Skip to Content
Big Data and Predictive Analytics

Applied Data Science and Analytics in the Insurance Industry

Dale Halon | December 15, 2021

On This Page

After 17 years as an underwriter, manager, and leader for insurance companies, I was lucky enough to have become involved in "predictive analytics" in 1994 when I was an early evangelist of credit-based insurance scoring and operational analytics for the insurance industry. For those of you who are wondering, a loaf of bread cost more than a nickel in those days, but there was no such thing as a data scientist in our industry.

The math people were actuaries, and most had yet to have been exposed to analytics. My colleagues were talented operations research-based mathematical experts, many of whom hailed from academia and few with business backgrounds.

As my fellow modeling evangelical colleagues and I visited insurers, presented at industry conferences, and wrote articles, we found ourselves sharing these newly discovered applications with people much smarter than we were but who just hadn't been exposed to how these mathematical techniques could provide more insight into decision-making broader than actuarial concepts.

Data and Educational Opportunities

Actuarial science students at the undergraduate and graduate level continue to study loss reserving, ratemaking, statistics, and other applied math techniques to do the work actuaries have always done to be able to perform the necessary and traditional functions of actuaries. The difference now is the curriculum has expanded to include more newly developed tools, which can squeeze more decision-making power out of similar data. Now, add a heavy dose of more data being available and modern computing power and BAM!

Actuarial, finance, and mathematics students are much more fortunate today. The Casualty Actuarial Society not only provides credentialling for actuaries but also provides texts and supplemental education through advanced classes and specialized conferences. Analytics are a regular part of the learning objectives and testing for credentials. Additionally, The Institutes offers both the Chartered Property Casualty Underwriter (CPCU) and the Associate in Insurance Data Analytics designations, both of which include analytics in the curriculum. CPCU and The Institutes also offer webinars and focused conference seminars on a variety of subjects, including the business deployment of analytics. There are scores of articles and books in the public domain that share a deeper understanding of varieties of related subjects and are there for the taking for ones that have an interest.

These and other educational opportunities provide gateways to an insurance industry with more variety of careers than most any other field. While actuaries used to be "behind the scenes" and may not have been part of the business decision-making process, the advent of analytics requires the quantitatively focused person to more deeply understand the insurance business from a wide variety of perspectives, from the process of analytical tool creation to the ultimate application to enhance the effectiveness of business processes.

Credit-Based Scoring and Beyond

When analytics was first introduced to insurance industry decision-making, the initial purpose was to aid underwriting in the acceptance or rejection of risks, usually combined with other underwriting factors. In fact, a couple of insurers initially applied the early credit-based insurance score without consideration of other factors. I referred to this strategy as "meat axe underwriting." As you can imagine, this "underwriting rule" drove agents and regulators crazy to the point it didn't take long for insurance departments to enact regulations that require that a credit-based insurance score could not be the only reason for an adverse action.

I'd like to share a historical anecdote back to these times regarding a past coworker who had become the CEO of a small company that wrote single line auto insurance. Having completed a study to show the loss ratios of their book by ranges of credit-based insurance scoring, the CEO made the command decision to draw a line in the sand at a score level where they would nonrenew policies or reject them based solely on the score. I begged him not to do so for obvious reasons, but he went his own way, citing the results of the study. Suffice it to say that he was forced (by the market) to end this procedure. Ironically, this company only survived another couple of years and is now only a distant memory.

Our industry is no different than others who had been emboldened by these new decision-making tools and began to use analytics in personal lines pricing (rating tiers), then renewal decisions, customer acquisition (direct mail campaigns using the credit-based insurance score as prescreening criteria), agency management, and offerings of optional coverages/limits of liability, to mention a few.

It didn't take long for actuaries and data scientists to seek other applications for deploying analytics. These quantitative masterminds began sharing the potential of mathematic decision-making with the other functional areas in their companies. Fast forward to today, data scientists have since educated business company leaders of the possibilities of making better decisions in a multitude of functional areas.

Along the way, data scientists discovered new data sources, new software, and new techniques better suited to varieties of insurance industry decision points. Today, artificial intelligence (AI) is widely used for customer service and in other areas that used to be human activities. These decision tools can be available 24/7. Bots don't call in sick or ask for a raise. Only more complex situations are referred to humans to make the decision, perhaps with a recommendation from the AI tool.

Common Analytic Applications

Consider the main functions that insurers must engage in when doing business. Below are some of the more common applications for analytics (of any type) for an insurer.


  • How to message certain types of potential and current customers based on their propensity to buy a policy, get a quote, renew, etc.
  • How much should marketing expense be applied to each type of customer based on their lifetime value as a customer?
  • What is the next most likely product to be purchased by an existing customer?
  • What are the renewal offerings to make for each type of customer?
  • Is the prospect more likely to purchase from an agent or directly from the insurer?
  • What is the likelihood of a claim or certain types of claims for groups of prospects or customers?


  • How big is a new claim likely to be?
  • How long will the claim take to close?
  • What type of resources should an adjuster deploy to adjudicate the claim?
  • What is the potential for fraud or settlement cost creep for certain types of claims?
  • Which claims should be assigned to less experienced or more highly trained adjusters?


  • Which payment plans should we offer to specific types of risks?
  • What is the nature of loss development for types of risk profiles?
  • Which investments are going to be most profitable and most manageable for particular lines of business?
  • Capital budgeting
  • Procurement of funds
  • Working capital management
  • Allocation of investment funds

Agency Management

  • Is an agent likely to be more profitable than others?
  • Is a candidate salesperson likely to be more successful than candidates or meet expectations?
  • What is the likely profitability of an agent's book of business (renewal ratios, loss ratios, large loss prediction, etc.)?
  • What is the likely growth (positive or negative) of specific agents?

The above list is merely illustrative and examples of ideas experienced by the author. They do, however, warrant thought and discussion and are likely to be tailored to any specific insurer as their design, compatibility with other business goals, and regulatory approval.


My main message is to emphasize the potential broad use of analytics in an insurance company. The breadth of tactical strategies listed merely illustrate the needs of expert data scientists working in partnership with business and functional experts. Analytics generally do not fail because the math is flawed. Such projects fail due to flawed design, business value, or execution. The many aspects of data used in analytic models, creation of data sets, testing, legal review, effects on business operations, acceptance by end users/constituents, and projected results—along with procedural documentation for implementation and tracking—need to be put in place and monitored. Models need to be constantly measured for effectiveness and applicability as they age. All corporate constituents need to be versed in the strengths and weaknesses of the underlying analytics so they can use them as a tool. They must also have a deep understanding so they can communicate the message the tool is providing to others in the insurance delivery chain.

Related workaround design and implementation of analytic tools also illustrate the vast number of opportunities for students and persons seeking career expansion and advancement in our industry. With well-trained people and rigorous adherence to checks and balances of design, the returns on investment are vast for both insurers and team players. Students and those seeking upward or horizontal mobility in their careers should become familiar with these abundant opportunities to broaden their scope of "greener pastures."

Opinions expressed in Expert Commentary articles are those of the author and are not necessarily held by the author's employer or IRMI. Expert Commentary articles and other IRMI Online content do not purport to provide legal, accounting, or other professional advice or opinion. If such advice is needed, consult with your attorney, accountant, or other qualified adviser.