Predictive modeling and predictive analytics are no longer news in property/casualty insurance.

For years now, innovative carriers have been mining public and private data to identify demographic, geographic, and economic factors that correlate with insurance loss.

Factors found to be reflective of loss experience have been incorporated into complex algorithms developed to underwrite and rate policies with ever-greater segmentation and discrimination among risks.

“There is a race on now in analytics,” said Joel Brown, vice president and director of personal lines for State Auto Ins. Co., Columbus, Ohio, in a presentation at the 2010 AAIS Main Event conference, April 11-13 in Fort Myers, Fla.

According to Brown, a company investment in the effort to develop and refine predictive models “is not only worth the effort, it’s key to survival.”

“I view the use of analytics today as merely ‘table stakes’ for insurers to be in the game,” said Greg Hansen, leader of the actuarial research and predictive modeling group at Westfield Insurance, Westfield Center, Ohio, and a co-presenter with Brown.

Beyond ‘gut’

Rigorous, data-based decision- making has supplanted reliance on “gut” decisions in leading organizations throughout the world, said Jeanne Harris, executive research fellow at the Chicago-based Accenture Institute for High Performance, and the Main Event’s keynote speaker.

“For many years, managers were promoted on the basis they could make good decisions without having good data,” she said. “The new generation of leaders is different. They are taught to expect to make data-based decisions.”

Yet, even today, she said, Accenture’s research indicates that 40% of major business decisions are based on “gut instinct” rather than data and facts.

According to Harris, in a rapidly changing world, organizations can no longer sustain competitive advantage by relying on old sources of value, be it a regulated monopoly franchise, proprietary technology, geographic access to customers or suppliers, or occasional breakthrough innovations in products and services.

Success today, she said, requires “analytics-driven optimization of key business processes,” an unrelenting use of data analysis to stake out and refine distinctive market strategies, find the right customers, charge the right price, minimize inventories, maximize availability of supplies, and manage financial performance.

“Companies that invest heavily in advanced analytical capabilities outperform the S&P 500 on average by 64%,” she said. “Companies that invest heavily in developing analytical skills and adopting an analytical mindset recover quicker from economic downturns.”

Commitment

A modeling project is only as good as a company’s level of commitment to it, said Hansen.

Hansen compared two modeling projects at his company, one of which took 40 months to complete, the other 12 months, and identified several characteristics of the latter that were instrumental to its success:

• The more successful project was a corporate priority from before work actually started, and a project manager was designated early on;
• The IT work ran concurrently with the project, rather than starting after the model was built;
• Experienced consultants were utilized, rather than relying exclusively on company staff;
• Company staff were devoted full-time to the project, rather than doing the work “on the side;” and
• The more successful project was the fourth of its kind; the less successful one was the first.

Somewhat surprisingly, Hansen finds that the actuarial work in developing models a relatively small part of a project.

“The actuarial work is dwarfed by the data collection activities and by the work of implementing the models,” he said. “Sixty percent of my job deals with culture change activities.”

Counter-intuitive

Companies must be able to recognize and accept “counter-intuitive” findings if they are to implement predictive analytics successfully, said Brown of State Auto.

“Models are not perfect,” he said. “Models will deliver counter-intuitive results, and new rating approaches produce new challenges.”

For example, he said, analysis of the length of time individuals had been in their job or home, and loss histories (whether at-fault or not-at-fault, paid claims or unpaid) indicated those factors were better predictors of loss than credit history.

“As we identify more and more variables, there’s less reliance on credit,” he said.

That won’t necessarily eliminate the controversy that surrounds the use of credit in insurance pricing, Brown added, “because the data mining [needed to develop new variables] may be more intrusive than the use of credit.

“There’s a privacy concern when we have gathered so much information,” he said, noting that proposals to use education level and occupation as variables could put companies on a “slippery slope” toward confrontation with regulators.

“These are brand new approaches [to underwriting and rating] and they require a lot of explaining,” he said.

External

When it comes to people who need an explanation of how carriers are using analytics, agents come at or near the top of the list, as the use of new or revised variables and algorithms can lead to abrupt shifts in the classification and pricing of accounts.

Many carriers will be pleasantly surprised to learn that independent agents embrace company analytics--as long as agents are treated as partners and consulted on the process, according to Angelyn Treutel, vice president of Treutel Ins. Agency, Bay St. Louis, Miss., and chairperson of the Agents Council for Technology.

“Independent agents have been able to capture more market share because of intelligent systems” implemented by carriers for underwriting and rating, Treutel said in her presentation at the Main Event.

According to Treutel, agencies are aggressively utilizing analytics to improve their own marketing, customer retention, collections, fraud detection, and other performance factors critical to success.

On their own, Treutel said, agencies have leaned that “how a customer pays [for policies] is a leading indicator of whether that customer is a good client for an agency.” For example, people who pay for their policies in full at the beginning of the policy period have fewer and/or lesser losses than those who pay in installments.

Yet, while agents are generally receptive to analytics, carriers complicate relationships when they commit certain common errors implementing and revising rating plans based on analytics.

“The credibility of the agent suffers when the carrier changes appetite,” Treutel said. For example, carriers should avoid having their agents promote a product, then abruptly announce that they capped their writings in the line and will accept no more accounts.

Another concern of agents regarding analytics, Treutel said, is having to “interrogate” clients for personal information carriers are seeking to analyze as potential correlates for loss.

“Ask yourselves what information you need to have, and what information you only want to have,” she said. “You have agents to sell for you. Don’t make them data entry people.”

Regulators

Of course, your rating analytics will never get to customers if they are stopped by regulators.

A fourth speaker at the Main Event told attendees not to expect state regulators to be enthusiastic about new applications of analytics in underwriting and pricing. But, regulators will be more disposed to accept innovations in analytics if they are convinced that pricing precision promotes consumer protection.

“Insurance regulators’ number one charge, by statute, is consumer protection,” said Mark Boozell, a former Aon executive and insurance commissioner in Illinois, now a government affairs advisor in the Chicago office of the law firm Dykema.

“Commissioners won’t always do that (protect consumers), but they’ll always hang their hats on that,” he said. “Prove to them that pricing precision is a consumer protection. That is the big sell.

“The better the decisions, the stronger the companies, and strong companies provide jobs and revenue, not only for the companies but for the state, as well.”

.