Building Plaid for Car and Home Insurance

Daniel Demetri learned to fly small planes over a decade ago and now does aerobatics in his spare time — footage of him flying upside down is searchable on YouTube. It’s a hobby he pursued specifically because it never stopped challenging a mind he describes as fast-moving and easily bored, and it required no small amount of persistence: earning his license in the New York area meant a Saturday routine of public transit into Manhattan, a 40-minute bus ride, then dodging highway traffic on foot to reach a Dunkin’ Donuts where his instructor would pick him up for the final mile and a half to the airfield — a full-day commitment just to fly for an hour.

That same persistence runs through his career. Self-taught in C at age nine, driven equally by a fascination with technology and finance (despite coming from a family of doctors — his parents and sister are all physicians), Daniel started trading with his bar mitzvah money and later found his way to Harvard, then Google’s Associate Product Manager program — a highly selective cohort of just 40 people globally, where he worked on part of Google’s ad exchange. After a stint at Kayak.com, he became the first product manager at Earnest, an online student loan company that originated billions of dollars in loans before being acquired by Navient. That’s where the idea for Trellis actually started to take shape.

About Daniel Demetri

Daniel Demetri is Founder and CEO of Trellis, a data aggregation platform for auto and home insurance, licensing technology that lets consumers securely share their insurance information the way Plaid enabled consumers to share bank data with fintech apps.

The Insight From a Student Loan Company

At Earnest, Daniel noticed something the insurance industry hadn’t connected to lending: the most common reasons borrowers requested forbearance were insurable events — health issues and car accidents. Most Americans carry a thin layer of equity beneath a stack of student loan, mortgage, and credit card debt, meaning a single uninsured or underinsured event can cascade through someone’s entire financial life. If a lender could verify that a borrower carried adequate insurance coverage — enough that a car accident wouldn’t cost them their job, or a disability wouldn’t cost them their mortgage — that borrower would genuinely present lower credit risk, and could reasonably be offered a better rate. Nobody was building that connective infrastructure, Daniel realized, not because the idea was novel but because the technical plumbing to link insurance data with lending and investment data simply didn’t exist yet.

Trellis Connect: The Plaid Analogy, Explained

Trellis Connect works, functionally, the way Plaid works for banking: a consumer authenticates with their existing insurance login, and Trellis securely retrieves their policy data — coverages, payment history, billing schedule — directly from the carrier. Daniel was careful to frame this as digitizing a process that has existed for hundreds of years, not inventing a new one: people have always brought paper policies to agents’ offices: Trellis just replaces paper with a self-service digital channel, on the consumer’s own schedule, more accurate and less error-prone than a call center.

He drew a direct contrast with Plaid’s own origin story. Plaid was a response to existing, clunky data aggregation tools (Intuit’s data API, Yodlee) that weren’t built with a modern developer experience in mind. Trellis, by comparison, isn’t displacing an existing standard — there simply wasn’t one in insurance, beyond faxed loss reports or expensive, centralized credit-bureau-style solutions. That means Trellis is opening a genuinely new market rather than disrupting an established one.

Why Insurers Shouldn’t Fear Commoditization

Daniel said he gets one question constantly: if consumers can port their insurance data out, won’t incumbent carriers get commoditized as customers shop around more easily? His answer draws directly on how the Plaid story actually played out for banks — initially skeptical that enabling apps like Venmo and Robinhood would bleed customers away, banks are, years later, still dominant, in part because becoming the trusted source of record makes a bank more valuable to its customers, not less. He expects the same dynamic in insurance: most Trellis users, in his direct experience, don’t actually want to switch to a smaller, unfamiliar carrier — they stay loyal to brands they already trust, but they’re more likely to buy additional or better-fitting coverage once an independent voice can tell them clearly what they’re missing. He used a baking metaphor to make the point: opening insurance data up via API doesn’t shrink any single participant’s slice — it grows the whole cake, largely by addressing the industry’s chronic problem of underinsurance.

Iterating Toward Product-Market Fit

Trellis didn’t land on its working business model immediately. Daniel described trying, and largely failing, to license the technology directly to other insurtechs (too early-stage and focused on their own core rollout), to large incumbent carriers (a poor fit between corporate pace and startup agility), and to individual independent agents (insufficient volume and digital reach to make the integration worthwhile). What ultimately worked was licensing Trellis’s technology to fintech apps whose users were already logging in for financial advice — Trellis could then match those same users with better-fitting insurance coverage or a more competitive carrier, creating and directly capturing value in one motion, rather than just theoretically creating it.

He framed this as a broader lesson for founders: before raising serious capital, work through four questions in sequence — can you actually build the idea (is it technically feasible), will people use it, does it create real value for them, and can you actually capture a share of that value. He used online dating apps as an example of the gap between creating enormous value (helping someone meet a life partner) and being able to fully capture it, since users are reluctant to credit an app for something that personal.

Funding: Profitable Almost by Accident

Trellis raised a $2 million seed round led by General Catalyst, after Daniel and a small team of Google-connected engineers proved the core data-exchange technology was buildable using their own time and money — deliberately de-risking feasibility before asking for outside capital. Trellis later raised a $10 million Series A led by QED Investors, chosen specifically for QED’s fintech network and credibility, which Daniel saw as essential to convincing the broader fintech and financial services community to take insurance seriously as an adjacent category.

He noted, with some self-deprecating humor, that Trellis reached $2 million in revenue having burned less than half its seed capital, became profitable within a year, and was profitable enough the prior year to actually owe federal income tax — a genuine rarity for an early-stage startup, and one he described only half-jokingly as an unintentional side effect the company was now actively working to reverse by investing more aggressively in growth.

What’s Next: Deeper Integrations

Looking ahead, Daniel described three parallel growth directions for Trellis: broader market penetration of the login-based data-sharing experience across the entire insurance shopping journey (from initial search through to speaking with an agent), deepening relationships with top-tier insurers now that Trellis has enough scale to matter to them directly, and exploring integrations further upstream — not with agency management systems like Vertafore’s AMS360 or Ivans, which he sees as less central, but with core policy administration and rating platforms like Guidewire, which hold the actual source-of-truth policyholder data and ultimately underpin any real-time bind API.

He also pointed to a specific, underappreciated industry statistic: online insurance quote-to-sale conversion typically runs around just 5%, meaning the vast majority of people who request an online quote never buy. Even modest improvements against that baseline — Trellis specifically targeting the digitally engaged segment of shoppers, not the whole market — can meaningfully move a carrier’s overall conversion rate.

Advice: Learn to Fly

For his closing recommendation, Daniel admitted candidly that he works close to double a normal workweek and doesn’t have much room for hobbies beyond flying — which he recommends unreservedly, describing America’s general aviation privileges (available to private citizens without a military or commercial airline path, unlike in many countries) as a genuinely underappreciated, joyful experience worth trying at least once via a sightseeing flight.

Key Takeaways

• Trellis’s founding insight came from Daniel’s time at Earnest, where the most common reasons borrowers requested loan forbearance were insurable events (health and car issues) — revealing an unbuilt connection between insurance coverage and personal credit risk
• Daniel’s framing of the “commoditization” fear directly echoes Plaid’s own history with banks: becoming a trusted source of data makes an incumbent more valuable, not less, and most consumers who gain visibility into their coverage choose to buy more or better insurance from a brand they already trust rather than switch away
• Trellis found real product-market fit only after three failed attempts (insurtechs, large carriers, independent agents) — the working model was licensing to fintech apps whose users were already logging in for financial advice
• Daniel’s four-question framework for de-risking a startup before fundraising: can you build it, will people use it, does it create real value, and can you actually capture a share of that value
• Trellis’s next expansion targets core policy administration and rating platforms like Guidewire rather than agency management systems — because that’s where source-of-truth policyholder data actually lives, and what ultimately powers a real-time bind API
• Online insurance quote-to-sale conversion runs around just 5% industry-wide — a statistic Daniel points to as a clear, underexploited opportunity for any technology that meaningfully improves it, even for a narrow, digitally engaged segment of shoppers