US property and casualty insurers are coming off their strongest underwriting performance in two decades. Soteris is betting that the next source of profit will come from finding the policies dragging those results down.
The Y Combinator-backed machine learning company has launched a new product designed to help P&C insurers and managing general agents identify unprofitable policies already sitting within their books,without changing rates, policy forms, regulatory filings or staffing levels.
The launch follows a strong stretch for the US P&C sector. The industry posted a $60.9 billion net underwriting gain in 2025, nearly tripling the $22.1 billion recorded the year before, according to AM Best's latest financial review, which found the industry's combined ratio improved 3.7 points to 92.9.
Separately, Verisk and the American Property Casualty Insurance Association put the 2025 underwriting gain at about $63 billion, with the combined ratio improving to 92.9% from 96.6% in 2024. Fitch Ratings, in an April report, called it the industry's strongest underwriting performance in two decades, with the combined ratio reaching its best level since 2006.
Soteris' original product, which analyzes loss ratios, has been used by carriers and MGAs since 2020. The company said that tool has scored more than 100 million policy submissions covering over $180 billion in premiums.
The new product addresses a gap the company identified between predicting a policy's loss ratio and predicting how much profit it actually generates for an insurer, since the financial outcomes of a single policy can be split across multiple entities. A company that sells and services a policy, one that holds the state license, and another that holds the capital backing it can each receive a different share of the economics involved, according to Soteris.
"Every insurer knows they're writing policies that will lose them money. They just can't find those policies with the resources currently at their disposal," said Sunit Shah, founder and chief executive of Soteris. "That's the blind spot we built Soteris to close. For the first time, an insurer can look at a single policy and know exactly what it's worth, in time to act on that information."
The company said insurers typically rely on spreadsheets and pivot tables to review policy performance, methods that can produce dozens or hundreds of usable analyses. Soteris said its approach can generate millions to billions of segment combinations by analyzing multiple policy characteristics simultaneously, treating each policy as its own credible segment.
Implementation takes under 90 days, according to Soteris, and once active, the system delivers insights in under 250 milliseconds through an application programming interface at any stage of a policy's life cycle, including at quote or binding.
The company said insurers using its original loss-ratio product saw improvements of five to 15 percentage points within a year of implementation. In proof-of-concept work tied to the new profit-focused product, Soteris said it observed increases in book EBITDA of between 70% and 125% among participating insurers.
Shah said gains identified through the tool could be reinvested by insurers into pricing and customer experience improvements.
Minsoo Chi, partner at Spider Capital, which led the company's seed round, said the new product builds on Soteris' earlier shift from analyzing segment averages to assessing expected loss at the individual policy level.
Soteris has raised more than $8 million in seed funding from investors including Spider Capital, Khosla Ventures, Intact Ventures, Amplify Partners, Foundation Capital and the Webb Investment Network. The company was part of Y Combinator's 2019 cohort.
Founder Shah previously built pricing models for life insurance before spending two years building a $750 million P&C insurer within hedge fund Pine River Capital, according to his Y Combinator profile.
