Modeling the Future
It is no wonder that predictive modeling is emerging as a profitable best practice in commercial lines insurance. The results are just too compelling to resist. Predictive modeling—a broad term that encompasses statistical techniques to analyze current trends and forecast future events and their consequences—boosts premium accuracy, finds sweet spots through market segmentation and saves on loss adjustment expenses and loss costs.
Also known as predictive analytics, predictive modeling applies statistical methods and algorithms to consider traditional and nontraditional variables indicative of future events. Predictive modeling typically confirms the conventional wisdom of experienced underwriters while finding new and variable combinations to forecast risk.
“It is a game changer in opportunities for [brokers] and will change the nature of the relationship with carriers,” says Roger Burkhardt, president and CEO of EagleEye Analytics, an international property-casualty predictive modeling provider whose clients include Guy Carpenter and Milliman. “There will be more transparency in the profitability of agents.”
Predictive modeling is taking place in different pockets of the property-casualty industry, which makes it challenging to gauge how many brokers are being affected and to what degree.
Many brokers are not addressing the results of insurer models, says Chris Gagnon, director of strategic technology at The Council and president of Tiebeam Partners. But those with the technological interest are actually developing predictive models to enjoy the same benefits as insurers, Gagnon says, though this practice is not widespread. “Everyone knows that data is power,” he says, “and that the ability to interpret it to help the insured is even more powerful.”
Off-the-shelf, plug-and-play predictive modeling software is not yet available, so enjoying this strategic advantage requires in-house predictive modeling capacity. Adopting predictive modeling is not coming any easier for brokers than it is for insurers.
Brokers are hopeful about predictive modeling’s potential, according to Gagnon. “The big angle is differentiation,” Gagnon says, which gives brokers new market segmentation and service opportunities.
Predictive modeling, Burkhardt says, can create a list of profitable prospects for agents and carriers to identify the combination of factors that define an attractive market. Using predictive marketing, he says, “an agent can become very expert in defining a better way to quote insurance for a narrow segment of business, like dry cleaning or auto shops.”
Agents can also use this approach “to identify the most important clients to retain and grow and then build sales and service programs to support these goals,” Burkhardt says.
Imagine making predictions right at the point of quoting and buying, and then showing profitability with the least deviation, Burkhardt says. “You can optimize your income because you know the profitability of the prospect,” he says.
And since agents will better understand how much profitability they are bringing to carriers, the profitability partnership between the two grows stronger, he says. “Over time,” Burkhardt predicts, “agents and carriers will be working together and developing loyalty programs.”
Predictive modeling also reveals the factors behind the premiums so brokers can become strategic partners with clients by helping them improve their risk profile. “Some insurance professionals are calling themselves portfolio managers,” Burkhart says. Such brokers, he says, can show they are best informed about the market and provide the best options for a potential risk.
This is already happening for producers who sell health insurance and use predictive modeling to help clients mitigate and prepare for future losses. “Predictive modeling took what we thought we knew and drilled it down further,” says Joe DiMaggio, senior vice president of Kelly Benefits Strategies, which has been using predictive modeling software for four years.
By using more specific information about an insured’s employee population, predictive modeling helps the brokerage provide more effective loss mitigation options. Think targeted wellness programs to improve employee health.
Predictive modeling considers new and different factors, often from augmented data, to boost premium accuracy, says Doug Johnston, vice president of partner services and product innovation at Applied Systems. This approach could change the give and take between brokers and their underwriters. Brokerages that are not front-end underwriters are generally not privy to the details of how the carriers’ predictive models are changing underwriting rules, Johnston says. Stoll adds that agents might see changes in their carriers’ underwriting appetite “without warning.”
These changes can cause several surprises. Insurers could choose to quit writing new coverage for certain risks, Stoll says. Thus the accuracy of predictive modeling to price a risk is a double-edged sword.
“It’s pretty typical that with predictive modeling you will have some companies that have not had a recent loss suddenly look like a bad risk according to the models,” Stoll says. Conversely, employers with effective risk-management programs who comply with laws and regulations and are financially healthy could see premium reductions.
To be competitive, agents must invest more time and effort to ensure they are on the same page as their carriers, Johnston says. “I see more and more agents having more serious annual meetings with carriers to ask what markets they want to target,” he says.
At the same time, Stoll explains, the biggest challenge carriers face is the degree of transparency for agents and the end customers. “For the companies, this is a very big issue,” he says.
Meanwhile, carriers are at various stages of adopting them. Insurers are still experimenting, and their predictive models can vary significantly by carrier, says Brian Stoll, a director at Towers Watson. Generally speaking, the larger the carrier, the more likely predictive modeling is making its way from actuary to agent. “Predictive modeling is also evolving at different paces depending on the insurance line,” Stoll says.
Modeling can benefit agents and brokers of all sizes. “How agents respond to predictive modeling is more driven by their adoption of technology and their overall attitude toward the advantages of technology,” says Karen Pauli, research director for TowerGroup, a financial services research firm. “I know some one-man shops that are totally tech-focused and have no problem with the value of predictive modeling.”
Despite the advantages, implementing predictive modeling is “not an easy culture decision,” Pauli says. Insurance executives who believe underwriting is more an art than a science are not yet adopting modeling.
But insurer interest and actual implementation in commercial lines is growing, according to a predictive modeling benchmarking survey of 69 carriers that Towers Watson completed earlier this year. Respondents who plan to use or are using predictive modeling are: insurers of commercial property and commercial multi-peril business owners (73%), workers comp (72%), commercial auto (63%), general liability (52%), and specialty lines (36%).
Insurers already applying predictive modeling to their pricing and underwriting practices report impressive results, according to Towers Watson. Eighty-three percent report rate accuracy improvements; 76% report loss ratio improvement; and 73% report greater profits. These insurers also report an expansion of underwriting appetite (49%); renewal retention (46%) and market share (39%).
Predictive modeling started changing the pricing game in personal lines about 30 years ago. At that time, Stoll says, Progressive Insurance began considering nontraditional variables to fine-tune risk selection. Predictive models in personal auto revealed risk relationships between a customer’s credit score, education, occupation and/or income for personal auto insurance.
The impressive financial results from predictive modeling in personal lines transformed it into an industry best practice. Seeing the benefits of predictive modeling in the personal lines marketplace prompted practitioners to apply it on the commercial side. But the transition to commercial lines has been slow going. Adapting the tried-and-true from personal auto, for example, to commercial auto, is not a big stretch. But for other commercial lines, Stoll says, slow adoption has been the result of data issues. Data in commercial lines is thinner because, compared to personal lines, there are lower claim frequency and fewer risks. The risks are heterogeneous as well, Stoll says, making data compilation more difficult.
Carriers also want to collect more predictive information from brokers but are in a “constant balancing act” of trying to figure out how much detail they can collect without making it too difficult, Stoll says. An extra data element for a workers comp insurer, for example, would be providing the total number of employees. That identifies one more measure of exposure.
Meanwhile, insurers are frustrated with brokers and agents who are not keeping up with technology. “Agents will be on agency management systems four versions old, thus not being in sync with current functionality,” Pauli says. “Insurers find it increasingly difficult to maintain relationships with agencies that require manual intervention and nonstandard workflows.”
In fact, Burkhardt says, carriers’ use of predictive modeling is going beyond claims and underwriting to assess the business from its producer channels. “This creates an important driver for an agent to understand the value of the business it is bringing to the carrier and ensure that its strategies are aligned with the carriers’ profitability goals,” he says.
Experts agree that to be successful in the future, agents and brokers must embrace automation and predictive modeling. “It’s like most tools,” Burkhardt says. “Agents who quickly adapt will be more competitive. If you are slow to adopt tools and are less knowledgeable, you lose the edge.”