There are many levels in between tracking at the "Amazon" level vs the "Meat:Pork" level. For example I currently track all unprepared food shopping as "Groceries", and would break down "Amazon" shopping into categories such as "Electronics" or "Books".

Ultimately it comes down to how much effort you want to spend categorising spending, but there are many levels of granularity and orthogonal dimensions to slice against

Sure, and my Amazon example wasn't very illustrative of the kind of information that could be gleamed from tracking this at a more granular level. A more concrete example is that my wife and I were flirting with the idea of reducing our meat consumption (for non-financial reasons) and wanted to see what the financial implications would be. Another example is to try to accurately price out past trips we went on: some things normally get filtered under other categories, such as clothes or other products purchased specifically for the trip.

This level of granularity probably isn't worth it to most people, but I found it useful as an exercise to do once because it opened my eyes to insights about my expenses I never would have thought about otherwise. And if there were software that made this super easy to do (<1 minute per entry), I might still be doing it.

You can do a lot with specific and very little with generic entries.

In NL we got the banks to allow exporting [YOUR] data kicking and screaming. Nothing real time sadly but at least it's a thing now.

I would like to force merchants to do the same.

Enforce some protocol that doesn't suck and it should take 0 seconds.

Imagine the things we could build and how many people we could fire.

[delayed]

You don’t need that level of accuracy, you just need “close enough”. Split things out that are important, but if you’re trying to get the sales tax calculated to the cent you’re going way past a level of utility that matters. Eyeball it and move on.