Imagine you are heading up into the mountains for a winter’s day of skiing. As you get on the lift, your cell phone chimes with a text message. It’s from your insurance company. For $20, you can text them back quickly and have a one-day policy rider added specifically to cover your skiing adventure… broken gear, lost wallet, sprained ankle.
It’s an OTI, or One-Time Insurance policy, and it may be the future of the insurance industry.
Most people are familiar with OTI’s like travel insurance or vehicle rental coverage. The on-the-spot ski trip insurance seems more exotic, but Boston Consulting Group, an insurance industry management consulting firm, laid out just such a scenario way back in 2013.
Your insurer could use cellular location data to figure out when you hit the slopes. Through their extensive databases, they could pre-qualified you for the OTI policy and send you the text. Further adjustments to the rates and coverage could happen in near-real time as the phone’s GPS data revealed your skill level and the danger of the runs you choose.
The Complicated Concept of Risk
Insurance is all about risk, but risk is a complicated concept. The insurance industry has always run on statistics to make sense of those risks. Actuaries (possibly the original data scientists), who calculate the likelihood of certain catastrophic events that might require policy payouts, have been a mainstay at insurance companies since the 17th century.
Today, actuaries are as important as ever, but the data from which they make their calculations has vastly expanded. Data scientists at insurance companies are now pulling in information like:
- Government-tracked health, climate, and epidemiological data
- Vehicle instrumentation and tracking data
- Satellite terrain and geographic information
- Credit reports and economic data
But Big Data carries a risk of its own for big insurers. As pointed out in a 2013 article in Information Week, insurers previously gained a competitive advantage by accumulating more and more varied sources of information. Today, however, much of that information is accessible to all insurers, and the competitive advantage will go to whichever of them figures out how to analyze it fastest and arrive at the right conclusions.
That job is going to belong to master’s-educated data scientists.
Calculating Risk More Precisely with Big Data from Unexpected Sources
In many cases, calculating risk revolves around individual behavioral patterns, patterns that an insurance agent or actuary couldn’t easily access or assess for each potential client. At one time, all actuaries had to work with was internal company data, compiled from corporate records of losses and payouts. Such figures could determine a broad scope of risk based around generalities such as age and sex.
But data scientists comparing accident and billing data noticed a curious thing: people who paid their bills on time tended to be safer drivers. Soon, many insurance companies began to incorporate credit check data into their actuarial calculations for individuals, even if that person has never been insured before.
Now, consumers tend to document their lives digitally in increasingly minute slices. Social media postings, GPS (Global Positioning System) location and travel, purchasing information. All of these things may be mined to establish risk factors for more and more precise demographic groups, right down to the individual. Where insurers previously rated a customer’s risk category based on perhaps 10 or 20 demographic factors, now they use thousands- everything from credit scores to credit card purchases.
For health insurers, devices such as the Fitbit and other personalized monitors offer a wealth of individual information. In 2015, John Hancock insurance became the first in the United States to offer discounts to customers willing to allow it to monitor their Fitbit data streams. Although the information can provide predictive diagnostics of potential health problems, the real goal of the program is preventative: insurers find that consumers who are more aware of their own health and fitness have fewer health problems and require lower insurance payouts.
At larger scales, insurers are learning how to integrate government health data with their own actuarial data. Regional trends tracked by the National Institutes of Health (NIH) allow insurers to adjust, for example, for the relative likelihood of contracting infectious diseases in different parts of the country.
It’s not just individual risk that can be assessed more accurately with data science. Insurers who cover large corporations have unprecedented access to economic and regulatory data to check up on the health of their big corporate clients and assess their prospects for filing expensive claims during the span of a policy. Government data on industrial accidents and workplace complaints of various stripes are factored into policy rates for businesses.
In every case, insurance companies that calculate risk more accurately can price their policies more accurately- often reducing claims payouts at the same time. Data science is the key to getting ahead in the insurance industry today.
Data Drives Auto Insurance Premiums Down
According to a 2015 article in the Washington Times, auto insurers are increasingly offering discounts to customers who are willing to offer up GPS data about their driving range, speed, and habits—information known in the industry as “telematics.” Progressive Insurance bases rates on braking data, for instance, assessing risk from how often you have to stomp on the brake pedal at the last second. And Allstate insurance looks at where you drive, comparing that data to road safety and accident locations to determine how high your risk is on those particular roads.
Interestingly, this monitoring has helped decrease risk in and of itself. A report from industry analyst McKinsey Consulting in 2013 found that one large U.K. insurer reported that the very fact that drivers were aware they were being monitored resulted in better driving habits, leading to a 30 percent reduction in claims from that group.
And telematics data also allows innovative new pricing models. Many auto-insurance companies now offer pay-as-you-drive policies, which charge consumers by the mile, based on telematics data. Low-mileage drivers can save considerably by taking advantage of such programs; insurers, on the other hand, can ensure they are pricing their policies most efficiently on the basis of actual accident risk.
Fine-tuning Policies for Customer Retention and Fraud Prevention
Individual behavioral patterns can not only predict risk factors that may result in policy payouts, but can help predict more nefarious acts as well. Insurers now use data to parse claims for possible signs of fraud.
A 2013 article in Claims Journal, an industry publication, details the process by which big insurers are beginning to integrate multiple data sources to predict potentially fraudulent claims for further investigation. Claims may be cross-referenced against other claims from related parties and open-source data from services like Twitter. Suspicious patterns of posting and claims will quickly result in a visit from a suspicious claim’s adjuster.
But insurers are using data science to find good customers to keep as well as bad customers to cancel. The same types of systems that can detect likely fraudsters can be used to search for behavioral clues when a customer may be considering canceling their own policy. An agent can reach out to such customers before they leave, offering incentives or rate reductions to ensure loyalty.