Well, GMAC tried to do that and did succeed in becoming a giant mortgage originator, among other things. Back when they began more tan 100 years ago they wanted financing in order to sell cars they otherwise might not sell. That worked astoundingly well. Tesla, with insurance, is probably copying that Sloan-era GM policy. That they have better data, if they can deploy it properly, is unquestionable.
@neroden is prone to hubris regarding actuarial science IMHO, but he's correct that Tesla often is unable to execute on their promises based on their superior data.
I am reminded of Long Term Capital Management again. To wit: Superior data and superior analytics do not guarantee excellent results.
For Tesla there is never a question about superior data nor superior intellect. Their question is always about successful execution.
After all making consistent profits in auto insurance is never a sure thing. Actuarial decisions based on anything other than collision loss severity and insured driving record is fraught. The always reliable postal code classification, included in the Tesla generic application, is a wonderful surrogate for those two factors. Auto insurance actuaries are truly adept in engineering multi-collinearity into their algorithms and very, very few people seem to notice, fewer know what that means and even fewer care. Basically that means that auto insurance rates are generally built on reliable layers of prejudice.
So the question is whether Tesla can actually do better. Maybe, but I'll be surprised if they succeed in any consequential way. OTOH, they may well be able to make a bit of money, and betting against Elon as a disruptor is by no means a sure thing.
A decade ago, I was an employee at GMAC Insurance. All the different GMAC companies were quite distinct though owned by GM. We had our own CEO, for example. GMAC Insurance was arguably the most reliable earnings engine for all of GM and GMAC.
In fact, underwriting profit is rather predictable and stable. What was significantly harder was to actually grow the business. Pricing is highly regulated and very slow to change. What matters more is the flow of new accounts and attrition. If you price some segment too high, for example, you have more difficulty attracting and retaining overpriced customers. So via adverse selection the composition of the book tends to shift where those that were overpriced get replace by those that are relatively underpriced. So loss rates increase to the level you priced at even as the book shrinks as a whole.
The advantage at the time that a carrier like Progressive had was that they had highly segmented and accurately priced rates. This was more robust against adverse selection than what most carriers could pull off. The strategy of showing prospects the rates of competitors was in fact very smart competitively. For example, if a competitor was offering a driver a substantially lower premium, chances were very good that that carrier was underpriced for that risk. By showing this to potential customers one of two things would happen: the driver would go to the competitor which was underpricing the risk and make that competitor less profitable, or the driver would trust Progressive enough to pay a higher premium than they could get elsewhere. In that case, Progressive would be better assured that they had not underpriced the risk and would lock in a more profitable and likely retainable customer. As an insurer you want to avoid rate chasers and attract stable customer willing to pay a fair price. So my view of auto insurance is that it is primarily marketing business. Attract and retain the best customer, and you can do very well financially.
So I think Tesla has several advantages it can exploit to build up a nice little insurance book.
1. Tesla has a very attractive customer base to market insurance to. Very high income and creditworthiness. This is a high mix of preferred.
2. Tesla customer base is also strongly engaged with the Tesla brand, mission, and is very loyal. This is an auto insurance marketers dream prospect base.
3. Tesla owners drive vehicles that are easily mispriced and likely over priced by the competition. Especially the passive safety of AP/FSD is likely to be mispriced by most of the industry.
4. Insurance sales can be integrated seamlessly into Tesla online sales channel. As customer builds car online, Tesla can capture sufficient information to provide insurance quote. This quote gives a shopper confidence that they can find affordable insurance. But then there is the convenience of buying that insurance at the time the order is placed for the car.
5. Tesla NN likely has advanced driver analytics already and can be used to modify and augment driver behavior where it is substandard.
It is probably hard for most people to appreciate just how important item 4 is. Consider that most insurance carriers spend $600 to $1200 in marketing and sales cost to acquire 1 new customer. It is the holy grail of auto insurance marketers to be able to touch prospects when they are actually in the market for buying new insurance. When someone is buying a car, they are definitely in market for insurance. And in fact, insurability questions can make it harder to decide to buy a given car. If you've already got the prospect online and can quote them, this is the absolute cheapest sales channel. Tesla can likely get info on the other vehicles in the household and quote all of them. And it is super convenient for the customer. Remember that you don't care to insure aggressive price sensitive rate shoppers. What you want are the laziest car insurance premium payers out there. This is virtually a "and would you like fries with that" sort of upsell. The most lazy auto insurance customers will simply click the box to add insurance to their order for a new car and click to insure their other car too, all without even calling their current insurance carrier or looking at their deck page. OMG, this is exactly who you want on your insurance book of business. So putting the pieces together, I would suspect an insurance acquisition cost well below $100, and this is huge considering how desirable this customer base is.
I'd also point out to people that the actual underwriting losses in auto insurance is only about 65% of the premium paid. Most of the other expenses beyond the priced actuarial risk is marketing, sale and administrative expenses. It's that other 35% of premium where insurance companies make their profit or loss. (Plus the premiums get invested too for a profit, but that is a separate post.)
Absolutely, Tesla does not need to compete on price for this business. It needs fair pricing, but not price competitive pricing. That is not what this play is about. Rather it about customer relationship marketing, low acquisition cost, and really high value prospects.