss Re
Sarah HincheyMilliman, Inc
Denise OlivaresLexisNexis
Jamie YoderPWC
Jamie Yoder has advised leading insurers on business design and transformation initiatives and leads PwC's global research program on the "Future of Insurance“. Mr. Yoder has met with hundreds of executives to help them understand and address potential opportunities and risks for the insurance industry.
Sarah Hinchey is a Predictive Analytics Strategist and consultant with Milliman. Ms. Hinchey has led cross-functional projects in data-driven marketing for life insurers in the US, Ireland, Spain, and the Netherlands. While in the Netherlands, she helped develop a cross-border data analytics program from the ground up.
Denise Olivares is a Marketing and Business Development Executive with a distinguished career designing and leading strategic product development programs. Ms. Olivares excels in identifying smart business opportunities, developing business infrastructures, and executing programs that maximize corporate growth and profitability.
JJ Carroll leads a cross-functional team working with insurance clients to shape the consumer experience around their insurance protection needs. Ms. Carroll is responsible for the coordination of market and consumer research including predictive modeling, behavioral economics and consumer engagement.
Analytical applications in marketing
Target• Lead qualification• Channel preference• Product preference• Likely to buy• Likely to lapse• Likely to qualify• Segmentation• Personalization
Campaign management• Response• 1st & 2nd year premium
goals• Product sales trends• Conversion
Nurture• Lifetime customer value• Propensity to lapse• Interventions to improve
longevity and persistency
• Cross-sell and up-sell
6
How can actuaries leverage data to help marketing teams build a robust view of customers?
7
How can actuaries leverage data to help marketing teams build a robust view of customers?
Finding the right talentTargeting the right customers
Increasing overall profitability of customers
Tailoring the right products/features
Internal and external data allows a more ‘holistic’ view of the policyholder
Life Events• Getting married• Buying a house• Having a child• Retiring
Income Statement• Salary• Expenses
1. Nondiscretionary 2. Discretionary
Balance Sheet• Assets
1. Home2. Financial assets
• Liabilities1. Mortgage2. Personal debt
Choices• Rational• Behavioral
1. Joint decision making
2. Financial literacy
Examples
Actuarial Data• Policyholder behaviour
assumptions Lapses Partial/Full
Withdrawals Annuitization Utilization of
Features Pricing Cash value
• Policyholder behaviour actuals
• Reasons for deviation between assumptions and actuals
• Claims history• Hedging strategies
• Fact: Financially stressed households lapse/withdraw during economic downturns.
• Action: Balance your customer portfolio by targeting the customer segments that are underweight
• Fact: Utilization of riders varies by socio-demographic, behavioural and health factors.
• Action: Analyze policyholder utilization of features and develop ‘product bundles’ that default to what each segment is most likely to use in the future (and sell accordingly)
• Fact: Certain segments of customers (e.g., risk averse) are more profitable and value risk reducing riders (e.g., LTC)
• Action: Policyholder behaviour analysis can determine the most profitable segments and their value under different economic conditions
What are some examples of how you’ve used analytics in sales or marketing?
11
Profiling your book can begin with the basics
Confidential © 2016 LexisNexis 12
Are you converting the age demographic you’re targeting?
Footnote: Sample conversion data for three U.S. life carriers, each operating nationwide with two or more individual life insurance product offerings.
Source: LexisNexis analysis of several hundred thousand records of in-force life insurance policies, spanning multiple lines of business and product types.
70+60–6950–5940–4930–39
<30A B C U.S. Population
100
90
80
70
60
50
40
30
20
10
0
10%2%
19%
25%
27%
17%
5%
24%
27%
20%
12%
12%2%
17%
33%
30%
10%
8% 16%
25%
21%
15%
19%
3%
33%40%
17%6%
2%
Company F
0%
2%
61%
15%15%
1%6%
2%0%
Company D
Carriers’ customers’ channel preference can vary widely
Confidential © 2016 LexisNexis 13
Footnote: Sample channel preference demographics for customers of three U.S. nationwide life carriers.
Source: LexisNexis analysis of several hundred thousand records of in-force life insurance policies, spanning multiple lines of business and product types.
ALL Direct Exclusive Exclusive/Direct Independent Independent/Direct Independent/Exclusive
43%31%
15%3%6%
2%
0%
Company E
Even within the Young Working Families segment, sub-segments with higher education shop and buy differently
Confidential © 2016 LexisNexis 14
Footnote: Middle-income, young working families with some college education or higher are more likely to shop for and buy life insurance.
Source: LexisNexis analysis of several hundred thousand records of in-force life insurance policies, spanning multiple lines of business and product types.
Less than High School High School Some College Bachelor’s Degree Graduate Degree
% o
f Seg
men
t
Young working families - mid income
General PopulationShoppers Buyers
Several life events occur more often in the Middle Market
Confidential © 2016 LexisNexis 15
Home
Renter to homeownerHome purchase in past 6 months
New or increased mortgage
Newly marriedNew child in
household
Worse wealthBetter wealth
Financial risk increase
% increase in trigger rate for middle market shoppers
Family
Finance
5%0% 10% 15% 20% 25% 30% 35%
Looking forward: Impact of data
16
Wearables Internet of things
Social MediaMobile
LexisNexis and the Knowledge Burst logo are registered trademarks of Reed Elsevier Properties Inc., used under license. Otherproducts and services may be trademarks or registered trademarks of their respective companies. Copyright © 2016 LexisNexis.
How can an actuary help predict which existing customers are likely to buy additional insurance?
17
We are Wombat Life Insurance Co.
18
We want to use our customer data in a more intelligent way
Where do we start?
Step by step approach to turn insights into action…
Business Drivers
Analytical Support
Data Gatekeepers
Step 1Identify and involve key stakeholders
Alternative approaches
20
Targeted Business Question
Data Exploratory Analysis Modeling Implement
Data Exploratory Analysis
Targeted Business Question
Modeling Implement
Starting with a question…
Starting with the data…
Step 2Gather and aggregate data
21
Policy admin systems Customer Relationship Management (CRM) systems
Other internal data sources External data sources
One single customer view across all products, distribution, and communication channels
Step 3Explore data for insights…
22
ProductNo other contract Term Life Disability
Universal Life
Fixed Annuity
529 Savings
Term Life 90% 1% 2% 0% 0% 7%Disability 85% 10% 0% 5% 0% 0%Universal Life 85% 0% 10% 0% 3% 2%Fixed Annuity 85% 10% 3% 2% 0% 0%529 Savings 75% 15% 5% 4% 0% 1%
Cross-sell 2011-2015
0
20,000
40,000
60,000
80,000
100,000
120,000
2011 2012 2013 2014 2015
Wombat : New contracts by business line
Term Life Disability Universal Life Fixed Annuity 529 Savings
$-
$20,000,000
$40,000,000
$60,000,000
$80,000,000
$100,000,000
2015 APE (Annualized Premium Equivalent)
Term Life Disability Universal Life Fixed Annuity 529 Savings
$180
$2000
$500
$5000
$2400
SituationTerm life sales are
declining
ComplicationDifficult to sell: 1% conversion ratio
GoalIncrease
conversions by cross-selling Term Life to select 529
Savings customers
…and define your campaign goal.
Some customers are more likely than others to respond positively to a cross-sell offer:
24
529 Savings
customers Term Life customers
Can we find these customers… …and move them here?
Step 4Build model on historical data
• Define target variable• Clean & prep data• Split data into training and
test sets• Fit model to training data• Evaluate model performance
on test data (compare predictions to truth)
• Select “best” model
25
By targeting the 30% of customers with highest propensity to buy, the Random Forest model captures 72% of the customers who actually bought, while the Logistic Regression model only captures 48% of the
customers who actually bought.
0%
20%
40%
60%
80%
100%
Buy
%
Selection %
Which model performed the best?
Random Selection Random Forest Logistic Regression
Interpret and explain model
Which variables have the most predictive power for determining whether an existing 529 Savings customer will buy Term Life?
26
Demographic
Behavioral
Trigger Events
Other factors…
Step 5Design & execute campaign• Objectives• KPI’s to measure success• Channel• Budget• Marketing materials /
call scripts• Target group• Exclusions• Campaign process flow
27
Oracle: “Process flow for campaign to telesales”
Score and rank customers based on propensity to buy…
28
0%
2%
4%
6%
8%
10%
12%
14%
1 2 3 4 5 6 7 8 9 10
Prop
ensit
y to
buy
Decile (customer ranking)
529 Savings customer propensity to buy Term Life
Average propensity to buy for entire base
…and deliver leads to sales force
• Keep score/group hidden
• Be mindful of exclusions
• Wait for results
29
Customer Score (Ranking)
Group
A 0.95 High
B 0.85 High
C 0.75 High
D 0.65 Control
E 0.35 Control
…
Step 6:Collect and Monitor results…
30
9,000Leads
9,000Contacted
4,500Reached
1,000Meetings
400Applications
360Conversions
1,000Leads
1,000Contacted
500Reached
100Meetings
15Applications
10Conversions
Target Group(HIGH propensity to buy)
Control Group(random propensity to buy)
1% Overall Conversion
Rate
4% Overall Conversion
Rate
100%
50%
20%
15%
67%
100%
50%
22%
40%
90%
…and evaluate the impact
31
Target group
Eligible contacted
Sales
50,000
40,000
15,000Top 3 deciles
10,0009,000 High1,000 random
4001% conversion
3704% high conversion1% random conversion
Before model
After model Wombat Life
Insurance Co. realized nearly the same level of sales while
reducing acquisition
costs by 75%
How can insurance companies build a framework for leveraging data in marketing?
32
Data / Predictive analytics framework
Confidential © 2016 LexisNexis 33
UnderwritingIn-force
Management ClaimsProduct / PricingMarketing
• Objectives• Outcomes• Considerations• Products
• Objectives• Outcomes• Considerations• Products
• Objectives• Outcomes• Considerations• Products
• Objectives• Outcomes• Considerations• Products
• Objectives• Outcomes• Considerations• Products
Data / Predictive analytics framework: Marketing
34
OutcomesKey
ConsiderationsObjectives
• Targeted cross sell / upsell / retention
• Data-driven segmentation: the right customer at the right time with the right message
• Enriched analytics and marketing program optimization
• Target the right customer at the right time
• Acquire proper data enhance target marketing
• Improve ability to conduct campaign analytics
• Enhanced distribution experience
• Increased sales
• Predictive models• Demographic data• Individual vs. Household• Non-FCRA• Contributory data• Campaign design and
measurement
LexisNexis and the Knowledge Burst logo are registered trademarks of Reed Elsevier Properties Inc., used under license. Otherproducts and services may be trademarks or registered trademarks of their respective companies. Copyright © 2016 LexisNexis.
What are the key considerations in the implementation of models?
35
Implementation of Models: Key Considerations
36
What techniques are we using?
What data do we need?
Do we have the right technology?
What business problems are we solving?
How do we build the right org & skills?
© 2016 PricewaterhouseCoopers LLP, a Delaware limited liability partnership. All rights reserved.
Implementation of Models: What business problem are we solving?
37
Strategy & Growth
Customer &
Marketing
Sales & Distribution
Products, Pricing &
Underwriting
Process & Operations
In-force Mgmt.
Capital, Risk &
Finance
1. How will changing consumer demographic and socio-economic shifts impact demand for our products?
2. How are key macro-economic and regulatory changes impacting growth opportunities?
3. Can we grow our business by creating partnerships with asset management firms?
4. Can we do a better job of identifying and valuing our best and worst customers?
5. How can we best use media and other communi-cations to acquire new customers or deepen relationships with existing customers?
6. What steps can we take to improve customer retention?
7. How should we structure the customer experience through each distribution channel, so as to maximize sales and profits?
8. What are the implications of refocusing distribution to lower-cost channels?
9. How do we improve the sales productivity and profitability of our agency force?
10. What is the growth opportunity of our products and services? How does this change under different bundling and pricing scenarios?
11. Can we optimize pricing by capturing new health data to apply to our underwriting process?
12. Do our product designs reflect the evolving needs of our customer base?
13. What operations or technology initiatives will reduce costs without limiting growth?
14. How do we optimize the multi-channel customer service experience for each of our segments?
15. What is the most cost-effective path for managing the flow of policies?
16. How can we improve the policyholder persistency?
17. How do we analyze mortality and morbidityresults to improve pricing?
18. How can we more quickly and accurately identify policyholder needs and behaviors?
19. How do we optimize our portfolio of investments given our strategy and external constraints?
20. How should our capital allocation strategy (such as voluntary reserves) respond to different economic or regulatory changes?
© 2016 PricewaterhouseCoopers LLP, a Delaware limited liability partnership. All rights reserved.
Implementation of Models: What data do we need?
38
Image
Audio
Text andNatural
Language
Video
Social andNetwork
UnstructuredData
Text and Natural LanguageQuestion AnsweringSentiment AnalysisCustomer Service Analysis
ImagesFacial Recognition
Object ClassificationDeep Learning
VideoPattern Recognition
Motion TrackingSecurity
AudioVoice RecognitionText-to-SpeechSentiment Analysis
Structured Semi-Structured UnstructuredInformation with a fixed data model or schema, typically stored in tabular format
Labeled or organized data with no guarantee of data model consistency
Information has no defined schema or structure
• Relational Database• Configuration Files
• JSON/XML• Excel workbooks
• Audio• Image/Video
© 2016 PricewaterhouseCoopers LLP, a Delaware limited liability partnership. All rights reserved.
Implementation of Models: What techniques are we using?
39
Trad
itio
nal
Em
ergi
ngStructured Unstructured
A/B/N TestingExperiment to find the most effective
variation of a website, product, etc
Natural Language Processing
Extract meaning from human speech or
writing
Complex Event Processing
Combine data sources to recognize events
Predictive Modeling
Use data to forecast or infer behavior
RegressionDiscover
relationships between variables
Time Series AnalysisDiscover
relationships over time
ClassificationOrganize data points into known categories
Simulation Modeling
Experiment with a system virtually
Spatial AnalysisExtract geographic or
topological information
Cluster AnalysisDiscover meaningful
groupings of data points
Signal AnalysisDistinguish between
noise and meaningful information
VisualizationUse visual
representations of data to find and
communicate info
Network AnalysisDiscover meaningful
nodes and relationships on
networks
OptimizationImprove a process or
function based on criteria
Deep QAFind answers to human questions
using artificial intelligence
Image/Video Processing
Identify objects and events from images
and videos
DataT
ech
niq
ues
© 2016 PricewaterhouseCoopers LLP, a Delaware limited liability partnership. All rights reserved.
Implementation of Models: Do we have the right mix of technology?
40
Structured Data
Warehouse
Data Warehouse Environment
Interlocking is critical for
success
Identify
Match
Log
Big Data Warehouse
Big Data Environment
Data Lake
Entities / Relationships(Graph DB)
Ontology(Shared Semantics)
User Interface
Create & Store
Data Transformation
Application(s)
Data aggregations / metrics, etc…
User Interface
Application(s)
Relational Database
Extract Transform & Load (ETL)
Consumer Insights Analysis
Innovation Labs
Extract Golden
Nuggets of Information
Model Conception, Testing and Promotion
ILLUSTRATIVE
© 2016 PricewaterhouseCoopers LLP, a Delaware limited liability partnership. All rights reserved.
Implementation of Models: How do we build the right organization and skills?
41
Business Functions Client Operations
Client Performance Data
Transaction Systems
3rd Party Data
Modeling Office
Modelers
Skilled resources that: •Structure Models•Obtain Data•Create and test models
Modeling Lead
Business Intelligence
Tools
Analytical opportunities identified jointly between business users and “Insight Managers” embedded in functions and business units
Insight managers work with Modelers to develop and test strategic models
Insights are interpreted and summarized for the business by Model Analyst to enable decision-making
Insights are validated and tested with the business and operationalized into repeatable reports/metrics as needed
Modeling Leadership prioritizes workload
Modelers acquire data and calibrate models
2 4
5
6
Corporate, Regional, Country,
and Product Insight
Mangers
“Insight Managers”
…
1
Corporate Functions
Regional
Business Units
Market Research
3
In-Country
Product• Insurance Market Data• Consumer Trends and Data• Macroeconomic Indicators• Etc.
© 2016 PricewaterhouseCoopers LLP, a Delaware limited liability partnership. All rights reserved.
Implementation of Models: Key Success Factors
42
Start from business decisions and questions to be answered1
Demonstrate ‘value’ through pilots before scaling2
Address ‘big data’ – but don’t forget to leverage ‘lean’ data3
Fail forward – institute a culture of test-and-learn4
5 Overcome ‘gut’ instinct to become a truly ‘data-driven’ culture
© 2016 PricewaterhouseCoopers LLP, a Delaware limited liability partnership. All rights reserved.
How can actuaries become marketable to the marketing department within their company?
43
44
Skills to developht
tp:/
/insid
ebig
data
.com
/wp-
cont
ent/
uplo
ads/
2014
/01/
John
sHop
kins
.jpg
http
s://
arch
ive.
bigd
atau
nive
rsity
.com
/bdu
-wp/
wp-
cont
ent/
uplo
ads/
2015
/06/
Data
Cam
p_lo
go_t
op.p
nght
tps:
//w
ww
.kag
gle.
com
/sta
tic/im
ages
/site
-logo
.png
http
s://
ww
w.d
ataq
uest
.io/im
ages
/bra
nd/d
ataq
uest
_log
o.sv
g
Q&A
Sample of marketing/marketing analytics termsList SegmentationSelecting a target audience or group of individuals for whom a mail or email message is relevant. A segmented list means a more targeted and relevant campaign,
Campaign ROIRevenue from a campaign divided by marketing spend on that campaign.
DeliverabilityThe ability to get an email into the intended recipient’s inbox.
PersonalizationAdding elements to an email that are personalized based on information known about a target. It could refer to addressing the recipient by name, referencing past purchases, or other content unique to each recipient.
Match rateCan mean the same thing as hit rate when referring to appends of data to a name/address record. Match rate can also mean how well a marketer can match a response to a campaign. For example, a prospect may come to a website and ask for a quote but the marketer wont know if they received an email or postcard in a campaign unless they entered the website from a link provided in that material.
Email sharing (forwarding rate)The percentage of email recipients who clicked on a “share this” button to post email content to a social network, and/or who clicked on a “forward to a friend” button.
Some of the terms above are adopted from Hubspot: http://blog.hubspot.com/marketing/hubspot-google-analytics-glossary#sm.0000fhechx7c4fbrzka1m2sllxgds
Suggested Sources for additional informationFrom The American Marketing Association: https://www.ama.org/resources/Pages/Marketing-Dictionary.aspx
Other references:http://blog.hubspot.com/marketing/hubspot-google-analytics-glossary#sm.0000fhechx7c4fbrzka1m2sllxgds
http://www.fathomdelivers.com/glossary/