Skip to main content
Survey Report

Predictive modeling: new applications, new questions

2016 Predictive Modeling Benchmark Survey (U.S.)


March 24, 2017

Predictive modeling use by P&C insurers is steadily advancing beyond risk selection. But questions remain over how best to exploit potential new IoT data sources.
This Willis Towers Watson survey examined how U.S. P&C insurers are using and plan to use predictive models and big data to gain competitive advantage.


Although yet to attain juggernaut-like momentum, U.S. P&C insurers are progressively expanding their use of predictive models, both in core risk selection and underwriting, and in other business operations. And while it’s no surprise that personal lines carriers continue in most cases to take the lead, more commercial lines insurers indicate they are intent on building their capabilities, according to the latest Willis Towers Watson U.S. Predictive Modeling Benchmark Survey.

As more carriers embrace predictive modeling, the variables and predictors used are also expanding and methodologies are evolving. But building value from expanding modeling applications brings new challenges, many of them linked to accessing and unlocking the potential of data from sources such as telematics and the Internet of Things (IoT) that could really set the juggernaut rolling.

Top uses of predictive analytics

Two-thirds of P&C insurers surveyed currently use predictive models for underwriting and risk selection, an increase of over 10 percentage points compared to the 2015 survey. The reasons behind such an increase are clear.

There is unanimous agreement from personal lines insurers about the fundamental importance of using more sophisticated predictive techniques to drive success in today’s market. Equally, many commercial lines carriers are recognizing that the traditional barrier of the relative paucity of homogenous risk data in commercial portfolios can be overcome, enabling models to contribute significantly in more unique underwriting environments. Eighty-six percent of small- to mid-market carriers rate more sophisticated risk selection as essential or very important to future success. Over half (56%) of large account or specialty lines carriers share that view.

The degree to which many commercial carriers are becoming converts to predictive models is all the more evident in how they expect to use them within the next two years (Figure 1). From a lower current base use compared to personal lines companies, commercial insurers have ambitious expansion plans in areas such as claim triage, fraud potential, litigation potential and case reserving. Survey results suggest these could become increasingly important areas for performance differentiation, building on what many carriers believe models have already helped achieve.

Figure 1. Current and projected top predictive modeling uses

Current and projected top predictive modeling uses

Already, the perceived positive influence of predictive modeling on various business metrics is strong. For example, over 90% of P&C insurers surveyed say models have had a positive impact on rate accuracy, loss ratios and profitability. Benefits have also fed through, albeit on a smaller scale, to top-line performance, notably on renewals, but also in the expansion of many companies’ underwriting appetites and improved market share.

Technical modeling issues

Modeling experience in North America continues to be heavily weighted toward either generalized linear models or simple one-way analyses, although with a ging recognition that the type of methodology used depends on the modeling application and goals. Within existing models, the value of various internal and external data predictors varies by type of carrier. For example, both standard and specialty lines commercial insurers place particular value on analyzing data such as credit and financial attributes, and account experience. Many carriers also now commonly model separately by coverage and peril, typically overcoming any thin or incomplete data by supplementing models with industry data or competitive market analyses. Alongside these evolving methods for determining technical price, insurers consider competitive positioning, tempering of rate impacts on policyholders and agent feedback as primary factors in making final pricing decisions.

However, insurers are also becoming more aware and proactive about the potential contribution of machine learning techniques, with over half of insurers saying they either do already or intend to harness what they have to offer. The most common operational targets for machine learning methods are loss cost modeling and claim analytics (Figure 2). What the survey can’t tell us is how prepared companies are for the demands this will place on their actuaries and analysts. Since machine learning techniques work by considering and filtering every conceivable variation of scenarios, the results they produce can be very difficult to understand and interpret without appropriate experience and sufficiently robust software.

Figure 2. For which business applications do P&C carriers use or plan to use these methodologies?

For which business applications do P&C carriers use or plan to use these methodologies?

Data drive predictive analytics power

Data are the essential fuel to move predictive models forward, so the degree to which carriers characterize themselves as data-driven has a significant bearing on how aggressive they believe they are in employing analytics. The majority of midsize and large companies do consider themselves data driven, whereas a majority of smaller companies don’t. This divide doesn’t necessarily reflect differing states of mind or levels of conviction, but it is commonly seen as linked to data access, warehousing constraints and problems with IT bottlenecks and coordination.

In our view, unlocking the value of data — both emerging data sources and information that can frequently be tied up in a web of company legacy systems — represents the most likely near-term source of enhanced pricing accuracy, product differentiation and business outperformance opportunities for both personal and commercial lines carriers. Big data, notably from vehicle telematics and the IoT, are opening up many new potential avenues for investigation and improvement. These opportunities apply as much to carriers that have invested recently in improved policy administration and quote systems as it does to others. Whatever the available level of hardware and software within a business, a lack of accompanying investment in data and analytics is rather like driving a sports car without fully revving up the engine.

Many companies have already made a start. Types of big data currently seen as most useful include internal claim, underwriting and customer information. These are also expected areas of focus over the next two years, along with expanded use of usage-based insurance data, customer interaction analysis, social media analysis and smart home/ building data (Figure 3).

Figure 3. From which sources do P&C carriers currently collect big data and plan to in the next two years?

From which sources do P&C carriers currently collect big data and plan to in the next two years?

What about obstacles to progress? Survey respondents say that far and away the biggest challenge to generating additional business value from data is the availability of people with the right training and skills. Cost is also an inevitable concern, as are issues linked to identifying and accessing data of the right quality and reliability. Even so, expectations about how broader data usage will benefit carriers in the next two years are strong. The biggest expansions of capabilities are anticipated in loss control and claim management, and also in understanding customer needs and behaviors.

Predictive modeling strategy: take control

For now, the P&C insurance predictive modeling momentum hasn’t ramped up to breakneck speed. But we are nearing the point where market momentum will accelerate as value-building big data, and diverse and ging analytics techniques take hold. Companies that want greater control over their destinies will marry flexible IT frameworks and partners with the analytics techniques and skills necessary to effectively harvest data.

About the survey

Willis Towers Watson’s 2016 Predictive Modeling Benchmark Survey asked U.S. P&C insurance executives how they use or plan to use predictive analytics and big data. The survey was fielded from September 7 to October 24, 2016. Respondents comprise 14% of U.S. personal lines carriers and 20% of commercial lines carriers.

Download PDF
Title File Type File Size
Predictive modeling: new applications, new questions PDF 2.7 MB
Related content tags, list of links Survey Report United States
Contact Us