When the chief executive officer of Plymouth Rock Home Assurance spoke to Carrier Management in mid-July, it just happened to be on a day that online shoppers of books, clothes, electronics and other household items eagerly anticipate.

“If I could be so bold as to bug you when you’re shopping on an Amazon Prime Day like today, and pop my [homeowners insurance] rate up because I know your address is in your record, I’d like to be able to” do that, Bill Martin said during an interview focused on innovations he has helped to introduce to the homeowners insurance market over the course of a nearly 40-year career and what he looks forward to in the future.

Martin noted that the ability to actually show insurance prices to Amazon shoppers is likely restricted (by rules barring the sharing of customer addresses). But the example served to illustrate his idea of being able to present insurance prices in noninsurance situations in the future—without paying for expensive advertising—a prospect made possible by an innovation that he has already brought to Plymouth Rock.

That innovation is the ability to pre-rate any home in the Northeast states where Plymouth Rock operates, something that Martin first championed more than a decade ago in a different region of the country—at Florida-based Bankers Insurance when he served there as president.

“There are obvious areas when you’re shopping for insurance, whether it be home insurance or auto insurance, for us to present a rate. But there are also places where maybe they don’t expect it. Maybe you’re shopping for a home on Zillow and you already see the average mortgage cost per month,” and Plymouth Rock delivers an average insurance premium with that too,” Martin said, looking to the future.

“We have more patience to refine the model and make it work. Every single one of these models is wrong…You have to have a very systematic and aggressive way to find the errors and fix the errors.”

Bill Martin, Plymouth Rock Home Assurance

He added, “I’d like to see us in just about any place that any insurance is quoted—life or commercial or whatever—because we can present the rate at a relatively low cost and it’s cheaper than advertising.”

The pre-rating innovation, dubbed “blind rater” at Bankers Insurance and known as @Home in the evolution that exists at Plymouth Rock Home Assurance, works a bit like some of the capabilities that InsurTechs started to bring to the insurance space more recently—delivering insurance quotes with just a few pieces of identifying customer information while accessing public information in the background. To Martin’s mind, however, 42-year-old Plymouth Rock has competitive advantages that are unmatched by the newer startups. One of them is a distribution force that can use the @Home product to provide home insurance quotes in a matter of seconds or to service those customers that will eventually seek out policies after being served up quotes on noninsurance platforms.

Another is patient capital that is willing to bear through the challenges of inevitable tweaks needed to adjust the performance of predictive models.

Martin spoke about all of that, and also shared his thoughts on industrywide challenges, during a virtual interview on Prime Day. (Related article, “Are Captives the Answer for Uninsured HOAs? Utah Opens Door for Personal Lines“)

Rating Blindly: Building Blocks of Disruption

The blind-rater project earned Bankers Insurance Group the Analytics Model Insurer of the Year Award from technology research and consulting firm Celent in 2014. “We were deciding how much to charge a risk without ever having met the customer,” Martin said, describing the goal of the model. “It was in advance of even quoting them or being asked for a quote. You could say we were going in blind, but we weren’t. We were using parts of the environment that were available to us—public data—and saying here’s how much it will cost to insure this particular home prior to giving the quote.”

When requested, not only would the quote be able to be presented to a customer, and be bindable, at a much faster speed, “but you could also target your marketing to those quotes that you’re adequately priced for,” he said, noting that Bankers would simply ask for an address to deliver a rate.

Martin explained why this capability is particularly valuable when carriers find that their indicated technical rates diverge from actual rates they’re able to charge—a consequence of the fact that regulators may cap the amount of rate change permitted each year. “In a state like Florida for homeowners, there was a lot of advantage to introducing a program like this,” he said, explaining that being armed with knowledge of the actual technical price of each property informs the carrier when it should decline to quote a risk—perhaps one with high catastrophe costs. “If you decline a quote every time your rate’s lower than your target margin, your actual average margin will be higher than your target,” he said.

In fact, he reported that in 2012, by only quoting when a rate was high enough to meet the technical price, and with only a small amount of nonrenewals to the existing book, Bankers saw “very rapid correction” of the existing programs’ attritional (non-cat) loss ratio—from the 30-35 range all the way down to the low-teens.

“We could cut off the tail on the existing book and only bring in business that improved it. [And] it was also really easy to quote and easy to market to people we wanted,” he said, noting that paying bonus commissions for the desired business was part of the strategy. According to Martin, Bankers was also able to rapidly grow its home insurance book from $50 or $75 million to $200 million-plus,” he said.

“It seems simple to say, ‘Don’t quote when you don’t have a high enough rate,’ but most people don’t do that on a per-risk basis. They do it on a class basis,” he said. Absent this type of risk-by-risk indication, “there are natural subsidies in everybody’s book because we’ve taken risks at rates that sometimes are too low and sometimes are too high.”

“If you have a sophisticated pricing algorithm, you can get rid of those subsidies,” he stated.

In Florida, Bankers was also able to use technical price indications to pick the policies it would take out of the state’s insurer of last resort—Citizens Property Insurance Corp., according to Martin, who said that Citizens “provides one of the biggest databases in the industry under one regulator and one carrier that you can obtain” for carriers to analyze. “It was actually more than 25 million home years,” he said, referring to historical data in addition to existing in-force policies.

Bankers, he said, took information it needed from the database—time insured, losses incurred, date and perils of the loss—stripping out other data fields and then attaching public data to it.

“We ran a kind of contest between analytical firms of who could get us the best model,” he said, reporting that Bankers aimed to move past generalized linear models in use at the time to embrace more sophisticated machine learning models. “With that, we were able to blow away the fit of the unconstrained rate vs. the actual risk observed…”

“All that technical talk is a way to say we were able to get a much better pricing model out there and do a much better job of selecting risk. The result was that we could do a big takeout at Citizens as well as grow our new business book at a profit.”

As for the public data elements that the new model used to determine the technical rates, Martin said different data providers, including one large one, provided 2,500 pieces of data per address. “We’re way higher than this now,” he said, referring to Plymouth Rocks’ iteration of a similar process and noting that data includes ZIP code-specific information, such as crime rate and weather patterns, as well as property-specific data elements like square footage, market value and replacement value.

“We used public data not because it was accurate but because it would be systematically inaccurate,” he stressed. He explained: “If I go to 50,000 customers and ask them the questions, they’ll all answer them differently. One actually knows his square footage, another doesn’t but wants to move on rather than go look it up.” The same type of hit-or-miss situation exists with customer knowledge of roof age. “It’s a random answer whereas a database of assessor values is more of a systematic error.”

Martin said data with systematic errors is better to use in a model that aims to get the right rate rather than the correct value to put on a dec page. “Getting the right rate is the important part because it is rare that an insurer pays for a total loss.” They pay for partial losses 99.5 percent of the time, he estimated. “Partial losses are important to predict, and using the public data gave us a much more accurate way to do it.”

(Editor’s Note: Seven years after Martin moved on from his role at Bankers, the carrier made a decision to stop writing personal lines business in Florida, remaining active in commercial lines.)

Patient Capital

Plymouth Rock’s pre-rater, which won silver in the Efma-Accenture Innovation in Insurance awards in 2021, has building blocks similar to the Bankers product, according to Martin, who joined the Boston, Mass.-based insurer almost eight years ago. It’s not exactly the same because data sources are different, modeling techniques have improved, and the humans that built it are different, he said, noting that prior loss history and credit, not included in Bankers blind rater, are incorporated in the modeling for @Home.

“It looks dramatically different today to us. To the consumer, the experience is very similar.”

He continued: “You have a lot of companies calling themselves InsurTechs saying, ‘If you give me a minimum amount of data, I’ll give you a quote right away…They all say that, but very few of them have built this pre-rating system,” he said. “It takes a lot of time, a lot of data engineering, a lot of cleansing, a lot of extraordinary learned technique in actually using it and coming up with a model that doesn’t have ruinous error.”

Coaxed to offer details about the competitive advantages of building the predictive model within an incumbent insurer rather than at an independent InsurTech startup operation, Marting listed three.

“There’s no great insurance story that isn’t a distribution story. We have existing distribution, and we’ve earned our good reputation with that distribution,” he said.

Another distinction is a large capital base, he said, noting that de novo startups need to go out and raise capital. According to the most recent annual report, stockholders’ equity for Plymouth Rock Company and subsidiaries stood at $935 million as of Dec. 31, 2023, supporting home and auto insurance operations. (In addition, equity for a managed reciprocal, was $646 million.)

“We have more patience to refine the model and make it work,” Martin added. “Every single one of these models is wrong, even in its latest iteration after lots of learning. It’s wrong. So, you have to have a very systematic and aggressive way to find the errors and fix them,” he said.

“The report card is your direct loss ratio,” he said. “Your report card to investors is your net combined ratio, but your report card for pricing is whether you’re generating a better loss ratio than competitors and thus charging the right rate,” Martin said.

Although Martin said that there is no comparison between the homeowners insurance direct loss ratios generated at Plymouth Rock and startup competitors, like other traditional players in the homeowners segment, Plymouth Rock has been struggling with high reinsurance costs and the impacts of inflation. As a result, the carrier’s net combined ratios in recent years have not been where leaders want them to be.

“We absolutely aren’t satisfied with our home insurance profits,” Martin said. He added that Plymouth Rock has a higher bar than InsurTechs to reach in order to be satisfied. “They’re trying to get an ‘A’ Demotech rather than an ‘A’ AM Best… They’re trying to buy as little reinsurance as possible while we’re trying to buy as much as we can.”

“So, our net results don’t look as good as we want because we are so disciplined about the financial protection for the bad [year] that we’re going to have,” he said.

Still, the patience that Martin referred to is evident to any investor in the privately held stock company that has read through a series of annual letters written by Plymouth Rock Chairman Jim Stone, a former insurance commissioner in the state of Massachusetts and a former chair of the United States Commodity Futures Trading Commission, who founded Plymouth Rock Assurance in 1982.

“It’s easy to forgive one unsatisfactory year when the long-run picture remains impressive,” Stone wrote in an annual letter for 2021, when Plymouth Rock’s home insurance business posted a 118 combined ratio—partially attributable to Hurricane Ida. Martin, Stone noted, still fulfilled a stated mission to grow homeowners business—tripling the premium volume while delivering underwriting income over entire period of “five rapid expansion years” with the introduction of “customer-friendly marketing tools.”

In earlier annual letters, Stone detailed the goal of aggressively expanding homeowners to counterbalance what he viewed as a long-term trend of declining auto insurance premiums across the industry that would arise as auto safety features become more widespread. Stone also forecasted that a company advantage of mathematical sophistication in auto insurance would eventually erode (as competitors caught on to the power of advanced analytics in that line of business).

In the homeowners world, Plymouth Rock had previously invested in a direct-to-consumer platform, Homesite, which was sold to American Family in 2013 for more than $600 million—bolstering Stone’s confidence in using sophisticated tools for managing homeowners risk.

“We would have paid a good bit of money to grow like that,” Stone said in his letter about 2021 results, which nonetheless directly called out the innovative tools of Martin’s group as a contributor to 2021’s unhealthy homeowners net combined ratio. “The Home team undercharged for non-catastrophe losses just about everywhere except Massachusetts,” he wrote.

“Caution is clearly called for now, but [Martin] still has the green light,” he wrote.

In subsequent letters, Stone continued to deliver the straight dope about the situation, using phrases like “self-inflicted wounds” to explain missed targets in 2022 and 2023. Inflation and a rapidly hardening reinsurance market added to “mistakes of our own making,” forcing the home insurance group “to start to climb from inside a hole it had dug for itself,” he wrote in the 2022 letter, which reported a 112 net combined ratio that was 15 points over a budgeted prediction for 2022.

The fast quote-and-bind innovation was “designed to be especially easy for customers and agents to use,” Stone wrote. “The notion is a good one that [Chief Operating Officer] Andy [McElwee] and I fully endorse, but the original details incorporated rather too much ease of use and thus invited adverse selection,” he continued, also noting that “base rates were set substantially too low in just about every state,” compounding the problem of adverse selection.

Expanding the view further back and into the future, Stone noted that the home insurance group’s volume grew from $75 million to more than $330 million over the course of Martin’s tenure, while providing a net profit for the whole period of about $20 million. “Andy, Bill, and I all see the 2021 and 2022 results as temporary bumps in the road for a business segment with superior prospects over the long horizon,” he wrote.

Updating the volume to $360 million in his February 2024 letter about 2023 results, Stone again balanced news of a disappointing 2023 combined ratio (113 excluding the 3-point impact of December 2023 storm) with a long-term view, also detailing Martin’s game plan to solve “fixable problems” by shedding policies that don’t meet Plymouth Rock’s risk appetite and raising rates on those that are simply underpriced while working to separate out the best policyholders that will be protected from price hikes.

Martin told Carrier Management, “It’s been a challenge for me because we would be there [at targeted returns] if it weren’t for the rapid inflation that we all experienced in the last few years. But in my regulatory environment, I have to methodically catch up to inflation as opposed to just charge for it tomorrow.”

Although part of the growth in Plymouth Rock’s home insurance book came from acquisitions of policies in New York and New Jersey from MAPFRE and Farmers in 2018, Martin confirmed that those books generally have performed better than organically grown business. “As we grow in states where we have small market share, which would be everywhere but Massachusetts, your distributors who are new to you are not necessarily giving you their best book first. It’s not an intentional bias…Their first check company is not me. It takes a while for me to be their favorite. I have to earn those stripes. And so, our pricing and our underwriting processes were not defensive enough to recognize that we weren’t going to get everybody’s best tranche first,” he explained.

“We’ve made a lot of adjustments. We’ve sharpened and added credit curves. We’ve added screening steps in our inspection and post-buying process to be sure we’re not getting risks that don’t fit [our] homeowners program. And we’re paring back the business that we got in the first couple of years of the program that was clearly rejected by other companies or improperly rated by us,” he said.

Martin reiterated a point he made earlier: “Every model is wrong. Most of this correction would’ve happened much earlier if it weren’t for catching up with the inflation and the limits of what regulators really should or could allow,” he added.

The benefit of patient capital is clear to Martin. When you raise money as a startup, “your capital is not patient—and might be rewarding the wrong thing if they’re not rewarding profit. Eventually you have a ‘come to Jesus’ moment where you have to have the profit. And that is very impatient capital when it’s time to deliver profit, whereas our capital said, ‘We see where you’re going with this. We want you to correct it and correct it fast, but we’re not abandoning the idea [because] we see that it’s working.'”

Looking back on his career, Martin describes most of his innovation projects as “sponsored startups”—backed by people who have a larger business or a larger capital pool. They are “willing to not necessarily write off what we’re doing but at least understand there’s a fire test to any startup that you have to persevere through and make up for,” he said, referring to the sponsorships of Plymouth Rock and Bankers, and perhaps to the capital pool of a liquor liability carrier (North Pointe Insurance Company) that hired Martin to create an automobile insurance division in the early 2000s.

At Plymouth Rock, “I’m in the make-up-for mode right now where we’re very likely, over the next few years, to make all the pain of catching up with the severity increases in our business worth it,” Martin said.

Stone offered a similar conclusion in his latest annual letter. “When inflation abates, the future will look much brighter,” …[T]he promise of rewards is substantial when we get out of the ditch,” Stone wrote. “No one at 695 Atlantic doubts that we will build a large and profitable homeowners business,” he concluded, referring to the address of the company headquarters.

At some point in the future, if Martin convinces online sellers and app creators to give him some screen space to display Plymouth Rock home insurance rates to customers engaged in activities other than buying insurance, he’ll have another avenue to profitability. In a line like homeowners, with catastrophe exposure, carriers need to spread risk, he reasons. But “the chances of the people that you’re willing to spread to going out and getting quote are really thin,” he said, expanding on the idea that Amazon Prime Day helped bring to mind. “If I could get there to them, I can do a better job of reducing the cat cost per customer because I’ve spread my cat risk on a wider area,” he said.

Now, if only he can convince the advertising world to catch up with his way of thinking. “It requires a lot of head changing,” he said, reporting that his propositions to share “a little screen space” haven’t yet resonated with potential renters of the space.