A 1998 Truck and a 2020 Truck Get the Same Quote From Every Incumbent

Before insurance, Ian White built maps — literally. His first company, Urban Mapping, started by printing tactile, 3D-textured city maps (New York and Chicago were early hits in the early 2000s) before pivoting into aggregated location data and powering the mapping backend behind Tableau for roughly five years, ultimately selling the company to Pitney Bowes. From there he moved into earthquake catastrophe insurance, drawn by a specific, nagging inconsistency he’d noticed living in San Francisco: consumer earthquake pricing leans heavily on zip code, even though the real driver of earthquake damage is soil type, and zip code has essentially zero correlation with actual soil composition. He briefly explored a low-cost accelerometer-based approach to directly measure soil characteristics at a property, before moving into a stint building alternative-data-driven stock-picking models for a company that was later sold to a hedge fund — work he ultimately found unsatisfying on a fairly fundamental level, since it was a zero-sum trading exercise that didn’t leave anyone better off.

Looking for a genuinely antiquated industry where his alternative-data instincts could matter, he was eventually pointed toward commercial insurance, and then specifically toward trucking — jokingly described to him at the time as “the Satan’s pit of insurance.” He found his co-founder, Mike Dorfman, a fourth-generation member of a Brooklyn moving-and-storage brokerage family who’d just left the family business, frustrated by exactly the kind of outdated, fax-machine-driven processes Ian was looking to disrupt. Together they built Koffie Labs.

In Episode 58 of InsurTechTalk, Ian and I covered why nearly every incumbent prices a 1998 truck and a 2020 truck identically, the “reptile theory” behind billion-dollar nuclear verdicts in trucking litigation, and the hard lesson that pushed Koffie Labs from a pure data play into becoming a full MGU.

About Ian White

Ian White is co-founder and CEO of Koffie Labs, a commercial auto MGU focused on trucking and transportation, underwriting fleets from single owner-operators up to roughly 100 power units — a segment representing more than half of all trucks on the road and 95% of motor carriers. Before Koffie Labs, he founded Urban Mapping (acquired by Pitney Bowes) and worked in earthquake catastrophe risk and alternative-data-driven equity research. He co-founded Koffie Labs with Mike Dorfman, a fourth-generation moving-and-storage brokerage veteran.

A Line That’s Been Losing Money for a Decade

Ian opened with the scale of dysfunction in trucking insurance specifically: the line has seen roughly 10% average annual rate increases for approximately the past decade — around 40 consecutive quarters of rate hikes — and yet the industry’s combined ratio still runs around 115%, meaning every dollar of premium pays out $1.15 in claims and expenses. Only the top quartile of the market is actually profitable; the other three quartiles are underwater, with combined ratios ranging anywhere from the 80s to 150 or higher depending on the carrier.

His diagnosis traces to two compounding causes. First, roughly a decade-plus of near-zero interest rates removed the investment-income cushion that historically let carriers tolerate mediocre underwriting results — once that tide went out, underwriting had to stand entirely on its own, and this line, in his assessment, never developed the discipline to do that well. Second, and more structurally, underwriting in this line remains coarse and heuristically driven rather than genuinely granular.

Why a 1998 Western Star and a 2020 Volvo Get the Same Quote

Ian’s clearest illustration of that coarseness: a fleet with a 1998 Western Star manual transmission truck with three million miles on it, and a 2020 Volvo VNL with full Level 2 driver-assist capability and modern ABS systems, will typically get essentially the same quote from most incumbent insurers — because underwriting at that level mostly considers actual cash value (for physical damage) and weight class, not the vehicle’s actual safety technology.

Koffie’s alternative: a proprietary model, called the CASE score, built on roughly 150 vehicle-specific features, sorting the entire population of trucks into ten deciles of relative risk. Ian’s headline finding from that model: the safest decile of trucks is roughly four times safer than the least safe decile — a distinction that simply doesn’t exist in how most incumbents price this line today.

Mandatory Cameras as a Self-Selection Filter

Koffie requires telematics on every insured vehicle, exclusively through Samsara, including both an inward-facing (driver) and outward-facing (road) camera. Refusing to install them is grounds for non-binding or policy termination — a hard requirement Ian described as a genuinely useful self-selection tool. Some safety managers love Koffie’s pricing but balk at the cameras; his response is unambiguous: that refusal itself is meaningful signal about the underlying operation, regardless of the specific reason given, and Koffie simply isn’t a fit for that fleet. He noted this segment is functionally impossible to underprice into compliance — even a hypothetically free policy wouldn’t convince a fleet fundamentally opposed to being monitored.

That telematics data feeds real-time loss control: instant crash notifications packaging telemetry and policy data immediately for Koffie’s TPA, live behavioral coaching (catching distracted driving as it happens), and underwriting based on where trucks are actually hauling rather than a generic geographic radius assumption.

Distribution: Stake, Golf, Trust — Then Premium and Technology

Koffie’s distribution is entirely agency-driven, consistent with the rest of commercial trucking. Ian was direct that the “myth of going direct,” as he put it, applies here just as it has in personal lines — agents remain genuinely valuable, though their role is shifting. He distinguishes the transactional side of brokering (binding an initial account, which he expects to become increasingly automated, even “robo-quote”-like over time) from the higher-value, durable relationship side — servicing accounts, issuing endorsements, and acting as a trusted risk advisor — which is where Koffie wants to help agencies grow, by giving them more tools and more touchpoints throughout the year rather than just at renewal.

His memorable formula for winning agency relationships: stake, golf, and trust first — genuine relationship investment — with good premium and technology integration following from that foundation, not substituting for it.

Nuclear Verdicts and the “Reptile Theory”

Ian walked through why liability, not physical damage, is the central concern in this line specifically. Koffie writes policies with $1 million combined single limits, and the real tail risk comes from what the industry calls “nuclear verdicts” — outsized jury awards driven by a plaintiff-attorney tactic known as the reptile theory: rather than framing a crash as a random accident, attorneys argue it’s the inevitable outcome of a fleet’s sustained, systemic negligence (unwashed trucks, skipped inspections, sloppy maintenance records), appealing directly to jurors’ instinct to protect themselves and their own families from a company they’re primed to see as recklessly dangerous. Ian cited a real example: a verdict exceeding $1 billion handed down in Jacksonville, Florida over a single fatality, contrasted with another case involving a fatality that settled at roughly $500,000 — illustrating just how unpredictable and disconnected from the underlying facts these awards can be.

Roughly 200 Data Points, and Why Harsh Braking Isn’t Always Bad

Underwriting at Koffie draws on roughly 150 vehicle-level features (inspection, crash, and violation history; ownership and usage history — noting, for instance, when a truck built for local hauling in the Northeast was later sold into long-haul use in the Southwest; mechanical configuration like day cab versus sleeper cab, transmission type, and axle configuration) plus an emerging category of technology-specific data: which ADAS and driver-assist components are actually present, which ECUs support them, and whether the relevant software is current and active. That’s on top of separate company-level data (financial stability, local versus over-the-road operation, cargo type) and driver-level data — a meaningfully more complex picture than personal auto telematics, which has moved toward narrowly isolating driver behavior and deliberately excluding proxies like credit history.

He offered a genuinely nuanced point about interpreting that data: harsh braking alone isn’t necessarily a bad signal. It can just as easily indicate that a driver, or the vehicle’s own safety system (electronic stability control, lane and collision avoidance), successfully avoided what would otherwise have been a crash — a distinction that matters more and more as vehicle technology takes on a larger share of real-time safety decisions.

Why “the Robot Truck Ate My Business Model” Isn’t the Right Fear

Ian’s longer-term thesis: as the driver’s mechanical role in operating a vehicle decreases with rising levels of autonomy, the underlying risk doesn’t disappear — it migrates onto the vehicle and its technology instead, and incumbents structurally can’t see or price that shift because their models remain anchored to driver-centric risk factors. He drew a historical parallel: aviation insurance was new when it emerged, but not unprecedented, since the industry adapted using existing loss experience from maritime insurance — which is also, he noted, why aviation coverage is still called “hull insurance” today.

Will OEMs Just Eat the Insurance Industry?

Asked directly about the common investor worry that OEMs (Tesla being the frequently cited example) will simply absorb the insurance layer themselves given their proprietary vehicle data advantage, Ian pushed back with a specific structural argument. Tesla operates its insurance offering as an MGA — leveraging real data advantage, but deliberately not taking on full carrier risk directly, a lesson he traced back to GM’s regulatory trouble in the 1980s for putting its own balance sheet behind a financial services push outside its core competency (GM now runs a similar, narrower model through OnStar). Critically, Tesla’s insurance only covers Tesla vehicles — a household that owns both a Tesla and a Volvo still needs a separate policy for the Volvo, meaning OEM-based insurance inherently fragments a household’s or fleet’s coverage rather than consolidating it.

The same dynamic plays out in trucking: OEMs can reasonably offer physical damage coverage directly, since assessing actual cash value is straightforward, but liability underwriting is far more complex and better suited to a specialized player like Koffie. He also flagged a practical embedded-insurance limitation specific to fleets: a small operator buying one new truck from a dealership without a national insurance program creates a genuine coordination problem with the rest of that operator’s existing multi-vehicle policy — a friction point Koffie is actively working through rather than treating as solved.

The Missing Line Item in “Total Cost of Ownership”

Ian argued that trucking’s conventional total-cost-of-ownership calculation — typically limited to fuel, labor, and maintenance — has historically ignored insurance almost entirely, despite it varying enormously: smaller fleets pay roughly three times more per mile traveled on insurance than larger fleets on average. His pitch to OEMs and dealer networks: since safer, better-equipped trucks earn real premium credits under Koffie’s model, and Koffie’s own numbers suggest a payback period of zero to two years, a higher-priced, safety-equipped vehicle can genuinely offer a better total cost of ownership story once insurance is properly included.

He was pointed, though, about what actually backs up a claim like that: real actuarial work, not marketing. He explicitly distanced Koffie from vague, unsubstantiated claims like “save up to 30% with our telematics,” insisting that any specific safety-technology discount has to be proven with real loss-cost data tied to vehicle miles traveled, comparing a given technology or vehicle configuration against a genuine baseline — a slower, harder path than simply asserting a number, but the only one he considers legitimate in a regulated line.

Why Koffie Became an MGU (After Trying Everything Else First)

Koffie’s original plan wasn’t to become an MGU at all — it was to sell a pure SaaS product: a proprietary safety score tied to a vehicle’s 17-digit VIN, sold directly to carriers and reinsurers. Ian and his team pitched that score to roughly 30 primary and reinsurance carriers, and the near-universal response was genuine interest paired with an inability to actually use it: most trucking policies are written on a composite-rated basis, with no real room in existing underwriting models to incorporate vehicle-level granularity like this. Selling the score alone would have gotten lost in the mix, leaving the same slow, roughly ten-week quoting cycle unchanged.

That directly forced the pivot into becoming a full MGU. Ian’s explicit, hard-earned advice to other founders: don’t jump to building an MGA or MGU because it looks exciting — take that path only once you’ve genuinely exhausted every other alternative, since it’s an enormously harder route than most founders appreciate going in. Over half of Koffie’s employees come directly from the insurance industry, including co-founder Mike Dorfman’s producer background and Chief Risk Officer Justin’s experience at Great West, a historically strong trucking-line carrier, giving the team the underwriting and relationship depth to actually execute on that harder path.

Structure and Growth

Koffie Labs operates as an MGU with full binding authority, underwriting through carrier partner Sutton National, a rated, admitted carrier still completing its own state rollout. At the time of recording, Koffie was live in Illinois and Tennessee, with more states and new products planned for 2022 — trucking fleets typically need between four and ten separate insurance products to meet state and federal requirements, and Koffie’s ambition is to become a genuine one-stop shop rather than making agents source additional coverage elsewhere. The company raised a seed extension round in July 2020 led by Lerer Hippeau, with Anthemis also participating. Looking ahead, Ian described plans to build relationships across the full vehicle lifecycle — OEMs, dealer networks, Tier 1 and aftermarket safety component manufacturers, diagnostic providers, and fleet management companies — framing these as complementary distribution and data partnerships rather than competitive threats, since Koffie has no interest in becoming a hardware manufacturer and manufacturers have no interest in becoming an insurer.

Advice: A People’s History of the United States

Asked for a closing recommendation, Ian pointed to A People’s History of the United States by Howard Zinn — a history told from the perspective of the marginalized and oppressed rather than the traditional dominant narrative. He was candid that he’d never encountered the book in high school or university despite its long publication history, and described reading it now as a genuinely valuable, if frustrating, exercise in re-examining assumptions he’d taken for granted — landing on the broader point that history is inherently a matter of perspective, shaped by whoever holds the pen.

Key Takeaways

• Trucking insurance has run roughly a decade of consecutive rate increases while still posting a combined ratio around 115% industry-wide, a dysfunction Ian traces to the loss of investment-income cushioning after years of near-zero interest rates, combined with coarse, heuristic underwriting that never adapted
• Koffie’s CASE score sorts trucks into ten risk deciles using roughly 150 vehicle-specific features, finding the safest decile of trucks is roughly four times safer than the least safe — a distinction incumbents pricing by power-unit count and base rate simply can’t see
• Mandatory dual-facing telematics (via Samsara) functions as a deliberate self-selection filter: fleets unwilling to accept camera monitoring are turned away regardless of price, since Koffie considers that resistance itself meaningful underwriting signal
• Trucking’s real tail risk is liability, not physical damage — “nuclear verdicts” driven by the reptile theory (framing a crash as systemic corporate negligence rather than an isolated accident) can produce wildly disproportionate jury awards, illustrated by a real $1 billion-plus verdict over a single fatality
• Koffie’s original go-to-market — selling a pure vehicle safety score to roughly 30 carriers and reinsurers — failed because most trucking policies are composite-rated with no underwriting model flexible enough to use vehicle-level data, directly forcing the pivot to becoming a full MGU
• OEM-based insurance (Tesla’s MGA model, GM’s OnStar) is structurally limited by only covering that manufacturer’s own vehicles, fragmenting rather than consolidating a fleet’s coverage — a real constraint on how far OEMs can realistically push into insurance themselves
• Any safety-technology discount claim has to be backed by genuine actuarial work tied to real loss-cost and vehicle-miles-traveled data — Ian was explicit in rejecting unsubstantiated marketing claims common elsewhere in the telematics space