The Keys to Analytical Innovation

By Scott G. Stephenson

Making the wold a better place

Thirty years ago, almost all insurers rated their personal auto policies using a fairly standard set of rating factors and traditional actuarial techniques. But beginning in the mid-1990s, a few carriers began exploring the use of predictive analytics to create rating models using new types of variables. If they could more accurately predict the likelihood of a loss and its magnitude, these carriers could more accurately set premiums — resulting in an advantage over competitors to attract and retain customers. After analysis revealed a correlation between credit score and expected auto losses, the forward-thinking carriers started factoring that data relationship into their traditional rating variables.

The organization that's able to capture information, analyze it with specific goals in mind, and implement change quickly is the one that will win in terms of profitability, market share, or both.

The early adopters of predictive analytics — the extraction of meaningful, actionable insights from raw data — gained a dramatic competitive advantage. Those who incorporated the new data relationship increased their share, while those who didn't were acquired or downscaled or otherwise left the market. In only ten years, Progressive rose from the eleventh largest personal auto carrier to the third largest. But more than 100 other carriers — a third of the total number of groups writing auto insurance — disappeared.

The characteristics of personal auto made it particularly well suited to this type of analysis — a huge, highly competitive market with a large number of standardized products and consumers for whom price is an important part of the marketing mix. However, each of the following components had to be present for change to come about:

  • reliable, readily available data
  • advanced computational resources
  • analytical talent
  • business knowledge

Data and Computational Resources
In recent years, advances in technology and computing power have greatly increased our ability to collect and manage vast amounts of data. These technological advancements represent a key factor, since much of the time and expense of developing a predictive model is related to the process of gathering relevant data.

Today, several industries — most notably healthcare — are formulating plans to improve the availability of relevant data. Vehicle telematics data will provide new information about how people drive. As we identify new predictive data, we also need to develop an infrastructure that enables reliable and even industrywide access to the data. Even more opportunities for analytical innovation will become possible as data sources and their associated infrastructures mature.

Analytical Talent
No matter how comprehensive, a data collection is only useful if it's analyzed, assimilated, and developed into insights that can lead to smart business decision making. However, over the past decade, those who do this work have become an increasingly scarce resource because demand for analytical professionals has increased throughout the business world. For the next round of innovations to occur, we must attract more of these professionals to our industry.

Insurance is fortunate to have a cadre of mathematically skilled professionals — actuaries — who can be trained in the discipline of predictive analytics. In fact, actuarial organizations have been adding more predictive modeling knowledge to their curricula over the past several years. Many actuaries have taken the opportunity to add predictive modeling to their skill sets. But there is still much the insurance industry can learn from other industries that have been incorporating predictive modeling into their business processes for a longer period of time. So we must also recognize and encourage analytical professionals with experience in other industries to join with us.

Business Knowledge
Cutting-edge analytical tools don't eliminate the need for sound business judgment. Rather, analytic exploration allows insights to be implemented in a way that enhances their impact on business performance. In the case of personal auto insurance, credit information was used in underwriting long before the first insurance specific credit models were developed. Underwriters had learned there was a relationship between an individual's credit history and the cost of providing insurance to that individual. This insight didn't begin as a formal statement of statistical prediction but grew from the experience of professionals who knew the business and recognized that a new source of information was useful in improving their estimates of risk.

The development of any analytic innovation begins with business knowledge. Understanding the business enables professionals to identify patterns that lend themselves to further exploration.

Putting Information into Practice
Today, the organization that's able to capture information, analyze it with specific goals in mind, and implement change quickly is the one that will win in terms of profitability, market share, or both. Research shows that high-performance businesses have a much more developed analytical orientation than other organizations. They are five times more likely than their lower-performing competitors to view analytical capabilities as core to the business.

Developing predictive analytic capabilities is mandatory. But crunching numbers is useless unless execution is swift, sound, and creative. The catch-22 is that robust analytic solutions often depend on the costly collection of large amounts of data, which favors large carriers over small ones. But implementing and executing innovation quickly may favor smaller carriers, who tend to be able to move faster than larger competitors.

Future Opportunities
Analytic gains continue to be made in the personal auto insurance market, and insurers' rating models are incorporating much of the easiest lift. Personal auto writers have largely segmented themselves into the predictive analytics "haves" and "have-nots," driving consolidation in this line of business as the "have-nots" get acquired or leave the field. The application of predictive modeling to the personal auto line of business is relatively mature, but new data sources such as vehicle telematics are emerging to form a whole new level of analytics.

The next step is to apply those same analytic techniques to other lines of business where untapped opportunities abound. Homeowners insurance is one likely candidate due to the size of the market and relatively standardized products. Some commercial lines, such as workers' compensation, businessowners policies, and commercial auto, are also attractive analytic environments.

Instead of looking backward to examine what happened, predictive analytics helps us answer the questions, What's next? and What should we do to improve efficiency, reduce risk, and increase growth? It's an organizational and management lesson that can make the difference between thriving and struggling to remain viable.

Scott G. Stephenson is president and chief operating officer of Verisk Analytics.