Adverse selection inflates premiums, simply because people who need insurance are the ones who buy it.

The point of insurance is to transfer* risk to the insurer. The insurer does that by identifying a group that is homogenous enough that their premiums are just slightly over the payouts.

So an insurer can improve competitiveness by selling multiple products that cover different risk groups. I imagine that for Stripe, the risk variance falls in a fairy narrow band: bounded at the low end by not being worth insuring, and bounded at the high end by merchants losing their account.

* As opposed to say retaining risk, e.g. you don't buy collision on a beater.

> If stripe allowed you to do that - insure only dangerous transactions - then it could not be 0.4% but rather a few %.

This would depend on how well stripe can detect fraud.

And if the price is higher to insure individual transactions, that still would be a nice/valuable service to have.

I would imagine stripe will always block charges it thinks are likely to be disputed. They aren’t offering carte-blanche to accept transactions that are likely to be fraudulent.

Perhaps it wouldn't work for stripe (to allow the merchant pick and choose which transactions use this service). But if it doesn’t it means that the service won’t be used by clients with low fraud rates.