Case Study 1: The Listed InsurTech Carriers and the Loss Ratio That Wouldn't Bend

A note on figures. The companies discussed here are real, public, and listed on U.S. exchanges, so their broad story is a matter of public record (IPO prospectuses, quarterly earnings, investor letters). Specific quarterly loss ratios, combined ratios, and stock prices are public but change every reporting period, so this study describes the pattern and the direction — which are well established and stable — rather than pinning a precise figure to a precise quarter. Where you want exact numbers, go to the companies' SEC filings. Nothing here invents a statistic.

Background

In the second half of the 2010s and into the early 2020s, a group of technology-first insurance companies came to market with an unusually clear and confident thesis. The legacy insurance industry, they argued, was slow, paper-bound, expensive to run, and miserable to buy from. A company built from scratch on software — a slick app, instant quotes, automated claims, a low overhead unburdened by legacy systems and a branch network — could win customers, especially younger ones, and could underwrite better using data and machine learning. Several of these companies grew fast on that thesis and then went public, which is what makes them such valuable teaching material: once a company is listed, it must report its real numbers every quarter, and the numbers tell a story the pitch deck could not.

The most-discussed of these companies attacked the high-volume, data-rich personal lines first — renters insurance, homeowners, and personal auto — exactly the lines where the law of large numbers has the most to work with and where a digital experience is easiest to deliver. Their marketing emphasized speed and delight: a renters policy bought in minutes from a phone, a simple claim paid almost instantly, a brand that felt nothing like the insurance company of a parent's generation. Some leaned into a social or peer-to-peer-flavored framing — pooling premiums, taking a fixed fee, and giving back what was left of a designated pool — to position themselves as a different kind of insurer. Investors, in an era of cheap capital and enthusiasm for anything labeled "disruption," rewarded the growth with rich valuations.

The insurance / underwriting issue

The issue is the one this entire book is built around, and these companies ran into it in the most public way possible: a low expense ratio is not a low combined ratio, because the loss ratio is the larger and harder term — and only underwriting moves the loss ratio.

Reconstruct the thesis as an underwriter would stress-test it. These companies genuinely did achieve something real on the expense side. Direct digital distribution and heavy automation can lower the cost of acquiring and servicing a policy, and to a degree they did. But recall the structure of the premium dollar from Chapter 3: it pays losses, expenses, and a margin. For most personal lines, losses — what the company pays out in claims — are by far the biggest slice. You can attack the expense slice with software all day; if your loss slice is too big because you selected risks poorly or priced them too low, you still lose money on underwriting. And the early evidence, visible the moment these companies began reporting, was that their loss ratios ran well above the level sustainable underwriting requires, for an extended period, producing underwriting losses quarter after quarter even as premium grew impressively.

Why were the loss ratios high? Several reasons, all of them familiar from earlier chapters, and all of them things the technology did not fix:

• Growth itself attracts adverse selection (Chapter 1). A company growing fast by being cheap and easy to buy from disproportionately attracts the eager buyers, who skew toward the worse risks. Some of the volume these companies celebrated was volume the rest of the market had priced more cautiously for a reason.
• New companies have thin loss data on their own book. A model is only as good as the experience it learns from, and a young carrier's own loss history is short and unstable (the credibility problem of Chapter 10). The confidence that data and machine learning would out-select the incumbents collided with the fact that the incumbents had decades of loss experience the newcomers did not.
• The hard, slow truth of insurance is the lag. You write a policy today and learn whether it was a good one over the next several years as claims emerge and develop. Early growth flatters the picture because the losses have not matured yet — the same timing trap the chapter described for MGA books (§34.2), here at the scale of a whole carrier.
• Catastrophe and severity inflation hit the property and auto lines hard in this period, and no app changes the weather or a jury verdict. A company concentrated in catastrophe-exposed homeowners, or in personal auto during a stretch of rising claim severity, was exposed to exactly the volatility that underwriting and reinsurance exist to manage — and that a low expense ratio does nothing about.

The underwriter's reading. None of this means the companies were foolish or their technology worthless. It means they optimized the wrong variable relative to the constraint. The binding constraint in personal lines is the loss ratio, and the loss ratio is moved by selection and pricing discipline — the craft of Chapters 7 through 13 — not by the speed of the quote. The companies treated underwriting as a solved engineering problem and distribution as the frontier; the math said it was the other way around.

What it shows

It shows the fifth theme of this book — technology augments underwriters; it does not replace them — demonstrated at venture scale and confirmed by audited financial statements. The customer-experience innovations were genuine and have been widely copied; that part of the thesis was right. The underwriting thesis — that data and automation could deliver a structurally better loss ratio than the incumbent industry, quickly, while growing fast — was not borne out by the results. The companies that internalized this adjusted: they raised prices (often sharply), tightened their underwriting, exited or de-emphasized the worst-performing lines and geographies, leaned on reinsurance to cede volatility, and in some cases pivoted toward licensing their technology or operating in a more capital-light way. Those are not the moves of a company that has repealed the loss ratio; they are the moves of a company that has met it and is now doing real underwriting.

It also shows something subtler and worth holding onto: the market's enthusiasm and a company's customer love are not underwriting signals. For a stretch, these companies were simultaneously beloved by customers, admired in the press, richly valued by investors, and losing money on their core business. An underwriter who confuses any of those external signals with the health of the book will be misled. The combined ratio is the scoreboard. Everything else is commentary.

Outcome

The outcomes have varied, and the story is still being written, but the broad arc is consistent and public. The rich valuations of the cheap-capital era compressed substantially as interest rates rose and as investors shifted from rewarding growth to demanding a path to profitability — a shift these companies' underwriting results made unavoidable. The companies responded with the underwriting moves above: repricing, tighter selection, exiting unprofitable segments, more reinsurance, and a generally more sober posture toward the loss ratio. Some have made real progress toward underwriting profitability by doing so; some continue to work toward it; the sector as a whole is a more chastened and more realistic version of its 2020 self. The permanent legacy is on the experience side: the bar for digital intake, instant quoting, and fast claims has been raised across the whole industry, including at the incumbents the InsurTechs set out to disrupt.

The thing that did not happen is the thing the original thesis promised: the underwriter was not automated away, and technology did not deliver a structural loss-ratio advantage that made traditional underwriting obsolete. If anything, the episode underlined how irreducibly hard underwriting is — hard enough that some of the best-funded, most technically capable new entrants in a generation found it the binding constraint on the whole business.

Lesson

The lesson is the chapter's master lesson, and it is worth memorizing in the form an underwriting committee would recognize: you can buy growth in insurance with a click; you can only earn a profit with underwriting. A low expense ratio is good and worth pursuing, but it sits on top of a loss ratio that only selection and adequate pricing can move, and for most lines the loss ratio is the larger term. Any business — or any underwriter — that mistakes the easy variable (cost, speed, growth) for the hard one (the loss ratio) is buying market share with future losses, and the combined ratio will present the bill on its own schedule. Technology is a powerful augmentation of underwriting and a permanent improvement to the customer experience. It is not a substitute for the judgment and discipline at the heart of the trade — which is, in the end, the single most important thing this entire book is trying to teach you, here confirmed by the public, audited experience of the people who bet otherwise.

Discussion questions

1. The InsurTech carriers achieved genuine improvements in the expense ratio. Using the structure of the premium dollar (Chapter 3), explain why that was the wrong variable to optimize relative to the binding constraint, and identify the right one.
2. "Growth is easy in insurance." Explain, using adverse selection (Chapter 1), why fast growth by being cheap and easy to buy from is especially dangerous — and why a year of strong premium growth can be a warning sign rather than a triumph.
3. These companies were beloved by customers, admired by the press, and losing money on underwriting at the same time. What does that teach an underwriter about which signals to trust when judging the health of a book?
4. Several of the companies responded by repricing, tightening selection, exiting segments, and buying more reinsurance. Map each of those moves to a chapter of this book, and explain why, taken together, they amount to "doing real underwriting."
5. The episode raised the digital-experience bar across the whole industry, including at the incumbents. In what sense did the InsurTech wave "succeed" even where the InsurTech companies themselves struggled — and why is dismissing the whole wave as hype the mirror-image mistake of the original over-optimism?
6. A young carrier's own loss data is short and unstable. Connect this to the credibility problem of Chapter 10, and explain why it undercut the confidence that machine learning would let these newcomers out-select the incumbents from day one.