With the recent buzz generated around data, the migration from telematics in auto insurance to advanced analytics in homeowner risk is not surprising; perhaps it is even inevitable. Several factors have come together to push the quest for information on homeowners insurance to a different level.
“The successes in other lines of business and applications have created capabilities and awareness of the power of predictive analytics,” says Greg McCutcheon, president of Opta Information Intelligence. “Many started with personal lines auto, claims triage and other related supply transaction costs… Personal lines property is a natural evolution,” McCutcheon suggests.
Given that homeowners insurance accounts for roughly 20% of the overall property and casualty insurance premium volume pie, it stands as a significant line of business. With losses mounting from severe weather, water damage and other claims, there is heightened pressure on the personal property product in terms of rating, underwriting and loss reduction.
“If you look at the results of Canadian insurers over the past couple of years, you can really see the impact of weather-related claims,” says Mary Trussell, partner with KPMG Canada and a member of the firm’s global insurance leadership team. “The shape of risk is changing. Insured values are rising. In what might initially seem like a row of uniform houses built in the 1970s, many have likely been redeveloped and renovated,” Trussell observes.
“One piece driving this is that personal lines property has struggled in terms of profitability, particularly in Western Canada,” says Simon Mellor, assistant vice president of pricing and reinsurance for SGI Canada. “Insurance companies are placing an increased focus on these products,” Mellor suggests.
It is not just troubling loss ratios, but also the availability of data that has helped spur change in homeowners insurance. Moving beyond internal transactional data to outside sources of information has proved highly valuable in drilling down into individual risks and offering even more thinly sliced segmentation.
“There is more and more external data available from all kinds of different sources,” says Shelley Toyota, vice president of personal insurance for RSA Canada. “Whether you are talking about flood models, postal code data – there is more access to different kinds of data than ever before. And then the dilemma becomes, when you are dealing with masses of data, do you have the technology to manage it?” Toyota asks.
Chris Van Kooten, senior vice president and chief underwriting officer for Economical Insurance, comments that in addition to data, p&c insurers today have access to more sophisticated analytic tools and greater computing power. “We are finding ways to do more things with larger data sets,” Van Kooten reports.
“New solutions coming out from a technology perspective put the old way of managing data behind us. Previously, you had to take all your data and structure it into a format that you can easily pull from; the new solutions allow you to just put blobs of data onto systems, not structure it,” he reports.
“Over the last few years, it has become somewhat of an arms race of being able to understand your data better than your competitors,” says Van Kooten.
“I think we are at a tipping point,” suggests Keith Walter, senior advisor with Deloitte in Canada, who specializes in analytics and actuarial work.
“The vast majority of significant p&c insurance players have either implemented or are in development with data analytics for homeowners insurance within the next 12 months. By the end of 2015, it will be a requirement for players in this space,” Walter contends. “Those that get left behind can be severely disadvantaged,” he cautions.
In a survey of 99 insurance homeowner representatives in the United States and Canada, results of which were released this past November, Verisk Analytics and Earnix found 57% of respondents now use predictive modelling for homeowner loss cost development. As well, 8% of those polled use a by-peril rating structure, which focuses on gathering data on individual perils to determine an accurate price for each level of risk.
One of the prime reasons for increased usage of data analytics is to curb worrisome trends in the loss ratio, which jumped from 58% in 2012 to 74% in 2013 for personal property lines in Canada.
A study of U.S. insurers in 2012 found that those with by-peril rating plans had loss ratios 7.4% lower than companies using traditional rating systems, reports Douglas Wing, assistant vice president of analytic products at ISO, a source of information about property/casualty insurance risk now part of Verisk Analytics.
“The prime benefits (of data analytics) are improved loss ratios with lower loss costs, better rate for risk or pricing, optimized loss control and better understanding of the risks your portfolio faces,” says McCutcheon, whose company works closely with Canadian insurers on predictive modelling, data analytics, property data and peril scoring.
Advanced data analytics for homeowners insurance can entail a number of head-scratching terms, such as univariate analysis, sampling, regression/general linear modelling, splines and spatial smoothing. However, at its heart, it involves finding useful data, or patterns of data, that inform business decisions in rating, underwriting and claims.
By-peril rating is one of the first tangible outcomes of that process.
“The whole point of analytics is to try to understand the information that your company has at a deeper level, whether that means doing it by peril or through different ways of grouping the data,” Mellor points out.
“What you are trying to do is extract new and actionable information with the data, and breaking it out by peril is one way of doing that,” he says.
SGI Canada moved to an “individualized rating environment” for personal lines insurance in January 2014.
Toyota explains that the peril of fire has been eclipsed by other risks, including inside water damage, outside water damage, ground-up water damage, hail, hurricanes and earthquake.
“All of these perils are emerging. And it is forcing us to ask: What other data is available? How do we couple external data with our existing data? Do we have the technology to actually pull out the insights?” Toyota says.
“By-peril rating is something that makes a lot of sense from a pure pricing perspective, but it also helps in giving consumers some options,” notes Van Kooten. “For example, if they live in an area where they get sewer back-up every year, maybe they can opt out of sewer back-up coverage because it is so expensive,” he says.
“By giving them more information about it, they can start to manage some of that risk themselves and take action to reduce their exposures,” he adds.
“Insurers are able to improve pricing because they have a better understanding of by-peril exposure, both likelihood and severity of specific losses,” McCutcheon observes. “Knowing that risk can never be completely eliminated, applying loss control or targeted underwriting action to specifically identified properties improves t
he return on investment for such initiatives,” he notes.
Greater accuracy in pricing is not the only potential benefit of data analytics in personal property insurance. Walter suggests that operational efficiency in areas such as claims management may emerge as a significant breakthrough for insurance companies.
“Insurers are trying to find ways to intervene in and improve the claims-handling process for better overall outcomes,” he says.
“We see things like the risk of exaggerated claims and the opportunity to do a better job working with the client to reduce claims costs – these areas are very ripe for improved data analytics,” Walter adds.
Mellor observes that data analytics can also lead to more innovative products in homeowners insurance. “Looking at data in new ways, I think, ultimately contributes to the evolution of the product and new offerings,” he says. “This creates more choice for customers, which could include loss prevention,” he suggests.
“Once companies start to manage their own data effectively, you start to look for external data sources that might supplement the information you have,” notes Van Kooten. “There are a lot of things you can do with that. It can go into your pricing, it can go into product design, customer experience and providing additional tools so that customers and brokers can understand what is happening with their risk.”
Innovation may find a more likely breeding ground in personal property lines, as opposed to the more rigid side of auto insurance.
“The interesting thing about home-owners insurance is the opportunity to introduce new approaches,” Walter says. “Auto insurance tends to have more regulatory control. You have more flexibility in personal property. And insurance companies are rolling out new methods and new technology.”
While insurance companies talk about enhanced transparency of pricing through data analytics and by-peril rating, this could be a difficult sell to consumers who see a spike in premiums as a result of repeated claims activity.
Clearly, there will be winners and losers in the personal property data game. One of the main challenges will be how to communicate those changes to customers.
“We have to translate the science into a language that customers and brokers understand,” says RSA Canada’s Shelley Toyota. “Sometimes, we can get caught up in our own sophistication. At the end of the day, we need to spend more time asking, ‘What does this mean to brokers in terms of their ability to better serve the customer?'” she points out.
“Instead of saying, ‘Here is sophisticated black-box underwriting,’ we need to boil it down to: ‘This is the outcome, here is how you can explain it to the customer to help manage the risk and premium,'” Toyota says.
“I really think it is going to be about how companies can find a way to use (data analytics) to enhance the customer experience, rather than just pad company profits,” says Van Kooten. “There is a lot of work to be done by the insurance industry on making insurance products and the entire experience more user-friendly,” he contends.
“Having good data and generating high-quality analysis doesn’t provide value unless the company changes the way it does business. It has to be in the interests of all stakeholders, and most critically, the customers,” Mellor adds.
Sources say, however, that there are several obstacles that line the path of data analytics and meaningful business results for homeowner insurers. Broad industry challenges include the availability of analytic resources, both human and technology, access to verifiable data and the ability to modify old rating structures and underwriting processes to reflect new analytic approaches.
“There needs to be a steady state of investment, but also an underlying process to make it work,” Toyota notes.
“There are different types of resources that go beyond pure actuarial work. The external challenge is making sure that our broker partners and customers come on the journey with us. The internal challenge is making sure the benefits that we are deriving out of data analytics can pay for the investments, so we don’t have to pass it on to consumers.”
KPMG Canada’s Mary Trussell points out that a global KPMG study, called Transforming Insurance, identified integrating data analytics into existing systems as one of the top challenges facing property and casualty insurers.
“It is really about turning theory into practice,” she suggests. “It is becoming much easier to migrate data and because of that, we are seeing insurers step up to that challenge. So instead of having fragmented legacy systems, they are investing in unified platforms to give them a much better grounding to analyze data,” she reports.
Van Kooten says Economical Insurance is in the process of changing its legacy policy administration system to allow for easier integration of data analytics.
“Going to a space where the industry is really interested in quality of data and collecting more data, most insurance companies are finding that their legacy systems are not getting the job done,” he says. “I think that will be a game-changer as well.”
Access to reliable, clean data is a necessary precondition for sound analytics, but some say that inaccurate information still plagues insurers on personal property files.
“Believe or not, address quality and related data has been a specific challenge for some carriers,” McCutcheon reports.
“Deciphering an address to match it with other data sources and to be able to pinpoint its exact roof-top location can be a challenge when dealing with years of renewal business or legacy data,” he adds.
“Postal codes and municipalities have been scrubbed and standardized in some cases, but rural addresses, streets and valid unit numbers still pose problems for carriers,” McCutcheon says.
Another potential pitfall is whether or not new data analytic approaches will be embraced by insurance companies rooted in traditional rating structures.
“One other challenge remains the adoption and acceptance in trusting the output provided by predictive models and solutions,” says McCutcheon. “Old ways of pricing risks rely heavily on the insured, or the broker or the agent, to address many subjective questions about the home, such as the quality of finishing throughout,” he points out.
Insurers that move to data-based predictive modelling remove the “art” aspect of rating and underwriting to a more scientific approach. “Now, predictive analytics can remove these subjective questions and, instead, draw from a greater pool of structured data,” says McCutcheon.
Walter contends the need for change in how data is analyzed and used in the business requires leadership from the top of the insurance company. “One of the most important issues is executive sponsorship,” he argues. “How does this fit within the organization? Who is the sponsor? Who is driving it forward? It is a cultural change issue and it does require executive leadership.”
Walter also cites Deloitte research into how companies in all sectors of the economy integrate data analytics into practical business decision-making.
“A key part of this is that winners are those who are able to achieve bite-sized success,” he notes. “It is not a ‘one-hit’ or ‘do-it-all-at-once’ approach. It is about ongoing improvements to the business that are most likely to be sustainable.”
Mellor echoes these comments in his description of SGI Canada’s approach to data analytics in personal
property insurance. “This is an ongoing exercise; it is not something that you launch and you get perfectly right the first time. There is incremental learning as you continue to implement analytics,” he says.
The evolution of data analytics in homeowners insurance may hinge on the ability of insurance companies to seamlessly integrate new techniques into their business pricing and underwriting processes – while also keeping customer needs at the forefront.
Toyota issues a word of caution on what she calls the potential dangers of “micro’ segmentation and a “hyper-analytical” approach.
“We have to make sure we have the customer needs and preferences in mind,” she advises. “You can keep segmenting and segmenting to a micro level. At some point, you have to step back and make sure the customers’ needs and wants have a dimension in what could be a hyper-analytical environment,” she goes on to say.
“We are trying to find that optimal balance between data analytics and customer needs,” Toyota suggests.
“I think it is about being agile in terms of being able to integrate data analytics into the business,” Trussell observes. “The interesting question is, Will it be the larger players who succeed or can the smaller players be more nimble?” she asks.
“Insurance companies of all sizes now have higher-quality data and more analysis and expertise,” Mellor comments. “So there has been a flattening of the playing field. Predictive analytics is no longer the domain of just the largest, most sophisticated companies; it has really become table stakes,” he says.
“There are a number of first-movers, from all tiers, who will drive implementations fiercely into their operations early on,” McCutcheon observes.
“Based on the interest we are getting, the increasing property claim costs, the natural evolution of existing strategies and the need to protect against anti-selection, a dozen or so more will adopt some type of analytics on property in the next two to three years.”
With early-adopters on board, many sources say the competitive advantage will narrow as quick returns gradually diminish in value.
“Companies will gain advantages in the short term through data analytics, but these will close relatively quickly as others adapt,” says Walter. “The focus tends to shift to finding new and valuable sources of data.”
These “new sources of data” could involve smart home applications and data from personal property monitoring devices.
“Some of this is in its infancy, so there is so much potential for data we could use,” suggests Van Kooten.
“You look at all the smart-home monitoring devices people have now, such as security systems with built-in water detectors and remotely controlled thermostats. Insurance companies would love to have access to that information,” he says.
While insurers seek out new forms of data for competitive advantage, Walter contends that it is the expertise of data analytics itself that will separate successful insurers in personal property lines from the also-rans. “The long-term winners are those who will see data analytics as a core capability of the organization,” he says.
“It is not a new product or new initiative. There are opportunities across revenue generation, claims cost management and operational efficiencies. The winners will be those that regularly invest in opportunities across that whole spectrum by building a core capability within their organization,” Walter suggests.