CSaaS
In this essay, I’ll talk about some startups that I like and why I like them. First of all, I’m going to put insurtechs into three categories; then I’ll talk about an insurance startup which defies that classification; then I’ll fit that startup into a broader category that extends beyond insurance.
There are basically three kinds of insurance startups.
Some are fairly generic traditional vSaaS plays that just happen to sell to insurance companies: think policy administration, claims management, document processing, etc. This category (very roughly) targets the expense ratio, and there have been wins here: Guidewire is now a $10bn public company, Indio had a nice exit.
Some sell insurance-specific data and/or analytics, giving big insurers and brokers new capabilities and reducing their loss ratio: the iconic example here is RMS, but anything that helps predict wildfires/credit default etc falls into this category.
Some compete directly, beating incumbents at their own game - think Lemonade, Hippo, Coalition, Metromile. Ironically, the best startups in this genre aren’t tech companies at all - they’re RenRe, Acrisure, and Ryan Specialty, all founded by insurance veterans.
The first category can generate great outcomes, but it’s very painful because you have to sell to insurers; it took Guidewire 18 months to get their first customer, another 18 months to get their second, they nearly got killed by an Accenture lawsuit 7 years in, and then they had to figure out how to pivot to cloud. For more on Guidewire’s story, listen to this excellent interview. Maybe Novidea or hyperexponential can win in this category.
I’m not convinced that the big outcomes are there in the second category. RMS was a one-off. You’re basically trying to productise academic research (into e.g. wildfire prediction), and while that might lead to a nice acquihire, it’s hard to turn that into a unicorn. Best case, you get something like Tractable.
If you want to win in that third category, you need to convince yourself and the world that you, the startup, can get to distribution before the incumbents get to innovation. Plenty of unicorns have been minted by following this approach. But incumbents have massive distribution, so you have to be incredibly innovative.
Lemonade gets a lot of stick, but the thing they did well was to leverage mobile as a way to genuinely revolutionise insurance distribution; synthetic agents might be a clumsy framing, but I believe that their GTM machine is/was great. In a different vein, Coalition and At-Bay’s innovation was strong enough to overcome incumbents because they are/were the only people employing genuine military-grade hackers. Perhaps Reserv will win here as a tech-enabled TPA.
On the other hand, if you want to see how not to do this, then check out this stunningly brutal (and sadly overlooked!) takedown of Metromile. It’s a work of art.
If none of those categories sound particularly promising… then that’s why insurtech is hard. Very few founders are well placed to do it.
One of the only insurance startups that I do like is great because it breaks out of those categories; that startup is Authentic.
Authentic, which announced an $11m Series A in June 2024, provides infrastructure for other companies to offer insurance to their customers: they set up a captive insurance program for companies like Mindbody or SpotOn to sell insurance to their gym and restaurant clients. Rather than buying a business owner’s policy from a traditional insurance broker/agent (or, uh, a startup like Rainbow, which falls into Category III), Authentic make it so that businesses can buy insurance from their existing vendors; your POS or ERP provider can now sell you a policy!
The reason this is genius is that a company like Mindbody isn’t focused on the economics of the insurance policy per se, because those policies add to the cost of switching away from Mindbody to a different core provider. For Mindbody, the goal doesn’t have to be underwriting profit; it’s improved net dollar retention. They could sell insurance to their customers at a loss if there are synergies with their main business.
That said, it might not be a loss-leader for Mindbody - they’re well-placed to understand their clients’ business, and they have a direct channel of communication with them. I guess there’s an alternative story you can tell where this is not purely a NDR play - it’s a way to exploit low CAC, upselling additional services to existing customers in a profitable way.
Maybe there’s two kinds of businesses here. Some businesses have profitable core offerings, and bundle on low- or negative-margin offerings to make those core products stickier; Mindbody, as a SaaS platform, should have great gross margins, and so it should fall into this category.
But other businesses have low- or negative-margin core products, but sell high-margin ancillary services. That might be true of some low-cost airlines and banks; and a fine art underwriter once told me that a company like Oak Furniture Land barely breaks even on the furniture, but makes a lot of money by selling warranties - insurance policies with incredibly low loss ratios. Similarly, I think the iBuyers (i.e. Zillow and Opendoor) didn’t necessarily expect to make money by trading houses - they wanted to make money on ancillary services around that core transaction.
Mindbody are not the first software vendor to sell insurance to their customers. I’ve seen vendors take two approaches in the past: partnering with insurance companies through a quasi-referral scheme, and setting up their own broker. The first approach probably nets the vendor company about 2% of insurance premiums; the broker approach would net 10-20% of premiums.
In 2018, Cisco partnered with AIG and Allianz to offer cyber insurance to their clients; and in March 2023, Procore, the construction SaaS company, launched a retail broker (Procore Risk Advisors) aimed at their clients. As of 2024, that business has roughly 20 employees, which would imply about $4m revenue, based on industry revenue/employee numbers. As I understand it, that brokerage performs a roughly equivalent function to an Authentic captive.
It’s worth making two points here: first, construction insurance is a much bigger line item than cyber insurance. As a rough, rough, rule of thumb, companies buy cyber insurance cover against 1% of their revenues, and pay 1% of that coverage in premium. By contrast, construction companies buy various forms of construction insurance against 10% of their revenue, and pay 10% rates - so all else being equal, Procore’s customers are paying 100x more for their construction insurance than Cisco’s are for their cyber.
Second, Procore is a really big SaaS company - they did $950m revenue in 2023, compared with something in the low nine figures for Mindbody. Only at that stage - March 2023 - did it make sense for them to set up an insurance subsidiary. Authentic, by contrast, offer a lower-touch, presumably lower-cost alternative, that means it makes sense for vendors to sell insurance much earlier in their lifecycle.
Let’s zoom out for a second. One of my favourite tech articles is John Luttig’s 2022 article about Rippling. The main reason I think this is a good article is that it persuaded me in 2022 that Rippling was a good company through, like, pure intellectual convincingness, and that impression has lasted. It’s great marketing.
In the article, John makes the point that most startups start as a niche product, and then aim to expand sideways to become a platform. By contrast Rippling, because of the nature of the founder and the space, set out to be a compound startup. Rather than tackling a single niche, Rippling spent a ton of money on R&D to go after all of them at once:
Rippling was not looking to build a software company. It was looking to solve the employee management problem, requiring coordination across employee data silos. Parker started with the employee data asset: by building on top of the employee record, Rippling could develop a compound product quickly.
…
A common middleware layer on top of employee data empowers not only internal product teams, but also customers to build complex workflows on top of their employee data that go well beyond traditional HR use cases.These workflows are otherwise nearly impossible to execute: for example, Rippling customers can configure custom workflows that allow salespeople to give customer pricing discounts based on their level within the org. Or ingest Zendesk tickets to see payroll costs per support ticket resolved. Or run reports across silos of business data.
It’s similar to what Palantir call the ontology - and if you have the founders and engineers to pull it off, it’s very powerful.
Very few startups are compound startups from the beginning; even Palantir only reached that point 10+ years into the business. But almost all successful companies eventually add new product lines. So perhaps there’s a category of businesses that offer new product lines as a service - compound startups-as-a-service? Go on, let’s call this CSaaS.
Authentic is one of these startups. It makes sense for most vendors to sell insurance, but it’s fiddly to get an insurance arm off the ground - and it’s also never going to be a top priority for management. Procore’s broker is still immaterial to overall revenue.
Another is Imprint, which helps companies like Brooks Brothers and Turkish Airlines offer co-branded credit cards, and announced a $75m Series B in November 2023. Delta Airlines are big enough to launch their own credit cards - but with Imprint, Texas supermarket chain H-E-B can do it too. A wise man on Twitter wrote a few months ago that Airbnb should have a co-branded card - maybe Imprint could fix that for them?
Just like Authentic, Imprint might be an NDR play, or a genuine revenue opportunity; I don’t know enough about their finances to tell you. But both Imprint and Authentic bring down the scale at which it makes sense to offer ancillary services; they make it easier for already-successful companies to mature into compound startups. It’s tempting to make this category really broad - to include any startup that helps already-successful companies take the next step a little earlier. Maybe we should include Deel - which reduces the scale at which companies can justify overseas employees.
What sets Authentic and Imprint apart, though, is that they’re creating new product lines for their clients - incremental SKUs that can be sold to their client’s customers. In that sense, the best comparison in traditional insurtech are B2B2C companies - above all, in travel insurance. It makes sense for companies like Booking.com or Skyscanner to bundle travel insurance with the purchase of flights and hotels - and insurtechs provide the infrastructure that makes that possible.
My favourite thing about Authentic is the intellectual purity of the idea; it’s just so clever! I wish I came up with it. Sadly, intellectual purity isn’t always enough in insurance: Byrne Hobart was once bullish on an intellectually pure idea that didn’t come off:
And speaking of regulatory arbitrages, what would insurance look like if you combined the emotional intensity of healthcare with the comparatively relaxed regulatory regime of property and casualty insurance? No need to speculate: that’s how pet insurance works.
That pet insurance idea didn’t seem to go anywhere - I’d be curious to read a postmortem. But another pet insurance company started with a different clever idea. What is now ManyPets began life as Bought By Many. They had a neat idea to reinvent the insurance value chain by combining unusual individuals into groups so that they could buy insurance together at advantageous rates.
Once a group hits a certain number of people (usually 100), the business will begin negotiations with insurers to get the best offer possible. So along with 50 stage hypnotists, Mendel’s team is also on the lookout for 20 yoga instructors, 50 scuba divers and a handful of Norwegian Forest Cat owners.
Unfortunately, their idea might just not have been clever enough: that aggregation is more or less the exact job of big brokers and insurers - to combine enough heterogeneous risks into one pool that the law of large numbers can work its magic. Ironically, Bought By Many subsequently pivoted to become - it’s not clear to me whether there’s a really clever idea at play here.
If companies like Authentic and Imprint fail, I suspect it’s likely to be due to TAM - there may be hundreds of companies ready to launch a captive or credit card, but are there thousands? I guess we’ll see - but I’m rooting for intellectual purity.