Skip to Content
Big Data and Predictive Analytics

Insurance Company Data-Driven Opportunities

Dale Halon | March 4, 2016

On This Page
Globe on top of financial charts and graphs

In this first article on big data and predictive analytics, the vision of possibilities for data-driven companies to focus their analytical resources is laid out. Each one has its own data source(s), data approach, operational considerations, and analytic methodology to be applied. In addition, all of these facets of potential solutions can vary by company, since every insurer does business differently.

My mother wanted me to be a doctor. Since my handwriting was too neat, I opted for a career in insurance, which, thankfully, has been a solid path for many years. My first job was as an underwriter trainee, and I worked my way through the chairs to an executive role in 15 years, later making the transition to sales. While it's been a rewarding career for me, I cannot help but think what life would have been like if I had opted for another fork in the road along the career path.

One doesn't have to look too far to notice the demand for data scientists, data managers, predictive modelers, actuaries, and the like. With merely one "zig" toward a different corner of the insurance world instead of the "zag" I chose, I might have crossed into a land of opportunities with enough runway to double the longevity of my career. That doesn't mean you shouldn't reap the benefits of my hindsight—you should!

Predictive Analytics

Predictive analytics began its commercial run through insurance in 1993 when Fair, Isaac and Equifax introduced Property Loss Score, the first consumer credit-based score for insurance. Twenty-three years later, it seems like we're barely applying analytics throughout our enterprises. Really, we've hardly touched the surface as an industry.

Over the years, I've listened to insurers about their business objectives and have been an evangelist for transforming to a data-driven culture. The early and continuing focus of data-driven decisions has been the use of analytics to support pricing and underwriting strategies. Nearly all companies with even basic analytical capabilities are using analytics to support pricing. Generally, the exceptions have been largely skewed to the larger companies with cubicles filled with analysts using or researching all sorts of modeling techniques to derive predictive signal out of any data they have or can get their hands on.

Of course, there are outliers. I've met with large insurers that are still late in the "data driven" game compared to similarly-sized peers. I've also met small insurers who use the little data they have to support models for any business decision possible. All insurers have the same opportunities and challenges, but some, of course, have more data to work with and/or ample resources to approach the problem than others.

Common Objectives for Insurer Analytics

Below is a list of "How do I?" questions insurers are always asking and, thus, are ripe for the Moneyball approach. In some cases, it's possible to seek answers through data and analytics. And, for some business questions, a less quantitative strategy may be the only tool available. This list is not a pipe dream when looking at the industry as a whole. All of these questions are being explored by an analyst somewhere, and I am sure I have overlooked a few. This list proves there is job security for those with training, ability, and the right tools to solve them.

  • Identify the best and worst risks using a multivariate approach, as measured by loss ratio. Find the better risks in groups believed to be poor-performing and the worst risks in segments traditionally being profitable. Quantify the questions, how good is good?, and, how bad is bad?
  • Explain the characteristics of "best" and "worst" risks.
  • Identify the best and worst retaining business.
  • Determine which customers to retain. Ask, am I retaining the right ones?
  • Provide loss ratio projections by type of risk.
  • Identify profit trends for any policy, rating variable, or market segments.
  • Identify trends of geographic profitability and/or retention.
  • Figure out which low-retention profile policies are not worth trying to retain and which high-profit policies are worth spending extra effort to retain.
  • Develop profit projections of your larger and smaller agents/brokers.
  • Ask, how can I describe to my agents/brokers or my marketing department the types of risks that are the most and least profitable?
  • Determine which agents/brokers to use and how to best reward and retain them.
  • Quantify individual agent/broker profit trends, retention, and production trends.
  • Ask, how do I know in advance whether the changes I made to my portfolio are taking me in the desired direction?
  • Attempt to figure out when a new piece of business will reach profitability.
  • Estimate the likelihood of a large claim at quote or new business submission.
  • Assess under what circumstances external underwriting information (such as loss history, motor vehicle records, or inspections) should be obtained.
  • Determine which claims seem small but have a high likelihood of being larger than expected (by company definition).
  • Determine which claims are likely to have low severity (by company definition).
  • Identify which claims have a very small, as well as very high, probability of being fraudulent.
  • Figure out which claims are likely to have attorney involvement.
  • Outline the factors that drive reserving practices.
  • Ask, how much expense will I incur on a policy?
  • Assess when consumer or business credit information makes a difference in the projected-loss cost of a new or renewal policy.
  • Determine which resources should be spent on subrogation activities for specific claims.
  • Ask, which telephone calls are going to take the most resources?
  • Identify which claims take the most company resources.
  • Try to determine which sales reps are going to be the most productive.
  • Develop strategies for assigning a nurse or case manager to a workers compensation claim.
  • Identify what geographies support profitable expansion.

Start Small

The paper "Thinking Big About Big Data In Insurance" (Chunka Mui, Jan 22, 2016), published on Forbes' website, is an interview in which Tom Warden, formerly chief data officer at AIG Life and Retirement, states

First, companies need to recognize that Job 1 is leveraging data to optimize the profit machines they already have. Best practice today is to embed more predictive modeling in all parts of the value chain. You don't necessarily need Big Data to do that. You need smart people, a disciplined approach and a culture of cooperation to monetize data-driven insights.

I'm a fan of Mr. Warden's approach of starting small. Tackling big data right away is a major challenge. Insurers have been more successful getting "little data" right first.

Strive for Quick Implementation

The biggest challenge for every analytic objective is implementation. Early successes are realized with less complicated implementations, probably for obvious reasons. It's hard to get buy-in around the company until results are evident. Quick wins are still wins! That means easy IT implementation (more on that in a future article) and a solid fit for the ultimate users of the analytic result (another article topic).

Given how long predictive analytics have been embedded in our industry, it still amazes me to see resistance from end users. My belief is the benefits to them are just not communicated well enough, and the end users are not part of the process from the beginning. Who else is better equipped to find the best use of the analytic result and to make sure the result matches business requirements?


Our business is here for the long haul, as are most of our careers. There is still so much to accomplish in our industry, and competition continues to accelerate with the availability of data and technology. Meanwhile, I will continue on my established career path and remain a casual observer of the road not taken.

  • I shall be telling this with a sigh
  • Somewhere ages and ages hence:
  • Two roads diverged in a wood, and I —
  • I took the one less traveled by,
  • And that has made all the difference.
  • Robert Frost*

*originally published in Mountain Interval, Henry Holt and Company, 1916.

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.