Continue Reading
The first wave of insurtech looked, from the outside, like it had the right idea. Take a product category that everyone hates interacting with — insurance — apply better design and technology, acquire customers directly, and build a better experience. The valuations reflected that optimism. The operating results, eventually, told a different story.
The problem was not the product. Better design and faster claims processing are real improvements. The problem was customer acquisition. Insurance is not a category where people naturally seek out better options. They buy insurance when they have to, typically at moments of transaction — when they buy a home, a car, a business license. The direct-to-consumer insurtechs were trying to acquire customers outside of those transactions, which meant competing for attention on generic digital channels at CAC levels that crushed the unit economics.
A typical direct-to-consumer auto insurance CAC ran $400 to $800 per policy in 2021-2022. The annual premium on a mid-range auto policy is somewhere between $1,200 and $2,000. After reinsurance costs, claims, operations, and the CAC, the first-year economics on many of these policies were negative. The companies were banking on multi-year retention to earn back the acquisition cost.
Retention didn’t perform the way the models assumed. Insurance customers are price-sensitive at renewal. The digital channels that made direct acquisition efficient also made comparison shopping easy. Customers who arrived through price comparison sites left through price comparison sites. The direct-to-consumer insurtechs were, in many cases, buying customers they couldn’t keep.
The embedded insurance model inverts the distribution problem. Instead of acquiring customers cold, the insurer or MGA is embedded in a transaction context where insurance is a natural, expected component of the purchase.
A few examples of where this works:
In each of these cases, the customer is already in a transaction context where insurance is relevant. CAC drops dramatically — in some cases to near zero, since the distribution partner absorbs the customer acquisition cost. Conversion rates are higher because the timing is right. And in B2B contexts, the software relationship creates ongoing data that improves underwriting over time.
The best embedded insurance products are ones where the distributor has data the underwriter can’t get anywhere else. A freight platform that knows shipment routes, cargo values, and carrier history can underwrite cargo insurance better than any traditional carrier. That’s a genuine advantage, not just a distribution play.
Most of the embedded insurance companies worth watching are operating as MGAs — Managing General Agents. They handle underwriting, distribution, and claims, but transfer the balance sheet risk to a carrier. This is structurally efficient: the MGA captures the value of better distribution and better underwriting data without the capital-intensive requirements of being a carrier.
The risk in the MGA model is carrier dependency. If your carrier exits the program or tightens capacity, your business model is at risk. The well-run MGAs are working with multiple carriers, building reinsurance relationships, and in some cases pursuing carrier licenses as they scale. Building carrier optionality into the capital structure from the beginning is the right move.
The embedded insurance categories that attract our attention are ones where the distribution partner has proprietary data that improves underwriting. Travel and product protection are relatively mature. The more interesting opportunities are in B2B verticals where software-native distribution meets genuinely better risk data:
Cyber insurance distribution through IT management and security platforms. The platforms already have visibility into a company’s security posture that no underwriter can replicate from a questionnaire. That data advantage should translate into better pricing and lower adverse selection.
Construction and trade credit insurance through project management software. The platforms know which contractors have completed projects on time and on budget, which have liens filed against them, and what their crew and equipment utilization looks like. That’s far richer underwriting data than a credit report.
The distribution problem in insurance is not fully solved. But embedded distribution, in contexts where the timing and data are right, is a fundamentally different model from direct-to-consumer. The companies that get this right will look very different from the first wave of insurtechs. They’ll have lower CAC, better loss ratios, and sticky distribution partnerships that compound over time.
Building in embedded insurance or insurtech infrastructure? We’d like to hear from you.