Linking for fewer losses: insurers need integrated data to prepare for the changing nature of catastrophic loss.

The numbers are already all too familiar: In 2005, Hurricane Katrina rang up between $38 billion and $55 billion in insured property losses spread mostly across homeowners, commercial property, business interruption, and marine and energy lines. Then there was Rita ... and then Wilma.

According to the Insurance Information Institute, "The number of natural and man-made catastrophes has been increasing on a global scale for 20 years.... Eight of the 10 most expensive disasters in U.S. history occurred within the past four years." The list of eight includes the Sept. 11, 2001, terrorist attack, which made all too clear that insurers had an extremely complex new element to weave into their planning for catastrophic losses.

But it's hardly the only new element. In the aftermath of the 2005 storms, some states have been considering legislation that would essentially wipe out the exclusions written into policies. Add to this an inadequate government response, which few, if any, foresaw when companies set underwriting guide lines and sold policies.

The only good news is that because 2005 was characterized by strong insurance industry profits, loss ratios, and capitalization, the industry was in a position to absorb the losses. Had disasters occurred with the same force and frequency three years earlier, when the industry's policyholder surplus was dangerously low, insurers' financial picture might have paralleled the physical destruction Katrina wrought.

Seen in that light, the 2005 storm season was an early warning signal for insurance companies: Since the nature and force of disasters have changed, disaster planning must change with them. A key element in that change must be the way insurers collect, store, analyze, leverage and distribute data throughout their organization.

More Than Lines of Business

In the past few years, property/casualty insurers can pat themselves on the back for having taken firm hold of their data, using it to properly calculate risk, and applying those calculations to underwriting in individual lines of business. It's one of the keys to vastly improved loss ratios.

Unfortunately, most recent catastrophes, be they natural or man-made, didn't affect just one line of business at a time. Consider Sept. 11. It didn't just affect those who wrote commercial property policies in World Trade Center buildings. Workers' compensation, business interruption and life insurance were among those hit throughout the entire neighborhood and much of New York City. For Katrina, if you were looking only at homeowners policies and not considering other exposures, you might have written too many policies in the region for other lines of business.

The clear implication is that in underwriting--and in making marketing decisions about where to sell and write policies--a separate snapshot of each line of business is not enough. What companies need is an incisive family portrait by a sensitive photographer, one who can reveal in just a few group photos the complex underlying relationships of all family members. This will enable new models and underwriting formulas to emerge that reflect an understanding of exposure across the entire organization.

That portrait can't be achieved unless insurance companies rethink their data infrastructure. Without the ability to assemble cross-functional data and position it in a variety of ways that reveal key relationships across lines of business, insurance companies will not have the insights they need for proper catastrophe planning.

Complex Modeling

In the aftermath of Sept. 11, at least one company was able to use sophisticated models that draw on integrated data from across the organization to quickly provide a picture of its likely losses to its reinsurer. It was enormously valuable in preparing both the primary insurer and reinsurer for the financial aftermath of the attack. It also illustrates how critical it is that actuaries change the way they do their modeling to calculate proper reserves and inform underwriting policies.

These new models have to contemplate a much broader range of assumptions than they've ever had to before. Various terrorist scenarios (dirty bomb, chemical or biological attack, suicide attacks, attacks on water supply, power generation) have to combine with the natural disaster models for earthquakes, hurricanes and tornadoes.

Moreover, those traditional natural disaster models need to incorporate new assumptions about governmental response and how post-disaster legislation might try to recover more than the industry had anticipated when they originally wrote in policy exclusions. The models must consider broad geographies in sophisticated ways and be able to make inferences about complicated chains of events that go beyond the initial disaster. (Consider how in New Orleans, the flooding was exacerbated by human response and the struggles that government had around evacuating and housing the displaced citizens.)

To build those models--to populate, test, and refine them in a way that ensures their validity--integrated data is essential.

Why? For one, drawing from isolated data marts often means trying to reconcile the different ways different business units or business functions define events or use vocabulary specific to their individualized business needs. In addition, the very act of extracting information from various locations for cross-functional analysis is not only time-consuming and expensive, but it also increases the chances for errors that the industry simply can't afford. As the World Trade Center example illustrates, that integrated data must be constantly updated in one place so insurance companies can constantly adapt their reserves and tactical decisions as circumstances dictate.

Integrated Data Platform

Conceptually an integrated data platform makes sense, but there are two practical characteristics that must be present to make the move to integrated data a success.

First, the platform needs the capacity to absorb near real-time data feeds so that all employees across a company always are working from the most up-to-date information when they make decisions that have implications for disaster preparedness. For example, when managers decide on a marketing effort based in a particular geographic area, they need an immediate picture of the company's cumulative exposure, regardless of line of business--and they should be able to drill down to understand the details of that exposure.

Second, the data must be organized in such a way that it supplies robust detail data--not summary data--for insurance-specific analytical applications, and provides rapid and secure access for each type of business use and user.

These two qualities are not just important for disaster planning; they also expand the power and value of integrating data. Those responsible for regulatory compliance and risk management are better able to get the picture they need and provide essential reports to regulators on demand. Senior executives who want to rethink strategic priorities around business retention or cross-selling can make decisions rooted in a single, up-to-date view of their entire business.

Avoiding Pitfalls

The question then becomes: how to get from here to there? Given the dramatic change an integrated data platform implies for business processes--most notably the understandably protective culture that has grown up around individual lines of business and business functions' data--companies can anticipate challenges. Two common mistakes can undermine the effort to overcome those challenges.

The first is trying to do too much at once. Far too many projects have failed for being overambitious, and insurers should keep their eyes on the prize. If catastrophic planning is to be the first beneficiary of this process, companies should stay focused on that. If they decide another strategic initiative should be the first beneficiary, then they should begin there and move to catastrophic planning later.

Another common mistake is that information technology organizations focus on chasing down the data, rather than first developing a detailed understanding of how business users--say the actuaries who will be doing catastrophic modeling--will put the data to work. This is a near guarantee of an underperforming integrated data platform that will frustrate users. In turn, they'll wind up more dependent than over on disparate, "homegrown" solutions, leaving the IT side unwilling to make such an effort again.

To avoid these mistakes, the implementation should begin with a commitment to a strong, ongoing dialogue among business users and the IT organization. Companies should also approach this as a step-by-step effort that builds on smaller, initial successes to foster company buy-in and finance the next step.

Three Steps to Success

Once the organization understands the potential pitfalls, emerging best practices indicate that three basic steps lead to a successful implementation.

In the first step, senior management engages with the IT organization to develop hypotheses regarding the greatest value drivers within an organization where integrated data can play a role. Again, given recent events, catastrophic planning seems a logical starting point, but not every organization will agree. Certainly, driving greater customer retention or regulatory compliance are other compelling reasons to bring data together.

The second step is to clarify and expand the hypotheses through sessions with the actual business users, which for catastrophic planning might include actuaries, underwriters and data analysts. The meetings should focus first on understanding the business users' unique goals and objectives. For example, actuaries might say that in order to build their models and answer key business questions, they need loss ratios from all lines of business, the data marketing people use to segment their markets, and data from external sources, such as figures from past catastrophes.

From there, people can work together to ferret out the current capabilities that can be used to provide that information. What data does the organization already have? What data does it need? How quickly must the information be available to be effective? Are the applications in place to make it happen? This process allows both groups to understand the existing information and application gaps and decide if they have efficient ways to close those gaps. It also helps both groups understand each other's primary concerns and capabilities and focuses dialog on the ultimate use and value of newly gained information.

In step three, IT must organize the data within an insurance-specific logical data model, and then link the business questions to the integrated data and ultimate sources from which the data originates. This will ensure a focus that quickly facilitates the transformation of data into valuable catastrophic planning or other strategic business information.

From there, the system can link the business questions (what is the effect of a dirty bomb and evacuation?) to an array of specific improvement opportunities (understand the proper level of reserves, how to mount a customer service response, spread geographic risk, adjust underwriting to account for new understanding of disaster impact), which are linked to the original overarching goals and objectives that senior management laid out (be better prepared for the new catastrophic environment without creating excess reserves that limit the company's business flexibility).

Time to Move

Insurance companies are understandably wed to their current ways of using data, especially to the common practice of putting it in silos that represent individual lines of business and business functions. Many individual lines have done quite well by looking only at their individual portraits.

But insurers' ability to absorb the hits of last year's storms may have lulled many into a false sense of security. An expanding array of disasters with much larger impact across geographies and lines of business has changed the environment in which all of us--insurers especially--operate. And the only way to be prepared for that is to have reserves that reflect complex new models that rely on integrated data.

Implementing Integrated Data Platforms

Common Mistakes

* Trying to do too much at once. Insurers must stay focused on the primary strategic initiative, whether it's catastrophic planning or something else.

* Chasing down data before understanding how business users will put the data to work.

Doing It Right

* Develop hypotheses regarding the most important functions within an organization where integrated data can play a role.

* Clarify and expand the hypotheses through sessions with the actual business users.

* Organize the data within an insurance-specific logical data model, Then link the business questions to the integrated data and ultimate sources from which the data originates.

Key Points

* For proper catastrophe planning, insurers need to assemble crossfunctional data and position it in a variety of ways that reveal key relationships across lines of business.

* New disaster models have to contemplate a much broader range of assumptions than they've ever had to before.

* Integrated data platforms must absorb near real-time data feeds and supply detail data for insurance-specific analytical applications.

Contributor Tracy A. Spadola is senior industry consultant, insurance industry, for Teradata, a division of NCR.