The function of your company would be similar to that of the traditional insurer except all of the premiums go to the pool for claims and admin instead of also having to share it with shareholders/investors.
What I'm curious is what happens when claims are more than what's already in the pool? My assumption would be that premiums would have to go up to cover the loss so you're banking on the fact that by group selection, claims will be less than the average. And in the event that claims are more than the average, there's probably insurance taken out on that.
> leveraging the fact that good drivers tend to know good drivers
I think humans might overestimate our ability to pick 'good' drivers. I wouldn't trust it very far... but it also doesn't matter much.
What matters is badly underpriced drivers, for example your mate who drinks and drives but doesn't yet have a conviction for it. Don't invite him. Everyone else is broadly fine.
> ...there could be a social pressure to not affect the group's members
That's a factor we're expecting to be able to measure, but it's not part of the business model per se.
> The function of your company would be similar to that of the traditional insurer except all of the premiums go to the pool for claims and admin instead of also having to share it with shareholders/investors.
Yeah, broadly speaking. Guevara's revenue stream is a flat fee, compared to a conventional insurer whose income is your premiums... so Guevara isn't motivated to inflate your premium or to look for technicalities when handling claims.
> when claims are more than what's already in the pool?
The UK motor insurance market requires policies to have uncapped liability, so you always need reinsurance. Even when you're insured through Admiral et al the larger claims are farmed out to Munich Re, Swiss Re, Trans Re, etc.
> ... premiums would have to go up to cover the loss...
To an extent, yeah. You do get price volatility, but we cap it at 100% of your (competitive) Year 1 premium. Your premium can drop as low as 25-30% of your typical market rate in a good year, and averages out costing about 50%. You get that huge discount in exchange for taking that volatility.
That discount is effectively the underwriting profit that would normally go to the insurance company.
> ...so you're banking on the fact that by group selection, claims will be less than the average.
We're not banking on group selection being meaningful to our claims rate. We hope for lower claims rates, and lower average claim size, but we'll be very successful without those things because it's simply a better business model. There's much less moral hazard, lower susceptibility to fraud, less information asymmetry... but more importantly (a) as you build up capital your premium drops, so you get a lower price and we get a higher LTV through reduced churn, and (b) we have a lower CAC compared to conventional insurers. I know I'm biased but I think it's a strong model even before any of the behavioural benefits.
I should add that I have now stepped back from Guevara (I can't commit enough time for family reasons), so my knowledge of product pricing, policy, and the rest of this stuff is dating rapidly.
It doesn't seem to me that customers would see full value of their "purchase" if they have never had a claim. Yet, it's crazy that their premiums go up because someone else had a major accident. On the flip side, insurers can't really choose to raise their rates like a normal business would do. For me, this is the fascinating part about insurance businesses.
What is curious to me is if there is a way to build a risk model that assumes that people will get into accidents by default? You've mentioned that people overestimate our ability to pick good drivers. Why not build a model that assumes people are bad drivers by default and optimize from there?
If you think about Guevara's insurance as capital expenditure (by describing it as a purchase) then, sure, you're right, losing some of it because of someone else's accident triggers our human aversion to loss.
But, the trick there is that while Guevara's insurance can be described as capital expenditure, conventional insurance is always operational expenditure. With conventional insurance you're not buying your insurance, you're only renting it. At the end of the year you have nothing.
> Why not build a model that assumes people are bad drivers by default and optimize from there?
I think Guevara's model is as close to that as it's possible to go. By moving an acceptable degree of the financial repurcussions for accidents onto you and your social group, Guevara is stepping back from needing to care directly about who is and who isn't a good driver.
For example, if there were a way let a group choose their own prices, or pay flat prices, then (in theory at least) Guevara would be able to entertain that idea. Pricing matters for Guevara in terms of customer acquisition, and in terms of making things feel fair for your group-mates, but other than that it's not a big issue.
Insurers care about pricing so much they're putting tracking devices in peoples cars to understand their driving patterns. They need to get more and more and more data because they're ultimately trying to predict the future. Guevara doesn't need to do that.
Totally, it's EPIC, EPIC, EPIC, moral hazard. The hazardiest of hazards. It's toxic; all of the mistrust and information asymmetry flows out it.
If insurers weren't constantly nudging prices there'd be no need to switch car insurer so often (45% of the UK motor market churns per year) and the price comparison sites probably wouldn't need to have existed.
If people could trust insurers to handle claims fairly then we wouldn't need to spend as much money managing complaints and government oversight.
Aside from a high regulatory barrier and the significant capital investment required, another reason the insurance industry has resisted serious disruption for so long is that a lot of the way things work are hidden behind closed doors, jargon, based on who you know in the industry.
In the UK at least, even the pseudo-standard industry APIs they use (based on EDI, EDIFACT to be specific) are hidden behind a licencing organisation (Polaris) owned by the ABI (the industry body) that charges per transaction. It's a cartel.
If it were in the future I'd describe it as dystopian.
In the US they use X12, which is an XML-based version of EDI, which was published publicly by ANSI at least.
> In today's peer-to-peer models, insurance policyholders form small groups online. A part of the insurance premiums paid flow into a group fund, the other part to the insurer.
In a broker-based model this might be true, but you may also be regulated as an insurer, or have a commercial relationship with a regulated insurer.
> For claims above the deductible limit the regular insurer is called upon.
Or the re-insurer, depending on whether the business deals with the underwriters or goes directly to re-insurance.
> The only requirement is that all group members must have the same type of insurance
Not necessarily. In fact ideally not. There are significant savings to be made by leveraging an existing float rather than having separate pots, and the social density of your group becomes sparse if you can only insure your ferret with other ferret fanciers.
> The providers are financed through brokerage commissions of insurance companies
In some models. Others not so much. As discussed above, dealing with the underwriters isn't the only way to run this kind of business.
Also "peer-to-peer" is a misleading name in some cases. Crowdsourced might fit better.