Big Data for Commercial Lines

Does such potential exist outside the world of personal lines?  What does big data mean for the corporate risk manager, the global broker, and the carriers that serve them?

To answer these questions, we have to look at how the dynamics of the commercial lines market challenge the analytic techniques typically employed by big data applications. Carriers have done far less here than those in the personal lines space.  Sample size is a significant issue, as even the largest carriers serving the commercial lines marketplace write a relatively small number of policies. Claim volumes also are typically low, which limit the effectiveness of the statistical models used to predict either the frequency or severity of loss and to price risk accordingly. 

Variability within the data also presents challenges.  Large commercial entities defy easy categorization; each has unique aspects to its organization, business activities and operations that make it, in many respects, a "sample of one."  Similar variability applies to the insurance products that large companies buy. Complex program structures, tailored coverages, and manuscripted endorsements make it difficult to structure policy data sets to enable true "apples-to-apples" comparisons. These issues have limited the take-up of big data initiatives among carriers focused on large commercial accounts.

An exception to this general rule has been in the area of public company directors & officers (D&O) insurance.  The widespread availability of class action lawsuit information documenting incurred losses and its correlation with publically available financial data has encouraged several leading carriers to develop sophisticated predictive models. These models score and price individual risks based on estimates of the likely frequency and probable severity of a loss based on hundreds of data elements relating to the historical financial performance of a company and its stock.  

Global Brokers Are Taking Advantage of Big Data 

The leading global brokers have taken greater advantage of the big data opportunity.  As intermediaries, Marsh, Aon and Willis have access to more policy data than any single carrier.  In recent years, they have begun to warehouse their clients' submission and program data in large consolidated databases. These allow new analytic offerings built on the foundation of these big data assets. Enhanced price and limited benchmarking, more finely tuned retention analysis, and competitive insights into the carrier marketplace promise better program structures and optimized placements. Risk managers can and should benefit from this scenario-based modeling of the total cost of risk (TCOR) provided to them by their brokers.  

Claims data is on the frontier of big data initiatives. Brokers have begun to aggregate claims information across their clients.  This work will take time, but the potential of such datasets is enormous. Statistically meaningful samples of high-severity claims offer the possibility of building predictive models for lower-frequency casualty lines. These models promise better estimates of both the probability and probable maximum loss (PML) for a variety of large-loss scenarios not captured through traditional catastrophe modeling techniques. With such models, risk managers can take more strategic approaches to risk transfer and enterprise risk management (ERM).

The Potential for Risk Managers

Risk managers do not need to depend on brokers to enjoy the benefits of big data.   Here are three areas where risk managers can take the initiative and deliver value to their enterprises today:

• Telematics: The inexpensive devices which power usage-based insurance in the personal lines marketplace are also widely used in commercial fleet-management applications. The data produced by these devices, properly mined and analyzed, offers great potential to drive risk management initiatives directed at improving the safety of driving practices throughout the enterprise. Telematics provides the foundation for proactive motor vehicle risk management and can contribute to substantial reductions in TCOR.
• Workers' compensation: Many large organizations have a significant number of attritional workers' comp claims that fall within their retention.  Data about these claims, whether managed through an internal risk management information system (RMIS), offshore captive or TPA, is an asset. Predictive models based on this data can identify factors driving lost activity and suggest risk mitigation approaches to lower claim frequency and reduce cost.
• Supply chain risk: The recent Fukushima earthquake highlighted the vulnerability of many enterprises to natural disasters affecting their suppliers. Companies can leverage the data stored in supply chain management systems to quantify and mitigate such risks.  This data can be mined to generate geospatial maps of principal suppliers. Property catastrophe modeling software can leverage these maps to calculate the PML for a series of potential supply chain disruptions. Risk managers can respond proactively, chartering supply chain diversification initiatives or purchasing contingent business interruption coverage.

The complexity of commercial lines business makes the adoption of big data approaches more of a marathon than a sprint. The deliberate pace of adoption should not be confused with a lack of momentum. Although it may not be a revolution, big data offers large commercial enterprises significant opportunities to improve both risk management practices and the effectiveness and efficiency of their insurance purchases.