1 of 25

Data Analytics in Insurance

Jerry Tuttle, FCAS, CPCU, ARM, AIM, ARe

Adjunct instructor Math & Data Analysis, SNHU

Presentation to Eastern NC Chapter of CPCU, Sep 18, 2025

2 of 25

The big question:

Can the decisions of an experienced expert be supplemented and improved by data analysis?

Experienced manager Art Howe, and a non-athlete data analyst.

3 of 25

My bio:

BA Math, Queens College, NYC
MA Math, University of Virginia
Fellow of the Casualty Actuarial Society
Chartered Property Casualty Underwriter
42 years as Casualty Actuary in NYC, NJ:

C&F, J&H, M&G Re, St Paul Re, Platinum Re, Ren Re

MA Educational Technology, NJ City Univ
MS Data Analytics, Southern New Hampshire Univ
15 years as Adjunct instructor math & data analytics
Volunteer Data Analyst for several non-profits

4 of 25

Agenda:

What is data analysis?
What do data analysts do?
What are some real-life data analysis examples?
What if you hired me as your first data analyst?
What about AI?
Questions and conclusion

5 of 25

What is data analysis?

The intersection among math, computer sci, & domain knowledge, to extract meaningful insights from data, translating into tangible business value.

6 of 25

What do data analysts do?

Define data-driven business problem to solve
Design databases to capture essential information
Design reports that summarize results
Design visualizations that summarize results and provide real-time information to management in easy- to-understand formats
Design models that forecast amounts

or probabilities

7 of 25

Some data analysis techniques

1. Regression finds best fitting line to data.
2. Decision tree is like a flowchart that helps make choices by asking a series of yes-or-no questions, leading step by step to the best decision.
Analogy to game “Twenty Questions”. “I’m thinking of a US President …”
Decision tree chooses the best question to ask ea time:

the question that best splits the data for the next question

3. Many other techniques.
Each technique makes diff assumptions about the data.

8 of 25

Real-life data analysis examples

Target Stores identifies a teenager as pregnant before she told her parents.
Amazon, Netflix, etc. suggest what you might like next.
Gambling casinos know what freebies will keep you at the casino longer.
Sports coaches check data before making decisions, such as what is prob. of success on fourth and one?

9 of 25

�Visualization example:�Call Center Dashboard

10 of 25

If you hired me as data analyst …

I would visit your various department heads,

e.g., underwriting, claims, human resources.

Try to understand what your department does.
Try to get you to share some of your data driven business problems that you have not had the time or expertise to solve.
Suggest some data-oriented solutions.
Ask if you think the decisions of an experienced expert can be supplemented and improved by data analysis?

11 of 25

Insurance Rating Classification

For insurance rating, we group (hopefully) similar customers into classes and charge an average rate for the class. Classification is rarely perfect.

Before classification After classification

12 of 25

Underwriter’s Biz Problem: �Which applicants to write?

If rates are adequate overall, and rates are adequate by class, does an underwriter accept every applicant?
No. Underwriter tries to identify the better than average applicants within a class.
Can the decisions of an experienced underwriter be supplemented and improved by data analysis?
Examine data variables that are not rating variables.
Classify into “Less likely to have claim” vs “More likely”.

13 of 25

Where my data came from?

I do not have any company’s real data. If I did, I would never share it publicly.
I used publicly available data from www.kaggle.com and www.github.com
This is based on real data, but it is artificial, not necessarily US, and old.
Think of it as illustrative. Please do not go back to your desk and use it to make real decisions.

14 of 25

Which applicants to write?

However, regression analysis makes some questionable assumptions of the data. .

15 of 25

Which applicants to write?�0 = probable no claim, 1 = probable claim

16 of 25

Which applicants to write?

.

17 of 25

Claims Biz Problem: Help identify fraud

Vehicle Category = {Sedan, Sport, Utility}

18 of 25

Help identify fraud

19 of 25

Help identify fraud

20 of 25

HR: Help identify employee turnover

21 of 25

HR: Help identify employee turnover

22 of 25

HR: Help identify employee turnover

23 of 25

More ideas about variables

● Telematics in cars uses technology to collect and transmit data about your driving habits.

My car knows when I am eating a muffin while driving !

● Telematics in homes collects data on thermostats, water leaks, security, etc. and transmits data about your household habits.

● Telematics in other lines of business?

● Can the decisions of an experienced expert be supplemented and improved by data analysis?

23

24 of 25

What about AI?

● Data analysis is about building models to extract meaningful insights from data.

● Artificial intelligence is a broader field of creating models that can perform tasks requiring human-like intelligence.

● AI uses data analysis models, including language models, to build human-like behavior.

● To some extent both DA and AI models learn with additional data.

● Will AI ever be able to evaluate intangible characteristics of a policyholder or claimant?

24

25 of 25

And in conclusion

● Check those policy exclusions!

Any questions? You can find me at: fcas80@yahoo.com

Credits:

Template: https://www.slidescarnival.com/cordelia-free-presentation-template/216#preview

Cover slide designed by slidesgo / Freepik

25