Visualize is pleased to present a recurring discussion series of timely topics written by Neil Spector, President of Underwriting Solutions for Verisk Insurance Solutions.
Underwriting, pricing, and managing insurance risk have long been done at arm’s length. Proxies for human behavior are at the heart of most underwriting and pricing. Understanding causation and best practices is the essence of risk management and loss prevention.
The 'missing piece'
The missing piece has been knowledge of what real people are doing in real time to cause or avoid losses. Today’s science of predictive analytics is taking steps to close that gap with ground truth about risk.
Moving beyond proxies
Underwriting and pricing through proxies assume that people with certain characteristics, on average, are more or less prone to risky behaviors. The crudest proxies are the most distant from behavior: gender, age, or other demographic categories. Modern tools such as credit are more relevant; a credit score reflects actual behavior and often a level of personal responsibility. But it’s still a proxy.
Predictive analytics looks differently at data, including behavioral data, to measure and price risk. For example, telematics-based driving data allows scoring of a customer’s actual behavior behind the wheel to price usage-based insurance (UBI).
Just as proxies equate certain traits with likely behavior, risk management and loss prevention tend to equate knowledge with action. With the right information, it’s assumed that people will do the right things to stay safe. Reality teaches otherwise.
But predictive analytics could drive convergence between pricing risk and reducing it. In this way, insurance can move beyond the transactional model of rating perils and paying claims to preventing more of those claims in the first place. Still, there’s more to it than throwing data at the problem.
Building the interface
Where human behavior meets cold data, New York Times columnist Thomas Friedman coined a new word—STEMpathy—to describe the intersection of people skills with science, technology, engineering, and mathematics. Applying this kind of capability may lead insurers to ways of reducing risk by influencing behavior.
One insurer ran a small-scale experiment in which a buzzer was installed in the subjects’ cars to alert them to unfavorable events such as hard braking or sudden sharp turns—items that might factor into driving scores for an actual UBI program.
The test was structured as a contest in which the highest-scoring participant won a $50 gift card. Driving behaviors improved significantly. But two months after the contest ended, the buzzers were turned off, and many of the drivers quickly reverted to their old ways.
What incentives change risky behaviors?
The lesson: Ongoing reinforcement is needed. But how much, how often, and most important, what kind of feedback and incentives change risky behavior? As big data and behavioral science become more intertwined, predictive analytics as applied to risk may evolve to something better described as experimental design: a quest for triggers to change what people do.
One key may be specificity. Credit reports can point to aspects of consumers’ behavior that affect their credit scores. Similarly, driving feedback might lend itself to customized tips targeted to the individual driver’s observed behavior. This could include the offer of a reward for improvement—perhaps lower premiums or more flexible policy terms to better meet the customer’s needs.
Wrangling the numbers
Before that can happen, the data needs to be tamed. Insurance data has always been big, but only in the past 20 years has the hardware become able to support predictive analytics that turns big data into smart data.
When bringing predictive analytics to the mainstream, it’s important to account for the human element—not only with customers but also within an organization. IT professionals can build tools to leverage new data streams, but without the users’ buy-in, veterans who trust their experience over technology may leave those tools on the shelf.
With a committed team adding the human touch, predictive analytics can help reshape risk from the inside out—how it’s measured, priced, and mitigated. The ground truth can be understood and even changed for the better, and the old proxies can be put out to pasture.