Launch HN: Goodcover (YC S17) – Cooperative renters insurance for half the price
I (Chris) worked in traditional insurance for 8 years. That time taught me to love insurance and how it picks people up after disasters, but it also gave me first hand exposure to the things people hate about it – the ever-increasing prices, the adversarial claims negotiations, the mountains of paperwork, and the byzantine decision making. All these inefficiencies kept us from really understanding and working for our policyholders – the people we were meant to serve. I became convinced technology was coming for this industry.
I moved to SF in 2016, which is where I met Dan through a family friend. Dan had co-founded Cloudkick (YCW09) and was now looking to start another company. He also knew technology was coming for insurance, especially after his early career at IBM where he saw just how many “tech consultants” were placed in State Farm. Meeting him was a breath of fresh air – we started Goodcover in 2017 and got into YC right after.
And then… it took us two years to get a product to market. We had opportunities to get going faster – you can get an agent’s license, buy off the shelf software, and sell other companies’ products in a matter of weeks. But they would be the same crappy, overpriced, adversarial products that everybody sells, and everybody hates. What good is that? Instead, we didn’t take the shortcuts and stuck it out to change the business model to cut the price in half, and Goodcover is the result.
The story of how we did this starts with how insurance prices are made. If you have lots of claims data, you can run regressions to learn how underwriting factors like location, customer data, previous losses, etc all affect the frequency and severity of claims for every dollar of coverage you are providing. You then load that claims model with your expenses and desired profit margin, and boom you have insurance rates. You then have to get the Department of Insurance in each state you enter to sign off on your rates (not too low so you lose money, not so high that the government calls out your gouging).
We knew technology would save us a lot on processing costs – with Dan’s technical background we knew that we could build technology that would run the business for a fraction of the cost that the typical industry vendors charge. We weren’t going to be saddled with huge agency forces or massive brand advertising. But, we didn’t have any claims data.
Enter Quirk 1 of the insurance industry: all personal lines insurance pricing is public. Since all product and pricing is approved by the state government, to start something new you need to essentially reconstitute work other companies have done, proving that the elements you choose work for your target market. This is why most new insurance offerings are basically just another version of the pricing model sold by the “Insurance Services Office” (ISO – yes, that’s a company, not a government agency). It’s approved everywhere and used by everyone, so it’s a quick start. Lemonade uses ISO with one important modification: they set their minimum premium at $60 instead of $120, allowing them to claim an introductory price of $5 a month. If you buy more than the minimum coverage though, you’ll quickly get to “everybody else’s price” territory.
However to cut the price without sacrificing coverage, we couldn’t use the same model that everyone else does. We needed a more granular model where we could charge the safest 99% people very little in exchange for charging the riskier 1% more. In my insurance career I had learned a lot about models designed for high-value homes, jewelry, cars etc – these models price in catastrophe (like hurricane and wildfire) very precisely to manage exposures in high-risk places. This granularity results in much lower prices for safer risks, since prices there don’t need to subsidize risky ones like they do in traditional “mass market” models. We decided to adapt these models to build our own that would be applicable to our target market, i.e. Renters. Our competitors that don’t use these models are in a bit of a pickle, because they can’t raise prices for high risks too fast thanks to regulatory constraints, meaning they have to keep prices high in safe places to balance the book.
It also meant we could start with a coverage baseline that was better than the usual Renters Policy, including coverage for mold remediation, water damage originating from other peoples’ apartments, etc. We modernized the coverage, getting rid of extra coverage for things like oriental rugs and replacing it with more computer coverage. And critically we lowered the expense base, allowing us to offer huge savings thanks to the compound effect of lower costs and more granular pricing for our target market. The biggest apples-to-apples discount we’ve clocked so far is 71%.
Custom model in hand, we were a critical step closer. But we still couldn’t offer insurance until we had the capital ready to pay claims. Insurance regulation sensibly makes sure that before you sell insurance to anyone you are adequately capitalized to pay your claims. Thus we set out to raise more capital.
Here’s where we ran into Quirk 2, and something I should have known all along. Insurance Claims Capital is inherently not venturable. VCs look for 10-100x return on investment. But, as we grew the company, by law our claims capital reserves would always need to be at least 3x our expected losses (8x is normal). That geometric growth pattern is not scalable. Even if we raised over and over again, by the time we “made it big” we would be selling shares to normal investors, people who value insurance companies on a multiple of their cash on hand. Today’s insurance monsters have grown their cash base slowly for about 100 years. We found it was inefficient for us to own this capital, and therefore impossible to pitch. Money we raise should go to scaling our operations, not sitting around in case we had more claims than premiums. This meant we need to get claims capital partners – aka, rent it from insurance companies.
We had hoped to avoid this because of quirk 3 of the insurance industry, “underwriting profits”. It used to be you could run an insurance business with huge expenses at a loss and make money because interest rates are great. No longer. The way insurance companies make money today is by keeping the difference of premium minus claims and expenses, or “underwriting profit”. This sets insurers up to be in conflict with their customers, and my experience in the industry showed me that if there was a root cause for all the reasons people hated insurance, this was it. So not only would we now need to ask insurers for help, we’d need them to give us their profit back to return to Members.
This process took over a year. Fortunately we secured in-principle approval from a great reinsurer (TransRe) early on. The next step was to find a “primary” insurance partner to get us set up in California. We still have the chart of our emotions on the whiteboard in our office from that time – with huge ups (like when we moved to board-level conversations with one of California’s best cooperative insurers) to huge lows (like when those talks collapsed because that insurer’s agency force wouldn’t allow the channel conflict of a digital partner). Eventually we got it done, inking a three-party deal that worked for everyone, providing a more or less stable return for our capital partners but with the excess profit returning to Goodcover’s Members. With model and insurance capital partners in hand we then moved to get approval from California, which went as well as a process like that could, thank goodness.
While we were working on these business objectives we built the necessary technology to service policies, quotes and maintain regulatory compliance. The easiest way to think of insurance is it works like an append only database. For instance, to remove coverage you would amend a person’s policy contract to remove a coverage, and so similarly in our tech stack we append an event that describes the changes to the policy which outputs a final policy. As a side benefit, this allows us to see the current state of a policy at any point in time. This model works well for us considering most, if not all, of our code is written in a functional language (Scala).
Which brings us up to late last year when we wrote our first Goodcover policy. Honestly it’s been quite an ordeal, but we think the changes we’ve been able to implement have been worth it. I am so thankful for the hard work and persistence of the team, and for all the feedback and help the HN and YC community have given us over the years (shout out to anyone who remembers our “Advice” Show HN from 2018! - https://news.ycombinator.com/item?id=17465114).
Renter’s insurance in CA is just the first step, (home and more states on the way) We have a ton more listening to Members to do, but we hope you enjoy the benefits – and are super grateful for any feedback you have!