Role Model

The use of predictive modelling continues to increase in nearly every business line – a positive trend that varies in degree by line of business, notes Towers Watson’s fifth annual Predictive Modeling Benchmarking Survey.

The survey of North American property and casualty insurance executives, conducted from last September 5 through October 21, examined where and how insurers are embracing predictive modelling – a means to enhance performance improvement efforts in underwriting, pricing, claim management and other core functions. (Respondents can provide their input on the sixth annual edition of the survey, which will be launched this September.)

The latest survey of 67 company groups (59 U.S. and eight Canadian) also found that, given marketplace realities, overall usage fluctuates significantly by line of business and company size, and in most cases, data-driven analytics are not applied uniformly throughout the enterprise.

This fluctuation in usage is evident when the personal and commercial lines markets are compared. Personal lines insurers operate in a highly competitive, mature market, so it is not surprising that a very high percentage of these insurers have adopted many aspects of predictive modelling.

Commercial lines, while still competitive, face less intense pricing pressure in some segments, in part due to the heterogeneity of risk and the heightened reliance on individual risk underwriting expertise, particularly in large risk/specialty lines. Specialty lines are considered by some in the industry to hold the potential for more profitable growth, and specialty lines insurers are showing increasing interest in applying predictive modelling to their businesses.

Nonetheless, some smaller insurers have chosen to focus on other competitive differentiators, including customer service and claims, rather than predictive modelling.

Key findings of the survey are as follows:

• large insurers are more active in applying predictive modelling analytics to claim applications, and few small insurers have plans for claim-related applications;

• small insurers seek to differentiate themselves in areas such as service and claims, rather than by modelling;

• large insurers see more favourable top- and bottom-line predictive modelling benefits, while some smaller insurers have concerns about adverse top-line ramifications related to defending market share and retention of existing business;

• personal lines insurers, particularly smaller personal insurers, find value in all forms of competitive analysis; and

• most insurers seek to increase the predictive power of their models, first by exploring new internal and external data, with personal lines insurers more likely to emphasize variable interactions and commercial insurers looking to leverage external, risk-specific variables.

Nearly half of personal lines automobile insurers have formal usage-based insurance (UBI) plans – up from a third last year – and insurers have progressed in executing those plans.

SOLID BASE FOR IMPLEMENTATION

A growing number of p&c insurers have the basic tools and capabilities to enable them to pursue integrating data-driven analytics throughout their organizations. This readiness is suggested by the consistent enthusiasm for predictive modelling, the measurable actions already taken and the proven success of early-adopting insurers.

• Enthusiasm: The enthusiasm for predictive modelling seen in Towers Watson’s two most recent surveys was evident again in the current survey. In Canada and the United States, over 75% of personal lines and small to mid-market commercial lines respondents view predictive modelling as essential or very important. All personal lines respondents attached some degree of importance to sophisticated underwriting and risk selection techniques for rating and pricing, with an overwhelming majority (81%) of personal lines respondents identifying predictive modelling as essential.

• Measurable action: The perceived value of predictive analytics was backed by action. Modelling increased for virtually all lines of business.

• Proven success: Many personal auto insurers that have successfully implemented predictive modelling have expanded to implementing UBI programs (which incorporate a predictive modelling component) as a logical next step in building on their predictive modelling initiatives and integrating them more fully throughout their organizations.

DIVERSE USES AND APPROACHES

Driven by different market segment concerns and needs, survey respondents reported that they are tailoring their predictive modelling programs to focus on specific market realities, and most have yet to achieve a more comprehensive, integrated, company-wide approach penetrating all core functions.

In personal lines, the competitive nature of the market, characterized by the availability of similar products from a number of insurers, is evident from survey responses. In particular, personal lines insurers are concerned that they are competitively aligned with their industry peers because the market makes it imperative to match competitors with comparable products and attractive prices.

In commercial lines, insurers have been slower to adopt predictive modelling, and as a result, respondents said there is less competitive pressure to use predictive modelling applications for risk selection and pricing in their main lines of business. Survey responses indicate Canadian insurers have been somewhat less active than U.S. insurers in commercial lines.

TOP- AND BOTTOM-LINE RESULTS

Canadian insurers have experienced greater top-line impacts than their U.S. counterparts, with U.S. insurers reporting greater profitability improvements. The survey findings offer a general sense of where Canadian and U.S. insurers are benefiting most. Insurers on both sides of the border agree that rate accuracy and loss ratios are improved.

With regard to claim applications, the survey found that applying predictive modelling to claims continues to lag risk selection and pricing, but activity continues to grow in the area, particularly among larger insurers.

In addition, U.S. insurers have been more aggressive than Canadian insurers across all claim-related applications.

The survey findings suggest there are significant opportunities for Canadian insurers, small U.S. insurers (and all other insurers, for that matter) to implement operational changes focused on improvements to the bottom line by applying predictive modelling and data-driven analytics to advance claim performance, given that the claim function is where most of the premium dollars exit the organization.

PRICE INTEGRATION AND OPTIMIZATION

While many insurers are not currently using price integration (i.e., bringing together customer behaviour, competitor and loss cost models to derive key business metrics, such as profit and volume, to test the impact of different rate scenarios) or price optimization (i.e., the application of a mathematical search algorithm to a price integration framework, aiming to identify the rates that maximize business metrics), they increasingly plan to do so.

Survey findings suggest that some lines may not be ripe for these techniques.

This is either because the data are lacking (price information is hard to collect or risk level modelling is not a viable option) or because companies have not prioritized price integration and, ultimately, price optimization to enhance performance while reflecting specific business goals.

Survey findings further suggest that some polled companies do not fully understand the benefits associated with price integration and price optimization, including how they are, ultimately, a means to measure and enhance customer value, and how those findings can be factored into everythi
ng from pricing to targeted new business marketing and portfolio defence.

THE ROLE OF DATA

As the sophistication and power of applying predictive models and data-driven analytics have increased, insurers continue to jockey for position, looking for the next predictor variable or interaction that will give them an edge over competitors. A majority of insurers continue to explore new internal and external data in an effort to improve the predictive power of their models. In addition, commercial lines insurers stress external risk-specific or sociodemographic data, while personal lines insurers tend to look for more variable interactions to strengthen their models.

Personal lines insurers are much more likely to apply predictive modelling in the form of rating plan adjustments by creating or revising rating/tier variables and relativities, while standard commercial lines insurers balance pricing and risk selection responses.

In particular, while a majority of commercial lines insurers also use predictive modelling to create or revise rating/tier variables and relativities, a much larger number also use predictive modelling to revise underwriting rules, including acceptance/rejection criteria and company placement.

These efforts need to continue – and insurers need to take it a step further and develop a more comprehensive, strategized and aggressive approach to predictive modelling implementation.