For most P&C insurers, a lack of underwriting profitability has been a long-standing problem. According to Insurance Services Office Ltd., insurers have posted net gains on underwriting for just 21 of the 113 quarters since the company began collecting quarterly data. And external factors, including persistently low interest rates, the glacial economic recovery and increasingly intense competition, over which insurers have little control, are exacerbating the profitability challenge. As a result, underwriting, due to its historically manual nature and its potential to increase profitability and reduce complexity, has for many insurers become a target for innovation.
According to “Data and Analytics in Insurance,” a report from insurance technology consultancy Strategy Meets Action (SMA), underwriting is a top-four opportunity for personal and commercial line carriers to apply data and analytics to improve profits.
More than half, 55 percent, of P&C insurers surveyed by SMA are using analytics to improve underwriting profitability and underwriting operations; 15 percent are evaluating or piloting analytics to increase underwriting profitability, and 28 percent plan to use technology to improve underwriting profitability in the next three years. Almost a quarter, 24 percent, are planning to use analytics for underwriting operations in the next three years, and 12 percent are evaluating or piloting analytics for underwriting operations.
“Underwriters’ value-add in the past was to make sure the company was not getting less than the rate that it deserved and that it filed with the state,” says Bill Martin, president of Bankers Insurance Group. “Now the system is there to make sure that the rate the customer is willing to pay is high enough to cover your modeled costs.”
At Bankers Insurance Group, a P&C carrier headquartered in St. Petersburg, Fla., automated underwriting of the Preferred Homeowner insurance product has led to a 300-percent increase in quotes by independent agents compared to 2012, a sixfold increase in new business and a 20-percent increase in the average profit margins for the new business. The time required to generate a quote has shortened from 15 minutes to less than a minute.
The system, which earned Bankers a 2014 Model Insurer award from Celent, an insurance business and technology consultancy, accesses a large business intelligence database that Bankers has repeatedly processed with machine learning technology, Martin says, to create hyper-accurate rating models. The models don’t use traditional underwriting or pricing methodologies, he explains, and are too complicated and proprietary to be filed with regulators. The technology stack includes Revolution Analytics, an SQL Server database, and IBM Cognos for performance analytics; extraction, transformation and loading is accomplished with SSIS and Cognos Data Manager, according to Celent.
For American households, insurance is a top-five expense, Martin explains, and historically the underwriting process has been long, difficult and unpleasant. Prior to the launch of the blind rating system, Martin says the process included 25 questions, 15 of which could immediately disqualify an applicant, including whether the applicant has a biting dog or lives on the shore.
“It was just one of the worst experiences for the money in any industry,” Martin says. “Then, the underwriter walked in assuming the consumer was trying to drive the premium down, not telling the truth or making a mistake. It’s no wonder the average insured was upset. From the beginning, we were trying to fix this ridiculous process.”
In addition to improving the customer experience, Bankers was looking to find the level of premium customers are willing to pay for an improved consumer experience. The intention wasn’t to find the maximum margin, Martin says, but rather to see if there was a market for people who would prefer to opt out of the normal underwriting process.
“We use the kind of variables that we’re used to, we just combine them in a different way,” Martin says. “There’s a lot of analytical, statistical work done in advance, and what you end up with is a table of addresses, minimum rates and expected rates; and then it’s very easy to deliver. You give me an address and I give you a rate.”
The automated underwriting system accesses policy and loss history data, quote conversion and agent histories, plus publicly available data from aggregators including LexisNexis and CoreLogic. However, that data is not used to figure the actual rates, Martin says. Rather, the system is used to verify that the rate the customer is willing to pay, and which is filed with the state, is high enough to cover the modeled costs.
“We’re using the traditional rating model that we file with the state to rate the policy,” Martin says. “That kind of expands the number of pieces of data and databases that we can use to test and to see whether we can get a more accurate model than we’d get with a traditional underwriting system.”
The system is currently designed for use by agents in Florida, who are prompted for an address and how much coverage the consumer wants to buy in terms of home replacement costs. While consumers aren’t expert at determining their replacement costs, Martin says Bankers doesn’t want to go below that number. “We want to make sure we get the minimum of what they want in terms of coverage.”
The system then returns an estimated rate. In Florida, regulators require insurers to ask whether a home qualifies for wind mitigation credits. “Florida requires us to give a big discount if the home has been retrofitted for hurricanes,” Martin says. “That’s not something that’s available on a public database.” If the customer decides to buy, the application then asks for just enough information for the customer to pay and to communicate with them in the future, Martin explains.
The project took about a year, and about 25 people across departments helped develop the Preferred Homeowner product, Martin adds. “This has gone a lot more smoothly than anywhere I’ve ever introduced change. The consumer is very willing to give you a bigger margin if you give them a better experience. So that is the right way to increase profitability, not just to have a more accurate rating model.”
Consumers benefit in other ways, too, according to Martin, as fewer renewals include rate changes. “Our underwriting scheme locks in their rate and doesn’t disrupt somebody just because we insured their next-door neighbor, creating aggregation. We have a much more stable rate than our average competitor, and that will add to customer satisfaction. Our early indicators are increased quoting and meeting our profit goals, and we’re beating what we expected.”
Martin says the company will continue to improve and expand its use of the system by offering it directly to consumers, online agencies and other distributors. “There’s a new clearinghouse in Florida that gives us an opportunity to quote more often; we’d like to see other states adopt it for their residual markets, for TRIA [the Terrorism Risk Insurance Act] and for Louisiana, maybe the wind pools in the Carolinas. We also will look at ways to do pieces of this with our other products.”
Farm Bureau Financial Services, a regional P&C insurer, has been using analytics to increase underwriting profitability in personal lines since 2005, explains Steve Wittmuss, commercial vice president at the insurer. Three years ago, the company began applying analytics technology and know-how to commercial lines, including business owners’ policies (BOP) and workers’ compensation products.
The company had significant loss ratio issues, Wittmuss says, and needed to fix them quickly. The first predictive model the company built for BOPs was a scorecard for quotes and renewals that assigned points to each of 20 characteristics to figure a final rate. That model was developed internally and without help from vendors. However, to build an underwriting platform for workers’ comp would have taken years due to the complexity, Wittmuss says. “We started looking for a partner that we could work with, that we could implement very quickly, and that’s how we came upon Valen Analytics’ InsureRight Platform.”
Farm Bureau Financial previously did not have an investigative underwriting process for workers’ compensation. Premiums simply were based on class codes and payroll, Wittmuss says, and workers’ comp was not sold as a stand-alone product. The company would take it on based on wanting the rest of an account’s business.
“We just wrote it where we had it listed; we had no scheduled pricing,” Wittmuss explains. “Workers’ comp can be a much better — or a much worse — risk. We don’t have multiple companies to tier with. That became an issue. When we did decide to give credits or debits, we didn’t know which ones to give it to. Maybe an underwriter got hurt on one specific loss and decided never to underwrite that again. But when you look at the whole universe of losses, it could be good risk. When you start using analytics, you get to use all that data out there. That’s what analytics can bring you.”
Giving underwriters a tool was critical to getting them engaged with the workers’ comp product, Wittmuss says. “I can’t say start underwriting work comp’ and expect them to know what to do. They’d either have avoided it or pretended to know what they’re doing. Valen, for us, works great. It gives them a quick snapshot.”
When agents quote workers’ comp business, the system directs them to an underwriter to enter the information into the Valen platform, which ranks applications on a scale of one through 10 based on desirability.
“The ones, twos and threes should be really good business,” Wittmuss says. “Underwriters don’t need to spend time with those. We can give them more credits, if we need to, to bring the business in. The fours, fives and sixes should be right around the scheduled rating. You get to the sevens, eights, nines and 10s, that’s where we now tell our underwriters to spend time.”
The return on investment has been substantial, according to Wittmuss. “We’re paying about $200,000 for the product per year, and we’re making about $600,000 back just on renewals, just by moving the premium up on the classes we need to: our eights, nines and 10s.” And there are other opportunities to use technology to improve underwriting profitability, Wittmuss says.
“We upgraded our business owners’ a few years ago, and now we’re working to upgrade our workers’ comp. That will include a number of different things. From a pricing point of view, it will include more of a scorecard process, and so we will use more than just a Valen score. We’ll use other data, too, to get an even better feel for what each risk should be. While agents are entering the data, [the system] will pull information from other sources, internal or external, and quote it for them. And, as far as the agent can tell, it’s real time.”
Change management was handled with constant communication and respect for the underwriters’ and captive agents’ knowledge and time, Wittmuss says. “If we put too many roadblocks in the agents’ way, they’re not going to be able to sell it or they’re not going to be interested in selling it,” he points out. “It’s a little bit of a black box, and agents always know somebody, their neighbor or somebody, who is the exception. And they’re not. You have to get them to understand: This person may be an exception, but he belongs to a group where he’s not the exception.”
Farm Bureau management made sure the underwriters were aware that the company was talking with Valen long before Farm Bureau signed the contract, Wittmuss says. “Some of our people have been underwriting for a long time. If you’re not careful, they think you’re challenging their knowledge,” he adds. “We still give them a band they can write that business in. They have a lot of ability to do stuff they need to do to bring the business on.”
Writing new business has been easy, but some renewals have proved to be a challenge, especially if they have been on the book for a number of years without a loss, Wittmuss says. “We really focus on those eights, nines and 10s and tell agents, Here are the ones where you really need to get the right premium for the risk.’ If we lose a piece of business because of pricing, it’s OK. We don’t want to write it at a loss. We have to trust that the data is correct.”
The story with consumers is slightly more complicated. “If their rates go down, you never hear from them. Those aren’t the people who complain,” Wittmuss says. “Many people, when we’ve tried to increase the rates, go out shopping and then they stay with us. That tells me that we were underpriced and that other companies are using predictive modeling. If we didn’t do this, we’d be at a competitive disadvantage.”
This story first appeared on the Insurance Networking News web site.