As risk managers from all industries look to better assess risk through enterprise risk management programs, the lessons of the insurance industry in incorporating data science and complex analytical models into the ERM process may well prove instructive, experts say.

Because insurance is an intangible product, data is the primary constituent of an insurance company. Accordingly, insurance companies have groups of employees, such as actuaries and underwriters, inherently well-versed at collecting and analyzing data. However, getting data and insight out of the models used in actuarial departments and enmeshing a broader ERM initiative within them remains a challenge, said David Cummings, vice president and chief actuary of ISO Innovative Analytics, a unit of Insurance Services Office Inc. of Jersey City, N.J.-based Verisk Analytics Inc.

“Risk managers and predictive modelers are going after different problems in some ways,” Mr. Cummings said. “However, there is an opportunity to marry enterprise risk management and predictive modeling together to produce new insights.”

Mr. Cummings said this synthesis between analytics and risk management is contingent on investment in complementary technologies that improve the availability and quality of data that flows between departments.

Kimberly Holmes, head of strategic analytics for XL Group P.L.C., said the insurer's intent is to embed analytic decision management process and application systems throughout the enterprise. To aid in this endeavor, the company recently implemented the SAS Visual Analytics platform from Cary, N.C.-based SAS Institute Inc.

%%BREAK%%

Ms. Holmes said the platform's visualization capabilities will make the insights that her team gleans from multivariant progression models and other predictive models more accessible to more people across the organization.

Conveying complex information visually is often more effective than presenting co-workers with a large table of numbers, she said.

“This changes how (analytics team members) communicate with decision-makers and eliminates the perception that getting insights from large data sets is hard,” Ms. Holmes said, adding that increased analytical rigor around decision-making in more parts of the enterprise is likely to yield new insights. “It will inspire questions that would not otherwise have been asked.”

Another important platform capability is speed, Ms. Holmes said. The SAS platform uses an in-memory architecture, which primarily stores data and models in the main system memory rather than on a hard drive, substantially reducing computation time for models.

“The speed is revolutionary for us,” she said. “The visualization of the data in real time allows the user to feel that it is more accessible, more tangible.”

Stuart Rose, global insurance marketing manager at SAS, said there are three major areas where analytics can help improve ERM: data management, risk analysis and reporting.

Mr. Rose said pressure from regulators and ratings agencies were causing many in the insurance industry to readdress how they quantify risk. “We are seeing more insurance companies implementing insurance-specific data models,” he said. “This is because a comprehensive data model serves as a single version of the truth for an enterprise data warehouse and is essential for regulations like Solvency II that have specific requirements for data quality.”

%%BREAK%%

Matthew Josefowicz, managing director and partner of insurance advisory firm Novarica, a division of New York-based Novantas L.L.C., agreed that an emphasis on data quality is critical to fusing analytics and ERM.

“ERM has always been about analytics,” Mr. Josefowicz said. “Improved data quality and analytical capabilities can have a huge impact on ERM.”

In addition to data quality, challenges surrounding the amounts and types of data needed by modelers will likely become more acute in the era of big data.

“Few companies are set up to handle unstructured data or super-high volumes,” Mr. Josefowicz said, noting that traditional data management and query tools are not designed for these tasks. “The biggest question is how will insurers adjust their business models to take advantage of a world of data super-abundance.”

Moreover, Mr. Josefowicz noted that much of the data insurers once spent operations dollars gathering now can be accessed from external providers at much lower expense.

Ms. Holmes said the new tools at her team's disposal will help them deal with the issue of data abundance.

“Historically, we have always accessed external data,” she said. “The difference in what's happening now is that we are starting to automate that access and directly incorporate it into our decision tools as well as store it for further analysis later.”