The insurance industry has undergone a transformation over the past five years. Once manual and paper-based, insurance professionals now recognize the overwhelmingly positive impact technology has had on efficiency and customer relationships.
We can expect the acceleration of tech-enabled customer experiences that promote dynamic customer interactions and empower the insurance industry to deliver tailored solutions and products to meet customers’ personal risk advisory needs. Here are a few of the innovations we will see take off in 2020.
Pay-as-you-go car insurance. This unique approach to car insurance is especially appealing for urban consumers who own a car but rarely drive it. Rather than forcing occasional drivers to spend money on the same caliber of coverage as someone who drives every day, on-demand insurance means car owners can activate coverage only when they need it or pay per mile. The program requires a significant investment in technology that provides direct, self-service options for consumers, but we’ve already seen Liberty Mutual and Metromile offer this flexibility. We can expect many of the larger brands to begin offering similar options.
Self-service capabilities. Sometimes it’s just not feasible to call an agent when a need or question arises. Modern consumers increasingly want the convenience of DIY options to allow them to secure coverage or access information on their own schedule. To meet that demand, we’ll see self-service portals offer a wider range of features, including risk rating, on-demand premium quotes, home inventory solutions and claims processing. There’s also a growing demand for virtual adjuster technology and drone/satellite integration that help speed claims processes. Many agencies aren’t aware that their carrier partners even offer these tools, so be sure to ask.
Smart home claim avoidance/prevention tech. You’re probably already familiar with tools that track driving habits and offer discounts on auto insurance. A similar tech is emerging to help homeowners head off costly insurance claims on some of the most common mishaps. For example, a large portion of claims are related to water damage, such as a pipe bursting or an old water heater that fails. By implementing smart home sensors, automated controls and other Internet of Things devices, insurers can help homeowners mitigate the damage. For example, a sensor installed on a water heater can help monitor its structural integrity and remaining useful life, giving homeowners a heads up if it looks like a leak is imminent. In addition to offering premium discounts for participating in such a program, carriers might also provide incentives for replacing risky appliances ahead of time to avoid the risk of claim. Other solutions might involve installing a flood sensor or automated water control valve, which would send a mobile notification if a pipe bursts while no one is home, or even automatically shut off the water to prevent major damage. Some carriers already offer smart home kits for free to avoid these types of claims, and, while some homeowners might be a little concerned about privacy issues at first, I expect we’ll see this become table stakes soon.
Deeper data insights to drive business growth. With threats of a potential economic slowdown beginning to percolate, we’re starting to see signs of a hardening of the market in commercial lines. In many cases, premiums are rising, and business customers are looking at all options to save money and reduce costs. Agencies need quantifiable data and insights from smart analytics to help commercial customers strike the right balance of risk, coverage and cost.
That means the demand for industry-wide data analysis will grow substantially, as agents and carriers need measurable and accurate insight into what’s going on across industries — what customers in similar business sectors are buying, what the risks are and what are any potential gaps in coverage. Combined with their experience and knowledge, comprehensive industry analysis tools will enable agents to see the big picture and give business customers data-backed recommendations to ensure they receive just-right coverage.
In the coming year and well beyond, consumers will increasingly demand greater transparency, a higher level of service, and on-demand solutions from their insurance carriers and agents. This means adopting innovative technology will become a norm rather than a novelty as the industry digitizes to provide a more modern experience.
The good news: A wide array of insurtech solutions are now available, providing accessible and affordable tools for even the smallest independent agencies. Adopting insurtech solutions can not only improve agency efficiency, allowing agents to spend more time with customers, but also help agencies spot opportunities for growth. This powerful combination can deliver substantial ROI for investment and position your agency as a modern, tech-savvy partner for both consumer and commercial customers.
We have previously evaluated and discussed the financial performance and operating results of the insurtech trio Lemonade, Root and Metromile. Based on the analysis of the last available data, we think that:
There is a pricing war
The trio is missing an edge and story with respect to gaining a sustainable competitive advantage
At first glance, it appears that all three firms are focusing heavily on containing and improving their loss ratios. Metromile and Lemonade continue to have a relatively stable loss ratio, while Root has drastically improved compared with the prior quarter. At 91%, Root delivered a better result compared with the 105% in Q1-19 but still has some distance to go before getting closer to Lemonade or Metromile in terms of quality performance.
Note: The expenses ratios are not significant because part of the expenses are paid by the parent companies and not reported on the Yellow Books.
There is a pricing war
We believe the insurtech trio is facing a pricing war. We find three data points in support of this view:
Viewed through the lens of growth rate, after a robust Q1-19 for all three players, Q2-19 presents a different story. We find the most interesting perspective by looking at the performance over the last year and half. Clearly, the premium evolution of Metromile is the less exciting story, as I previously wrote. The pay per mile doesn’t seem particularly effective in attracting customers. The “pay per mile” model introduces an element of uncertainty for segments of customers who want to save money and know what insurance coverage will cost them. The only comfortable customers are those who almost never use a car. (We will cover the customer experience and expectations for usage-based insurance in future articles.)
Lemonade has shown consistent growth in the last two quarters and appears to be on target to meet the $100 million annual revenue target. This revenue target is a far cry from the “massive disruption effect” that was expected during its debut. The revenue curve is not yet showing the vaunted hockey stick.
Root is the only of these three players with exponential growth in revenue. However, something happened in the second quarter, and growth slowed significantly as the loss ratio improved. As mentioned in our last article, insurtech D2C seems to be a “price game.”
Let’s go back to look at the top-line numbers. Root and Lemonade registered a net increase in direct premium written compared with Q1-19. Metromile, on the other hand, registered a marginally lower DPW than in Q1-19.
Having looked at the top line, let’s switch our attention to the loss ratios. For an insurance carrier, the loss ratio is really the litmus test that assesses the strength and quality of the top-line numbers. Loss ratio is a fundamental insurance number, and the fact that all three players have improved this crucial metric is a sign of increased maturity for the “not so” fast-growing trio.
All three improved compared with the prior quarter. This bodes well for the trio. Of the three, Root continues to have the highest loss ratio at 91%, suggesting that between new sales and renewal it is still under-pricing risks.
The Q2-19 loss ratios are significantly better than what the three firms exhibited in Q2 of 2018, when Lemonade had a loss ratio of 120%, Metromile was at 95% and Root at 112%. The loss ratios are still far higher than the respective market segments.
Missing an edge and story with respect to gaining a sustainable competitive advantage
In our last article talking about insurtech direct-to-consumer (DTC) as a “price game,” we highlighted how the companies have not been able to make customers fall in love with anything other than “saving money.”
We would like to share some thoughts on the business models of these three full-stack carriers, investigating where and how their approaches might both enable and impede them in terms of gaining a sustainable competitive advantage. What might be the proverbial sling that the “insurtech Davids” can use against the entrenched Goliaths of State Farm, Geico, Progressive, Allstate, et al? Let’s explore.
If we consider the economics of an insurer, there are three areas where you can obtain a competitive advantage that can allow financing this kind of “pricing war”:
The loss ratio
The administrative expenses
Investment income even in the current market conditions characterized by low interest rates represents the main source of profit for U.S. P&C insurers. A recent report by Credit Suisse has pointed out how, “over the last five years, approximately 90% of the industry’s profits have been generated from the investment income (float) component of the income statement.” In 2018, U.S. P&C insurers generated $53 billion net investment income, which accounts for 8.6% on the premium written. Well, all three insurtech carriers were far from this investment performance in the same period.
Probably, hidden in the parent companies income statements, there is some additional investment income obtained investing the cash received by their investors.
However – as of today – the return obtained investing the floating is clearly a competitive disadvantage for the insurtech players.
As explained in a previous article, the loss ratio is the key measure of the technical profitability of an insurance business . The U.S. P&C market showed $366 billion net losses incurred, which means almost 61% loss ratio (net of loss adjustment expenses). Can any insurtech element allow Metromile, Root or Lemonade to have a competitive advantage on the loss ratio?
Metromile and Root are telematics-based auto insurers. Matteo is a fan of the usage of telematics data on the auto business and an evangelist of these approaches through his IoT Insurance Observatory, an international think tank that has aggregated almost 60 Insurers, reinsurers and tech players between North America and Europe.
Based on the Observatory research, four value creation levers have been the most relevant in telematics success stories:
The telematics approach has demonstrated around the world a consistent ability to self-select risks. Simply said, bad drivers don’t want to be monitored;
Some players such as UnipolSai and Groupama have achieved material results by improving claims management through telematics data;
Some other players have been able to change drivers’ behaviors, e.g., the South African Discovery and the American Allstate;
Many players have been able to charge fees to customers for telematics-based services, providing a revenue stream.
Metromile and Root are mainly using telematics for pricing purposes. Root is currently using a try-before-you-buy approach that allows potential customers to generate a driving score in a couple of weeks and obtain a tailored quotation. If consumers buy the policy proposed by Root, they will not be monitored anymore. This usage of a driving score at the underwriting stage could represent a way to price better the risk (if the price is settled at the right level), but the absence of further telematics data doesn’t allow Root to extract any value from the value creation levers above mentioned.
By contrast, the Metromile pay-per-mile approach is constantly monitoring the driver for the duration of the policy, which is the necessary foundation for those telematics value creation levers. As of today, the only area where Metromile seems to exploit the value of telematics data is claims management. The biggest limit for Metromile seems to be the nature of the pay-per-mile business, which has found only a limited market fit (niche nature of the mileage-based approach).
So, we think that both the players are still far from telematics best practices, currently represented by a few incumbents. However, the evolution of their telematics approaches can generate some competitive advantages against many incumbents that are less advanced.
About the possible capability of Lemonade to generate a better loss ratio than other renter insurers, the game is around its famous focus on behavioral economics and the attempt to influence behaviors with the iconic giveback. Everyone (among insurance executives) remembers Lemonade for the fixed percentage of premium it charges — the iconic slice of pizza — while all the rest is used to ensure the company will always pay claims; whatever is left goes to charities. As of today, nothing in the loss ratio shows benefits from this approach. Would you avoid submitting a claim or resisting the urge to file an inflated claim (an unfortunate reality in our industry) because “whatever is left goes to charities.” We find it hard to believe that the fundamental economic incentives will move from individual gain to community gain. Ancient Latins were saying homo homini lupus. (Man is wolf to man.)
Moreover, a recent comment by Lemonade CEO Dan Schrieber suggests that “AI may have played a role in higher loss ratios.” The increase of straight-through processing combined with a vast reduction or elimination of legacy checks and balances can increase the risk of fraud. Sri and his team at Camino Ventures, an AI/ML fintech, believe that this is a natural phenomenon with respect to AI/ML adoption. Ilich Martinez, co-founder and CEO at Camino Ventures, says, “Companies need to take a holistic view of AI/ML integration into business processes and business rules. Piecemeal or spotty inclusion of advanced AI/ML capabilities can be akin to installing a Tesla induction motor on a gas guzzler. It simply does not work.”
Sri expects to see a spurt in fraudulent claims in certain domains where autonomous claims management is introduced or expanded. As bad actors try to game the system, it is imperative that insurance companies expand the fraud detection models and be willing to enter a phase of continuous and strenuous test-and-learns. Eventually, usage of AI/ML can be crucial to achieving better administrative cost positions, but the path to that destination will not likely be a straight line.
The U.S. P&C markets showed costs that represent 37% of the written premiums on 2018 (of which 10% was loss adjustment expenses). Unfortunately, we are not able to see any more the real costs of Root and Lemonade; part of the costs are on the income statement of their parent companies, and we can only enjoy the results of their goal-seek. Efficiency – if it will ever be achieved by these insurtech startups — can be a way to compete and to sustain a pricing war, but the scale matters.
We ask whether the AI-driven approach of a player such as Lemonade will become a key competitive advantage against the current established incumbents. Can an early, extensive and immersive adoption of AI/ML provide insurtechs a competitive edge? Will incumbent leaders be fast followers or late adopters in a domain as critical as AI/ML?
We are already seeing the incumbents make early and deliberate moves to gain an AI/ML capability advantage. From domains like telematics to computer-vision-enabled image processing for claims, we see the Goliaths not waiting to be left behind. They are quickly moving from a “test and learn” approach to wide-scale adoption of numerous AI/ML capabilities and are finding early success. This adoption is accelerated by nimble, agile fintech and insurtech enablers like Camino Ventures that help industry leaders quickly move from opportunity to outcomes. Given these market conditions, we believe that advanced AI/ML capabilities would give insurtech an edge only against incumbents that are less focused, capable or invested in integrating insurtech solutions into their value chain.
We believe the current business models are not yet showing any clear competitive advantages that can make the pricing war sustainable for the three full-stack insurtech carriers. However, they have a lot of cash combined with highly talented teams that can experiment and find new ways to build moats and forts to gain competitive advantages.
The three P&C venture-backed U.S. insurtech start-ups — Lemonade, Metromile and Root — finished 2018 with pretty good results. Quarterly growth was the slowest ever, but all three paid out in claims less than they collected in premium. All three start-up carriers have more work to do to achieve sustainable financials.
A year ago, when I started with my friend Adrian a public conversation about insurtech statutory results, the picture was ugly — loss ratios well over 100%, an aggressive focus on price and promotional messages on company blogs that dismissed traditional measures of success in insurance.
Since the first post titled “5 Dispatches From Insurtech Island,” the conversation has shifted dramatically. Fast forward a year, and one founder said he “messed up an entire quarter” because premium growth turned negative, when in fact the company generated its best quarterly loss ratio ever. In the months since, several start-ups have hired top underwriting talent from their traditional competitors, showing that they increasingly recognize the value of traditional insurance skills.
Skeptics point out that a quarter doesn’t mean much, that there’s a long way to go before reaching sustainability and that each additional point of loss gets harder to take out. True, but the increased focus this year on reducing losses and increasing prices is making a difference.
Here are the quarterly results:
I think – as already mentioned in the previous articles – these companies have strong management teams who could ultimately create valuable businesses. This will take several years, but all three companies are well-funded, even if the combination of statutory capital injections and operating losses consumes tens of millions in capital each year. (The Uber/Lyft model of growing rapidly while also incurring large losses is doubly penalized in insurance because carriers have to maintain statutory capital that increases with premium.)
Here is a year-over-year comparison.
The three companies have sold in the last 12 months between $40 million and $110 million, less than some of the early 2017 enthusiastic forecasts that Lemonade (for example) would hit $90 million of premiums by the end of 2017. In auto, I pointed out at my IoT Insurance Observatory plenary sessions that the pay-as-you-drive telematics approach seems to attract only the niche of customers that rarely use cars – maybe a growing niche, but not a billion-dollar business (in premium at least).
Loss ratios have all been below 100%, which is a great improvement from the 2017 performances. The quarterly dynamics show a positive trend, but these loss ratio levels are far from the U.S. market average for home insurance (Lemonade) and auto insurance (Root and Metromile).
While loss ratio is a fundamental insurance number – claims divided by premiums — I’ve been asked how to normalize/adjust the loss ratio of a fast-growing insurtech company.
Imagine a fast-growing insurer with the following annual figures:
Premiums written: $10 million
Premiums earned: $6 million
Claims paid: $2 million
Losses incurred but unpaid: $5 million
Any of the following numbers might be called a “loss ratio”:
Claims paid divided by premiums written: 20%
Claims paid divided by premiums earned: 33%
Claims paid and losses incurred divided by premiums written: 70%
Claims paid and losses incurred divided by premiums earned: 117%
The least attractive is the right one.
Claims paid and losses incurred divided by premiums earned is the loss ratio, for a fast-growing start-up as for a large incumbent.
The others are only “exotic loss ratios.”
I’ve heard people say that accounting rules cause the loss ratio to be overstated based on the following unlucky scenario:
There is a book of business done by a one-year homeowner’s policy sold for $730. This policy will earn $2 of premium each day. If a $100 claim (net of the deductible) is received on that first day, the loss ratio is 5,000%. That’s how it works, but is it overstated? Well, as long as premium is being earned, more claims could arrive and the loss ratio could go even higher still.
Obviously, you expect the following 364 days to be less unlucky for this portfolio. But I don’t think there is any need to adjust that loss ratio…only to know that is not (statistically) relevant.
The analyzed start-ups have portfolios of more than 100,000 policies, so the bad luck can’t be accountable for eventual unfavorable loss ratios.
It could be that some approaches are specially targeted for fraud, and it only takes a few fraudsters to cause big problems in the loss ratio on a small book, as the above illustration shows. Some start-ups have advertised how quickly they pay claims, sometimes not even having a human review them, which invites unsavory people to pay a small amount to start a policy then “lose” a valuable item. Early on, when less premium has been earned, this fraud has a particularly great impact on the loss ratio. Over time, in a bigger and more balanced book, fraud gets tempered by the law of large numbers.
Additionally, some start-ups have offered large new-business discounts. If they can retain customers, reducing the premium leakage, their second-year loss ratios should be more reasonable, but the overall loss ratio will be elevated for however long they are acquiring customers with aggressive discounts.
I would love to discuss also the other lines of the income statements, but unfortunately they are not meaningful or comparable any more, because companies now move expenses among their entities not represented in the yellow books. The cost amounts represented in the yellow books are only a part of the real costs necessary to run the insurance business.
The statutory information I’m commenting on is reported only for insurance companies, not agencies, brokers or service companies. The term “insurance company” or “insurer” has a very specific meaning: “the person who undertakes to indemnify another by insurance.” Within an insurance holding company, it is typical to have an insurer and an affiliated agency, and sometimes other affiliates such as claims administrator. The insurer pays the agency to produce policies. This may feel like moving money from one pocket to another, but there would be reasons for it — which I won’t get into now.
The point for commentators and investors is to beware of this: If an insurer (for whom the public receives financial data) pays an affiliate 25% of its premiums to provide certain services, then the insurer’s expenses (which are reported in the yellow book) are set at (or close to) 25% of their premiums, regardless of what they actually are.
Major investors are typically privy to large amounts of information and can disentangle the back-and-forth between the insurer, agency and holding company. For smaller investors, or those who simply pick up a statutory filing, it is easy to be misled.
At the beginning of 2018, Lemonade was no longer consolidating its parent and affiliate expenses into Lemonade Insurance Co., its statutory entity. Lemonade’s CEO commented that this change was at the request of its home state regulator. Root followed suit in October 2018, so the 4Q18 expense ratio is moved to 28% from the 70% in the 3Q18. So the “new” expense ratios (and therefore combined ratios) are artificial and not comparable with the previous ones or with competitors’.
While regulators may have reasons for their actions, it is better for students of insurance innovation to know the full, real financials, so as to determine if the start-ups are ever able to “walk the talk” of better expense efficiency from “being built on a digital substrate.” Unfortunately, bloggers are not the main audience of statutory filings. Nonetheless, innovation cheerleaders, investors and journalists …please pay attention to these accounting differences before commenting the performances.
Since the beginning of the quarterly discussion of U.S.-based insurtech carriers’ financials based on their public filings, many have responded that these players needed to be evaluated on other metrics, too. I agree, so let’s look at one of those measures and talk about some questions to determine whether the measure really stacks up.
Misleading vanity metrics
The insurance value chain is complex and difficult to compare across models. This can lead to comparisons between very different companies. Take these two hypothetical companies:
Company A — flashy Start-up Insurance Co. uses outsourced call centers, bots, incessant Instagram ads, comparison rater websites, third-party claims administrators and a slick app. It sells one line of insurance, only personal, with low limits, and has no complicated old claims (yet). If your house burns down, you open an app and wait. The company has a low expense ratio, high acquisition costs and a high loss ratio.
Company B — old Traditional Insurance Co. uses a mix of direct sales, captive agents and independent agents. Claims are handled mostly by agents and in-house staff. If your house burns down, your agent turns up with a reservation for a nearby hotel, billed directly to the insurer. The company sells 12 lines of insurance, including small commercial, and a wide range of products within each line, with bundling encouraged. The company has a high expense ratio, high acquisition costs, strong customer loyalty and losses less than the industry average.
A “vanity measure” could easily make one of these companies look better than the other.
The start-up, for example, may claim performance several times better than the incumbent on a “policy per human” KPI, considering in the count of “humans” agents and brokers. Why does a policy-per-human number matter at all? And why is more policies per human better than fewer?
Company A and Company B are two different business models, with two opposite approaches about humans — neither of which is necessarily better. Steve Anderson and I wrote a heartfelt defense of the model based on agents, brokers and other distribution partners a few months ago.
To measure efficiency, I prefer to use the two traditional components of the expense ratio:
General operating expense ratio = general operating expenses ÷ earned premiums
Acquisition ratio = total acquisition expenses divided by the earned premiums (for high growth companies, it’s acceptable to do the division by written premium). This metric includes advertising, other marketing expenses, commissions and other distribution expenses. However, this number (like CAC) can be difficult to compare – for example, are fixed marketing expenses included or excluded? And the economics of customer loyalty are different between direct (where initial CAC is high but renewal is low) and agent sales (where initial CAC is lower and variable but renewal commissions are significant).
I love numbers and – as shared in an interview with Carrier Management – the absence of quantitative elements in self-promoting website articles, conference keynotes, whitepapers and social media exchanges have been one of the reasons for starting the publication of articles about the full stack U.S. insurtech start-ups.
Although I’m sometimes described as “critic” or “cynic” about insurtech companies, I’m only critical of the misuse of numbers and am a big fan of those who get the old school insurance KPI right. I’d love to see innovation succeed in the insurance sector, and I wish all the best to these three players and their investors.
Our three U.S. venture-backed nsurtech carriers have mo’ money, mo premiums’ and mo’ losses. Around the time their losses became notorious, there were some signs of improvement, but not enough to prove viable business models. It’s still early. No 2Pacalypse is imminent, as the startups have tons of cash. But forget about East Coast/West Coast feuds: It’s still insurtechs against the world.
Premiums and losses grow at venture-backed startups
Industry-backed startups stay focused on underwriting
Gross or net – which matters?
Correcting rookie mistakes, experimenting and still underpricing?
1. Premium and losses grow at venture-backed startups
As we have said before, it’s early days. Growth is rapid. But all three startup carriers are a long way from profitability and need to continue to raise prices substantially (and thus potentially churn customers), tighten underwriting (which could also materially affect growth) or improve operational aspects that drive losses such as claims performance.
In insurance, product-market fit is only demonstrated when selling a product that makes money.
Here is the summary of the 3Q18 details of the three venture-backed U.S. insurers that we’ve been tracking. The fourth venture-backed insurer, Next Insurance, did not write premium in 3Q18.
On expenses, it is important to note that Lemonade received permission from New York State to hide some expenses in an affiliated entity effective at the start of this year, so Lemonade’s expenses aren’t comparable to the others. (Lemonade’s CEO told us that the State of New York requested the change.) Root has followed suit and will now start to hide expenses in an affiliated agency, effective Oct. 1, 2018.
Unfortunately, with less transparency, it is harder to verify whether the overall business model actually adds up. Lemonade, for example, makes a big deal about the power of an insurer built on a “digital substrate.” We’re inclined to agree, and we hope they’re right, but we won’t know from their statutory results whether a better expense ratio is really possible.
The third quarter’s “most improved” award goes to Lemonade, which grew premium by 57% over the prior quarter to $15.5 milliion (perhaps helped by seasonal effects in the rental market) AND turned in a 96% gross & LAE loss ratio, its best ever. The company’s reserves developed slightly adversely in the quarter, with YTD unfavorable development standing at $254,000, modestly worse from $245,000 the prior quarter.
Lemonade has a ways to go — its reinsurers continue to subsidize the loss ratio (but now paying “only” $2.44 of losses per dollar of premium), the average net loss & LAE ratio is still well above the 40% at which charities receive a giveback, and the company is far from profitability. But re-underwriting and higher rates — which we discussed in our last post — may be paying off. There are other explanations, too — shifting to cat-exposed risk, for example, lowers the loss ratio but raises the cost of capital.
The company has written $33 million in the first nine months of 2018, which is well behind some exponential expectations published last year. Management also says that they will take out 60 loss ratio points: “A similar progression in the year ahead will get us to where we need to be.” Good luck. (Despite the good quarter.) Taking out loss ratio points gets harder for every point you get closer to the industry average. The breeziness of management’s comments makes us wonder how much focus will be placed on the hard spadework of loss ratio improvement vs. embarking on a glamorous European Grand Tour.
Respected industry analyst V.J. Dowling, whose IBNR Weekly publication (subscription only) was cited in a Lemonade transparency blog supposedly praising Lemonade, recently published a tear-down of Lemonade’s strategy. IBNR called the giveback “a joke,” the company’s marketing “overly simplistic, constantly changing and borderline dishonest” and “B.S. On Lemonade’s ‘Transparency’ & Model.”
We have long taken issue with Lemonade’s definition of “transparency,” which (as with many startups) is transparent only insofar as the company wishes to tell a story, which is just another form of marketing. (We are working on a longer article about ways startups bend numbers and welcome your ideas). We also agree with IBNR that a giveback at 1.6% of premium seems cynical. Middle-class insurance agents are some of the biggest charitable donors and sponsors in many towns, and it’s hard to see the social benefit in Lemonade’s spending the advertising budget enriching amoral tech bros at GAFA rather than sponsoring the town’s Little League uniforms as an agent might do. Even insurance veterans don’t seem to realize that all of Lemonade’s entities are for-profit. Don’t be fooled: B-Corps are still for-profit corporations. And railing against gun violence is easy, but social consciousness is messy in reality.
Metromile grew at a steady and respectable 21% from the prior quarter to $24.5 million of gross premium but ran a 98% gross loss & LAE ratio (not improved from 2017). Among the three players, it has been the ant: slower growth and rather steady results. After seven years in business, the company is barely generating an underwriting profit and is showing little improvement. The company has been taking rate, but not as much as the actuaries say is needed. The company’s latest California indication, for example, is rates up 38%, but the company only took 15 points of rate, effective July 1, 2018, which will work through its book over the next couple of months.
Matteo recently published a discussion of why its PAYD (the pay-as-you-drive) approach confirms its niche nature.
If Metromile is an ant, Root is a grasshopper. Root again grew extremely rapidly through its TBYB (Try-Before-You-Buy) mobile approach, as one probably would expect of a company with a $1 billion valuation (as we discussed in our previous article). Root doubled the volumes underwritten by Lemonade. But Root continues to struggle with rookie mistakes that drove losses — discussed more below. Root grew premium by 120% vs. the prior quarter to $33 million for the third quarter but turned in 128% gross loss & LAE ratio – its worst of the year and the worst of the three U.S. insurtech carriers.
With the three carriers resuming rapid growth after a slow 2Q, the 2Q slowdown that we observed in our prior article may have been seasonal or a one-time coincidental slowdown.
2. Industry-backed startups stay focused on underwriting
We first pointed out in the prior quarter’s analysis that companies run/led by well-known underwriters were growing slowly but were far more profitable than those with venture capital backing. That hasn’t changed.
Here are the four subsidiaries of big companies that are selling direct or have a claim on being an insurtech. These companies often depend on parents for reinsurance and infrastructure, so we show mainly the gross figures.
3. Gross or net – which matters?
One well-regarded venture capitalist challenged us on why gross premiums and loss ratio matter when net is what actually sticks to current investors. The difference between gross and net is reinsured premium and loss.
The gross loss ratio is provides an unbiased view of the profitability of the book of business, while the net loss ratio may benefit from savvy reinsurance buying, as in the case of Lemonade.
What matters most is for startups probably the higher of the two loss ratios. If reinsurance reduces the loss ratio, this is only sustainable over time if reinsurers make an adequate return. Reinsurers, like all businesses, tend to demand higher prices when customers lose them money, which in time makes the net loss ratio look more like the gross.
In the case of the two auto startups, they are paying reinsurers to take away their biggest losses, so this “cost of reinsurance” needs to be reflected, and the net numbers matter more.
4. Correcting rookie mistakes, experimenting and still underpricing?
At the Nov. 30 plenary session of the U.S. IoT Insurance Observatory, Denese Ross of DRC Consulting shared an in-depth 50-page review of Root’s filings, which total over 130,000 pages. Our opinions, based on her factual/non-opinion analysis, is that Root is experimenting aggressively, correcting some rookie mistakes and still underpricing some business.
The following is Matteo & Adrian’s interpretation and analysis:
Raising rates, but is it enough? Root has begun correcting its early underpricing by raising rates substantially in several states, but not as much as the actuaries indicate is needed. The company has filed rate increases in seven of its 20 states since late summer, ranging from +5.5% to +34%. The highest is in Texas, where the filing is pending, and where Root now writes 1/3 of its premium (whereas Ohio was the largest state last year). The indication in Texas was +200%. An indication is the change to the overall rate “indicated” by an actuarial analysis to achieve a specified pricing target based on analysis and adjustments to historic trends. For many reasons, the rate level chosen could be less than the indication, but a big gap between rate taken and indication suggests management may still be oriented toward growth even if it is at a loss. Indeed, Root recently put up a blog post claiming to be up to 52% cheaper. Further, Root also added or changed discounts, which are not part of the headline rate. In other states, per-policy rate caps limit the actual amount of rate taken on renewal business, possibly to avoid churning existing customers, since the acquisition cost of a renewal customer acquired via a direct channel is quite small. These rate-making decisions illustrate the fine line a company like Root has to walk when balancing growth, profitability, retention and acquisition cost. Similarly, investors should be wary of individual numbers presented in isolation — it can be good to churn underpriced customers.
Technology that isn’t as predictive as expected: Root collects telematic data via a smartphone and scores a driver once he or she has tracked 500 miles, using features such as hard braking, acceleration, turning, time of day, mileage, consistency and distractions. The company started with a third party’s telematics model that apparently gave excessive discounts to drivers in better tiers – see the steep blue curve below, which is from a Root filing. Root then developed its own model and is giving more modest discounts to better drivers (gray line). Telematics contributes information and allows more granular clustering, but the law of large numbers and class rating variables (e.g. demographics) still matters more than perhaps anyone would like. Good drivers get hit by uninsured bad drivers. Bad drivers can drive well for 500 miles when they’re being watched. Good drivers’ cars get stuck in hail storms and hurricanes. Nonetheless, Root is the first insurer around the world to acquire large numbers of customers through a TBYB (Try-Before-You-Buy) app, something many incumbents have tried in the past few years.
Me-too is harder than it looks. Root has filed several modifications to its rates to correct mistakes, such as correcting household structure data, moving from ISO vehicle symbols to factors assigned directly by VIN and correcting for “double-discounting” and “double-surcharging.”
Claims difficulties. Most startups lack a high-quality claims infrastructure, which is one of the hardest aspects of an insurer to build, but it appears that Root is now taking more claims activity in-house instead of using a third party claims administrator (TPA). To date, Root’s customer claims experience hasn’t been as easy as the company says. The 15 Better Business Bureau complaints are mostly claims-related, and only 1/3 of Clearsurance users were satisfied with the claims experience. On the same website, Progressive and Geico are around 80%, and Metromile is at 70% claims satisfaction. As one Clearsurance review of Root reads: “I was in an accident back in July, and it’s now November, and my simple simple claim has not been paid on or closed out. It was originally with an adjuster at Crawford & Co. but without my knowledge that changed hands to some other adjuster actually within Root. The original adjuster back in September said that she would get back to me within the week….Months later and I’m still here with an open claim.”
It will take time for the corrections Root has made to show through in results. And particularly on pricing, they may be too little. Thus, the question remains: Is Root’s growth simply the result of selling something far too cheaply, or is it quietly experimenting its way to becoming the next Progressive?
5. Can a startup compete with Progressive and Geico?
The most common refrain of startups is that insurance is broken, and their solution will fix it. Oui, c’est vrai. But two companies have their act together more than almost any other insurers: Geico and Progressive, the #2 and #3 auto insurers in the U.S. Geico is famous for its lean expenses, while Progressive runs a loss ratio about six points below the industry average of 69%.
It wasn’t always this way. Back in 1996, Progressive was #7, and Geico was #9. There are few other examples in insurance of companies so steadily and regularly gaining market share across decades. In our first article, we showed how startups historically have only won where incumbents leave the door open. Progressive and Geico are not leaving the door open, at least in personal auto as a stand-alone line. Progressive, in particular, has been quite experimental throughout its life, pioneering telematics in the U.S. and experimenting with various ways of paying for auto insurance and using telematics, as some startups are now doing, but with none of Progressive’s advantages.
So why do startups try to compete with such well-run companies? We think there are several beliefs (or hopes):
A belief that the startup can duplicate public filings to set rates. As noted above, it’s harder than it looks. Even professional comparison rating websites have trouble estimating a person’s actual premium, and, in a thin-margin business, small differences matter. This is not a new story – eSurance’s first me-too nearly 20 years ago didn’t work particularly well. Progressive, having outperformed the market for decades, knows that its filings are scrutinized. Like a game of cat and mouse, the company has become expert at fooling competitors that try to replicate filings, which run thousands of pages and are deliberately obscured and complicated. And rate filings are only one aspect of underwriting. Credit models, underwriting rules and marketing plans for attracting the best customers may not be public and may be difficult to replicate. The result is that unwary and inexperienced competitors may grow by unknowingly writing the risks that their competitors didn’t want. There is no CAC at which an underpriced customer is a desirable customer
A belief that there are profitable untapped segments that incumbents won’t cannibalize. Progressive and Geico are very intelligent companies with a long history of innovation and extremely detailed customer segmentations. It defies all logic to suggest that there is a large segment of customers in a fragmented and highly competitive market that are being greatly overcharged simply because these companies refuse to offer them good rates to avoid cannibalization.
A belief that the startups can easily acquire customers through direct channels. Direct distribution is not new in U.S. auto – Geico and Progressive both spend billions of dollars a year in advertising and have extremely sophisticated digital marketing. To this end, one of Root’s interesting innovations is its referral program, which we showed in our last article to be quite powerful. (And we probably underestimated its power, because we didn’t account for multi-car policies.)
A belief that the startups can be “good enough” at claims. Scale matters in handling claims, whether it is hiring and managing high-quality defense counsel, detecting fraud patterns or negotiating with vendors such as garages and roadside assistance services. Further, incentives for third party administrators must be carefully set to avoid excessive claims cost and customer dissatisfaction. For example, when a TPA is paid per open claim, speed of settlement may suffer, which can lead to regulatory fines or bad faith judgments – not to mention customer annoyance. In an industry where 96% is considered a best-in-class combined ratio, even a little bit of claims leakage can be quite hurtful. Metromile is exploring the reinvention of the claims process through the usage of telematics data. Considering the international best practices of doing this, the size of the portfolio might not allow the company to have enough claims yet to train an algorithm for a reliable crash kinematic reconstruction.
A belief that “we only need 1% of a $220 billion market to be huge.” So 1999…
This analysis isn’t to suggest that it’s impossible for a startup to win in auto insurance, but that the moat around Geico and Progressive in personal auto is wide and deep.
The annual figures that are published in March contain a wealth of additional information. To be notified when these numbers are available, please follow the authors on LinkedIn.
Usage-based pricing is a fascinating topic for insurers. A technology that allows persistent monitoring of risk exposure during the coverage period could potentially enable insurers to price each risk at the best rate.
The potential, however, is not the reality.
In 2017, 14 million policies sent telematics data to insurers around the world, of which 4.4 million were in the U.S market, based on an estimate by the IoT Insurance Observatory, an insurance think tank that has aggregated almost 50 insurers, reinsurers and tech players between North America and Europe. (In the U.S., there were a further 3.6 million policies that are still active and commonly defined as telematics but that in the past had a dongle only and didn’t send any data to insurers last year.)
However, less than 9% of the global insurance telematics policies were characterized by usage-based pricing, which is a mechanism that charges the policyholders for the current period of coverage based on how they behave (mileage or driving behavior) during this period.
Instead, the vast majority of the telematics policies bought by customers around the world today have a defined up-front price for the current policy term. Moreover, the telematics data registered during the policy period does not affect this price in any way, and is used only for proposing a renewal price at the end of the policy. So, these policies are not usage-based because at the beginning of each policy term the customers are sure about the amount they are going to pay for the policy, regardless of their behavior during the months of coverage.
These existing implementations of telematics-based pricing are somewhat validated from consumer perceptions toward insurance. In a survey of 1,046 U.S. consumers, the Casualty Actuarial Society Insurance On-Demand Working Party has addressed and demystified some of the behavioral economics assumptions on the insurance products. The research showed that only 32% of consumers reviewed their personal lines (auto and home) coverage more than once per year. Furthermore, 89% of consumers said they would rather pay a single, stable price per year compared with paying per usage without a certainty of total price. Usage-based auto insurance, across the entire on-demand category studied by the working group, is attractive to people penalized by traditional insurance products, that is, consumers with low usage who would otherwise have to pay for more coverage than they need.
Potential and Success Stories
The usage-based approach persistently monitors the policyholders and charges (potentially) each customer a rate commensurate with actual exposure, minimizing the premium leakage in each coverage period. The resulting minimized earning volatility from usage-based pricing allows insurers to increase the leverage and through this to improve investment return and the return on equity of the company. This approach also allows for increased retention of good risks, at any pricing level, which are penalized by competitors with less accurate pricing mechanisms. The quality of the portfolio is improved (with more profitable customers) at each renewal.
The resulting lower volatility from usage-based pricing and better quality of the portfolio over time would also enable insurers to negotiate lower reinsurance costs.
But while usage-based insurance could theoretically be a profitable option for insurers, the problem seems to be the lack of customer demand for an insurance product where there isn’t a defined up-front price for all the entire coverage period..
Newcomers to the insurance market are bringing a different perspective to the problem, recognizing that small clusters of drivers who have been heavily penalized by the current insurance rates—such as extremely low-mileage drivers, or extremely safe drivers without a credit score—could be enough to start a niche business. There are a few success stories of insurtech startups, such as Insure The Box and Metromile, which have been able to build portfolios around 100,000 policies and relevant company evaluations within six to seven years.
Driving Scores at the Underwriting Stage
One way to combat the lack of market fit that has affected the usage-based adoption could be to use a driving score at the underwriting stage. This way, insurers will make an up-front quotation by using—together with traditional data—the driving data.
The value created through this approach is clear and similar to experiences the sector has had integrating new risk factors (e.g. credit scoring) in pre-existing risk models. This telematics-enhanced risk model enables more accurate pricing. This, in turn, allows insurers to generate favorable selection by attracting the best risks for each pricing level (leaving the worst to the competitors). Through the creation of smaller and more homogeneous clusters of clients, this approach even reduces premium leakage, reducing the volatility. And, if the driving score is used at each renewal, there is a chance of improving portfolio quality over time (at any pricing level), with insurers using driving scores for underwriting, benefiting from retention of the most profitable customers–those who are penalized by competitors with less accurate pricing mechanisms.
The ROI of this approach is extremely positive, but the current scenario for obtaining the customer driving score seems very different from the scenario we have known for the credit score. The credit score (or the granular data necessary to calculate it) is available on the entire customer base and certified by reliable third parties, so each insurer can gather this data any time a customer requests a quotation via an agent, a broker, a call center or even online. Moreover, anyone who doesn’t have a credit score is considered a nonstandard risk. So, the concretization of the driving score dream requires the availability and reliability of third-party data for the insurers and, most importantly, the creation of frictionless purchasing processes for the clients.
Data exchanges, which bring OEM data to insurers, have been present in the U.S. customer market for a few years, but because there are many points of friction throughout OEM funnels, they still represent only 2% of the U.S. telematics insurance portfolio. This customer fatigue is due to the need to opt in to request a quotation. Eligibility for the opt-in comes in a moment when he is not shopping around for insurance coverage (a few months after the purchase of the new car). The quotations, which are done with anonymized data, are only indicative, so the customer needs to add data later to receive the real proposal.
Try Before You Buy
A different way to concretize the wish to access a driving score any time an insurance price quotation is calculated is by using a try-before-you-buy app. Given the current level of smartphone penetration, such an app likely provides an easier way to address a large part of the market than with the data exchanges and may also reduce customer frictions. As insurtech carrier Root is currently doing, an insurer can ask a prospect to download an app on his smartphone, calculate the driving score through collected data and, after a while, calculate the quotation incorporating the customer’s driving score. Using this approach, this less-than-two-year-old auto carrier startup wrote 1.5 times more premium than the more-talked-about carrier Lemonade. (Both are insurtech carriers, although Lemonade is writing renters insurance, and Root is writing auto). Root even entered in the insurtech unicorn club in August, thanks to a $100 million round of funding raising the valuation to $1 billion.
Tailored renewal price
As mentioned, 90% of the current global telematics policies only use the driving data for tailoring the renewal price to the customers after having monitored them for a few months (rollover approach) or for the entire coverage period (leave-in approach).
Are insurers achieving any economic value through this pricing approach?
They can increase the retention of the most profitable risks at each pricing level by providing a discount at renewal. However, this additional discount reduces the profitability of these policyholders. So the chance to create some value through this “discounted retention” is linked to the presence of a high-level churn rate. If surcharges to the worst risks at each pricing level are added, insurers will have the opportunity at renewal to partially reduce the premium leakage they have identified on these risks, or push some of them toward competitors.
The accompanying chart (right side) summarizes these pricing thoughts: The expected ROI of the “discount at renewal” is definitely lower than the driving score scenario—it structurally misses the ability to have a positive up-front selection by attracting the better risks at each pricing level—but it is positive if surcharges are added.
The IoT Insurance Observatory has found that a large portion of the policies using driving data for tailoring renewal prices have not resulted in any bad driver penalties.
So, are these telematics portfolios destroying value instead of creating it?
The reality is that there is value created on these portfolios, but the value is not tied to pricing. And some of the pricing approaches are even reducing that value.
First, there are many examples of the risk self-selection impact of all the telematics-based products around the world. Even if two customers seem to be equal based on their characteristics, the one who accepts the telematics product has a lower probability of generating a loss. The stronger the monitoring message on the product storytelling, the higher the self-selection effect. The most statistically robust study is on the Italian auto insurance market, where this risk self-selection effect has accounted for 20% of the claim frequency. In this market, telematics products currently represent more than one-fifth of the personal lines auto insurance business, and the storytelling of the product is hugely focused on monitoring and customer support at the moment of a crash.
Other than risk self-selection, three other telematics-based use cases have been exploited by insurers.
Some international insurers have reinvented their claims processes through telematics data: Their new paradigm is fact-based, digital and real-time. Insurers such as UnipolSai have introduced tools for their claim handlers that allow a quicker and more precise crash responsibility identification and have been providing precious insights to support the activity of all the actors involved in the claim supply chain (both loss adjusters and doctors).
A second well-demonstrated telematics use case is the change of driver behavior. VitalityDrive introduced by the South African insurance company Discovery Insure is the first insurance telematics product entirely focused on promoting safer behavior. All the product features—from gas cash-back (up to 50% of fuel spending per month) to active rewards through the app (including coffee, smoothies and car wash vouchers)—are contributing to the risk reduction of the book of business and to increased retention of the best risks.
Both the Italian and South African experiences have even been characterized by the insurers’ ability of enhancing the insurance value proposition by adding telematics-based services bundled to the auto insurance coverage. The fees paid by customers for these services almost offset all the costs of the telematics services on the insurers’ income statements
Based on the experience of the IoT Insurance Observatory, global insurance telematics best practices have generated more value through these four use cases than through pricing as of today. So, the sum of the self-selection effect, the claim cost reduction and the economic impact of changes of behavior allows an insurer to provide an important up-front discount at the same level for all the new telematics-based policyholders.
This relevant level of up-front discount — 20% or more — has been able to drive the adoption (overcoming any eventual customer privacy skepticism) because it fits with the customer desire to save money, contrasting the low adoption rates generated for more than a decade in the U.S. where up-front discount offers are typically only 5%.
The discount should be maintained, on average, at the same level at the renewal stage. Moreover, an additional economic value can be generated—at each pricing level—by providing additional discounts to the best policyholders and reducing the discount to the worst ones.
This is what the international best practices are doing today.