Latest / Fintech Conversations with Fexingo: Banking, Payments, and Financial Software Companies / How Fintechs Use Real-Time Vehicle Data for Dynamic Insurance
Transcript
- Lucas: So earlier this year, Lemonade officially shut down its acquisition of Metromile's insurance operations, bringing the pay per mile model under its own brand. And that got me thinking about how far real-time vehicle data has come in reshaping auto insurance. Luna: I remember when Metromile was still a standalone company — the whole pitch was that you pay based on how much you actually drive, not some actuarial average. Lucas: Exactly. And that's the core idea behind usage-based insurance — UBI. Instead of grouping everyone by age, ZIP code, and credit score, you look at miles driven, time of day, hard braking, acceleration patterns. All that data comes from a tiny obd ii dongle plugged into the car's diagnostic port, or increasingly from the smartphone itself. Luna: And from automakers too, right? Tesla collects driving data natively. GM has OnStar. Ford has its connected services. Lucas: Right — that's the next frontier. But let's start with the dongle model because it's the most proven. Metromile had about 100,000 policies at its peak, and the data showed that pay per mile customers saved an average of 40 percent compared to traditional premiums. The math is simple: if you drive less than 10,000 miles a year, you're subsidizing high-mileage drivers in a standard pool. Luna: Forty percent is huge. But the industry adoption has been slow — I think the biggest barrier is trust. People get nervous about a black box in their car. Lucas: That's real. A 2024 survey by J.D. Power found that only about 15 percent of drivers have usage-based insurance, but nearly 40 percent said they'd consider it for a meaningful discount. So there's a willingness, but the privacy math has to work. You're trading granular driving data for a lower premium, and not everyone is comfortable with that. Luna: And it's not just privacy — there's the question of where the data goes. Could your insurer raise your rates because you drove through a pothole neighborhood? Or share data with law enforcement? Lucas: Those are live concerns. The regulatory framework is still catching up. California actually banned usage-based pricing for a while, though it's come back in limited forms. But the technology itself is getting more sophisticated. Some insurers now use smartphone sensors — accelerometer, GPS — to detect driving behavior without any hardware. That lowers the barrier to entry. Luna: So you download an app, enable location permissions, and the insurer builds a risk profile over a few weeks. That's what Progressive's Snapshot does, right? Lucas: Snapshot started with a dongle, but they've moved to an app-only option. And Allstate has Drivewise. The difference is that these are still discount-based models — you get a retrospective discount after a period of monitored driving. True real-time dynamic pricing, where your premium changes week to week, is rarer. That's where the fintech edge comes in. Luna: Because to adjust premiums in real time, you need a modern policy administration system that can compute risk on the fly, not just a batch file run once a quarter. Lucas: Right. And that's where companies like Root Insurance came in — they started with a smartphone-only model, using a test drive to give you a quote. Root went public via SPAC, struggled post ipo, but the concept showed that you could underwrite based on actual behavior rather than proxies. The challenge was adverse selection: good drivers loved it, bad drivers self-selected out, so the risk pool was weird. Luna: So the actuarial models had to adjust. But now you're seeing a second wave — not just based on miles, but on driving quality. Hard braking, rapid acceleration, cornering forces. Even the time of day you drive matters — driving at 2 a.m. is statistically riskier than 10 a.m. Lucas: And some companies are layering in contextual data: road conditions, weather, traffic density. In theory, an insurer could charge a higher rate for driving in a snowstorm than on a dry highway. But that's still mostly theoretical — the real frontier is the OEM data. Luna: Tesla launched its own insurance product in 2019, using vehicle telemetry to calculate a safety score. They adjust premiums monthly based on things like forward collision warnings, autopilot disengagements, and following too closely. That's probably the most advanced example of real-time vehicle data in insurance. Lucas: It's fascinating because Tesla owns the full stack — the car, the data, the insurance policy. No dongle, no app friction. The customer is already generating the data just by driving. And Tesla can price risk more accurately because they see exactly how the car is used. They claim some drivers save 20 to 30 percent compared to traditional insurers. Luna: But that also creates a lock-in: if you want the good rates, you stay with Tesla Insurance. And the safety score algorithm is a black box — drivers don't always know why their score dropped. Lucas: Right, and that's the tension. Greater data granularity can lead to fairer pricing, but only if the model is transparent and the customer has some control. There's a lot of research showing that when people understand the data being collected and how it affects their premium, they're more willing to participate. But if it feels like surveillance, backlash is quick. Luna: Do you think we'll eventually see a federal privacy standard for vehicle data? Right now it's a patchwork — some states have laws, others don't. Lucas: It's possible. The Biden administration proposed guidelines for automotive data privacy, but nothing binding has passed. The FTC has brought enforcement actions against companies that misuse driving data. But a unified standard would help adoption — it would give insurers a clear legal framework and give customers a baseline of protection. Luna: So where does this leave the traditional auto insurer? Are they going to be disrupted, or will they just buy the tech? Lucas: I think it's both. Lemonade buying Metromile is a good example — a digital-native insurer acquiring the technology and customer base of a pay per mile pioneer. You also have State Farm partnering with Google's Sidewalk Labs on urban mobility data, and Allstate investing in Arity, a telematics data platform. The incumbents are not asleep. Luna: But the real winners might be the data aggregators — the companies that collect and normalize vehicle data from multiple sources and sell it to insurers. Because not every insurer can build its own telematics pipeline. Lucas: Exactly. Companies like Cambridge Mobile Telematics — they power a lot of the usage-based programs for major insurers. They process data from smartphones, dongles, and connected vehicles, and provide risk scores. That's the infrastructure layer. And it's a scalable business model, because they don't take insurance risk — they just provide the data. Luna: And they're expanding into commercial fleets too. Real-time data for delivery vans, ride-share drivers, trucking — that's a huge market where every mile matters for cost. Lucas: Fleet insurance is actually ahead of personal auto in terms of adoption. Because if you're running a logistics company, you already have GPS trackers, you already monitor driver behavior for safety. Adding insurance telematics is a small incremental step. Some fleets have seen accident frequency drop by 20 percent just from feedback alone. Luna: And that's real-time feedback — a driver gets an alert after a harsh braking event, rather than a report at the end of the month. That's a behavior change loop. Lucas: Right, and the same loop applies to personal insurance. The more immediate the feedback, the more likely drivers adjust. Some insurers give you a weekly score with tips to improve. That's actually a shift from insurance as a financial product to insurance as a behavior-modification service. Luna: And that's a whole different relationship with the customer. Instead of an annual renewal notice, you have a monthly or weekly interaction. That's more like a subscription. Lucas: If these conversations are useful for what you're building or running, a couple of dollars a month is genuinely what keeps these going — buy me a coffee dot com slash fexingo, if you've gotten something out of them. Luna: Yeah, it's a small way to keep the show ad-free and focused on the real details. Lucas: So back to the insurance side — we're also seeing embedded insurance pop up. When you buy a used car on Carvana, you get an insurance quote at checkout. That's real-time vehicle data — Carvana knows the VIN, the mileage, the condition — and they can quote you in seconds. Luna: And that's powered by APIs from insurtechs like Trov or Cover Genius. They bundle the insurance into the transaction, so the customer doesn't have to shop around separately. Lucas: Exactly. And that's where real-time data really shines — at the point of decision. Whether it's buying a car, renting one, or starting a delivery shift, the ability to get a personalized, accurate premium in seconds is a massive improvement over filling out a form and waiting for a quote. Luna: But that also requires the data to flow instantly — from the vehicle, to the aggregator, to the insurer's pricing engine, back to the customer. Latency matters. Lucas: Ten seconds is the benchmark that most embedded insurance players aim for. Any longer and the customer drops off. So the tech stack has to be tight — real-time API calls, pre-computed models, fallback logic if data is missing. It's a different engineering challenge from traditional batch processing. Luna: And the data quality matters enormously. If the vehicle's mileage is wrong or the telematics signal is noisy, the quote could be off. That creates a trust problem. Lucas: True, and that's why the best implementations use multiple data sources to cross-validate. For example, combining the obd ii mileage with the vehicle's own odometer reading from a CARFAX report. The more data points, the more accurate the risk profile. Luna: So we're moving from a world where your insurance rate is based on who you are — your age, gender, credit score — to one where it's based on what you actually do. That feels more equitable in some ways, but also more invasive. Lucas: It's a trade-off that every consumer will have to weigh. But I think the direction is clear. As vehicles become more connected, the data is going to be generated whether you opt in or not. The question is who controls it and how it's used. Luna: And that's the big policy debate still unfolding. The EU has strict data portability rules under GDPR — you can take your driving data from one insurer to another. In the US, it's much more fragmented. Lucas: Yeah, and that fragmentation is probably the biggest obstacle to a truly dynamic, competitive insurance market. If you can't easily switch insurers with your driving history, you lose a lot of the benefit. Portability is key. Luna: So what's the one thing you'd tell a listener who's considering usage-based insurance today? Lucas: If you drive less than 10,000 miles a year and you're a safe driver, it's probably worth trying at least the discount-based version from a major insurer. You can always opt out after a few months. But read the privacy policy — know exactly what data is collected, how long it's kept, and whether it can be shared. Luna: And if you have a newer car with built-in connectivity, check if your automaker has a partnership with an insurer. You might already be generating data that could save you money. Lucas: Right. Real-time vehicle data is still early, but the trend is unmistakable. In five years, I think the default auto insurance will be some form of usage-based, with traditional fixed premiums becoming the exception. And fintech companies — whether they're insurers, data platforms, or embedded distributors — will be at the center of that shift.