predictive analytics and accelerated underwriting survey ...€¦ · predictive analytics and...

Predictive AnalyticsandAccelerated Underwriting Survey Results

Al Klein

October 6, 2017 IAA Mortality Working Group

Agenda

Background

Results

2

Predictive analytics

Accelerated Underwriting

Concluding thoughts

Background

▪Survey was conducted in June/July 2016 of US companies

▪We initially had fewer responses than we wanted so we called companies we knew had implemented programs and asked them to participate▪Response to this follow up was good and we believe most of the companies that had a program when we conducted the survey participated

▪Responses were received from both direct companies and reinsurers who helped implement programs

3

Background (cont’d)

▪Goal was to learn about company practices on three timely issues:

▪Predictive analytics

▪Accelerated underwriting

▪“The use of tools such as a predictive model to waive requirements such as fluids and a paramedical exam on a fully underwritten product for qualifying applicants without charging a higher premium”

▪Enhanced underwriting – Insufficient response

▪“The use of supplemental information (e.g., criminal history, credit rating, prescription histories) and a predictive model to refine the underwriting process for a simplified issue product”

▪Avoided questions on proprietary information to maximize participation

4

Background (cont’d)

▪Started both sections of the survey with a large question to establish what was:

▪Implemented

▪Being worked on

▪Not worked on or considered

▪Focus of all subsequent questions was on the programs implemented by the respondents

5

Caveats

▪Original survey is out-of-date as additional companies have implemented new programs

▪However, I believe the information is still good and useful for both those with programs and those considering new programs

▪ I will be covering results at high level

▪Please find complete survey at:

▪https://www.soa.org/experience-studies/2017/predictive-analytics-underwriting/

6

Predictive AnalyticsSurvey Results

Predictive Analytics

Implementation Choices

Implemented

Working on and

Plan to implement within 1 year

Plan to implement within 1-2 years

Plan to implement longer than 2

years

Not sure if will implement

Not currently working on but

Considering it

Considered it and/or worked on it

but decided not to do it

Not considering it

Quick Summary of 2015 PA Results

34 companies responded to the survey

26 of these companies implemented one or more

PA programs

9

117 PA programs were implemented

Two companies implemented the most PA programs

(12 each), others implemented 1-10 programs

Predictive Analytics – Marketing

Marketing Programs

Program ImplementedWorking

on

Not working

on but

considering

Not working

on and not

considering

Total

Customer more likely to buy 12 6 6 5 29

Cross selling 10 2 7 9 28

Target market determination 9 8 4 7 28

Up selling 9 3 7 10 29

Customer less likely lapse 7 9 5 8 29

Customer health profile 5 5 11 7 28

Agent selection/hiring 4 6 4 9 23

Predictive Analytics – Underwriting

Underwriting Programs

Program ImplementedWorking

on

Not working

on but

considering

Not working on

and not

considering

Total

U/w risk class 12 9 7 5 33

Deciding on u/w

requirements9 11 8 2 30

Stretch criteria for

selecting u/w class5 4 10 13 32

Business decisions 1 2 4 18 25

Table shave 1 1 1 21 24

Predictive Analytics – Post-Issue Mgmt

Post-Issue Management Programs

Program ImplementedWorking

on

Not working

on but

considering

Not working

on and not

considering

Total

In force mgmt. – pre-lapse 7 6 7 9 29

Targeted conversion 5 2 7 12 26

For term, post-level premium

term conservation mgmt.2 7 8 11 28

Agent monitoring/mgmt. 2 6 11 7 26

In force mgmt. – post-lapse 2 5 8 11 26

In force mgmt. – Other

customer interaction1 2 4 8 15

Other types of PA programs that have been implemented

13

1Marketing (4) – Attract new reinsurance business, Prospecting models, Identifying prospects, UL vs. Term Prospecting

2

3

Underwriting (2) – implemented another type of underwriting PA program, Working on something

Post-issue Management (8) – Implemented another, working on other, or considering another type of post-issue management program (6), Ongoing claim study, Considering for business considerations

Sources/types of data used to develop PA Models

14

1

Vendor

(17)

2

Financial

(16)

3

Lifestyle

(13)

4

Application

(12)

5

Internalexperience

(12)

Individuals/Areas involved in developing PA models

15

1Marketing – Internal Actuary, Marketing, Data scientist/statistician

2

3

Underwriting – Internal Actuary, Internal Underwriter, Marketing

Post-issue Management – Marketing, Data scientist/statistician, Internal Actuary

Other Interesting findings

▪Most PA programs were implemented within the last few years, but some PA marketing programs were implemented earlier

▪Most PA programs were implemented as a pilot and many of the underwriting and post-issue management programs remain as a pilot

▪Most PA programs impacted only 0-10% of the overall business and none impacted more than 75%

16

Top Obstacles in Developing PA Models

17

1

Data Sources

(20)

2

Agent Buy-in

(13)

3

Internal User

Buy-in

(13)

4

Implement-ation

(12)

5

Designing/ Building

the Model

(12)

Accelerated UnderwritingSurvey Results

Accelerated Underwriting

Accelerated Underwriting (AU) Programs

ImplementedWorking

on

Not working

on but

considering

Not working on

and not

considering

Total

10 12 1 3 26

Accelerated Underwriting Program Limits

▪Maximum issue age ranged from 35 to 85 and most common was 60

▪Maximum face amounts ranged from $100K to $3M, with most common $1M

20

Accelerated Underwriting Decision-making

Data sources used for AU decision-making

22

1

MIB Checking Service

(7)

2

MVR

(7)

3

Rx History

(7)

4

Application

(6)

5

Lifestyle

& MIB IAI

(5 each)

Most important data sources for Accelerated Underwriting decision-making

23

1

Rx History

(6)

2

Application

(6)

3

MVR

(5)

4

MIB Checking Service

(4)

Data sources used for Risk Class decision-making

24

1

MVR

(7)

2

Rx History

(7)

3

Application

(6)

4

MIB Checking Service

(6)

5

Financial

(5)

Most important data sources used for Risk Class decision-making

25

1

Rx History

(7)

2

MVR

(6)

3

Application

(5)

4

MIB Checking Service

(5)

5

Personal History Report

(4)

Individuals/Areas involved in developing Accelerated Underwriting programs

26

1 Internal Underwriter (all 8)

2

3

Internal Actuary (7)

Internal Marketing (4)

Other Interesting findings

▪5 of 9 accelerated underwriting programs were implemented as a pilot program and one remains as a pilot program

▪4 of 9 companies randomly check some applicants to test their assumptions and/or model

▪4 of 8 use predictive analytics in the decision-making process for AU programs

27

Other Interesting findings (cont’d)

▪7 of 8 indicated time to issue decreased

▪6 of 8 indicated they were not sure if mortality changed since implementation of the AU program

▪7 of 8 plan to expand their AU programs

▪4 of 8 indicated that their reinsurers participated in the AU program

28

Biggest challenges encountered in developing AU programs

29

1 Data sources (4)

2

3

Justifying cost/benefit analysis (4)

Implementation (3)

Concluding thoughts

Both PA and AU programs are growing at a rapid pace and I expect that to continue over the next several years.

I also expect to see new methodologies and hybrid approaches emerge over this same time period.

I believe this is a great time to be a PA actuary and to offer creative and constructive solutions.

30

Thank you

Al Klein

(312) 499-5731, al.Klein@milliman.com