Is that true? Polling the API might be too much, but I would expect Stripe to be able to handle a few GETs for every payment they take, even big companies will probably only be taking ~1-10k payments/second, doubling or tripling that isn't much. And ingesting webhooks for payment is necessary to be able to perform access control or whatever is necessary for the business to function. Conversely, $0.03 of margin to patch over bad engineering is quite a lot of margin for some types of business.

Lower tech businesses perhaps, but most of them probably aren't doing data analysis anyway.

Yes, it’s true. For a hint of some of the utility and complexity, consider backfilling historical data; or, how you’d ensure consistent views (as of time X) of historic data (all charges before time X) when each row in the dataset is mutable.

can't you just design your database so as to be able to do that independently ?

You’ve got plenty of paid options with Stripe, but here’s the catch: trying to match what you see in the Stripe dashboard with exact cent accuracy using queries is a total headache. in addition, it would be simpler to add an option to pull that data via an API call.