Data gets messy, because the real world is messy. Once you start ingesting from other sources, it gets very messy. There is a reason ideas like the "data mesh" or semantic web (distributed schemas) were attempted.
Sure, but that complexity is something you do not want to deal with in your reporting API layer. A reporting API should report correct figures, you don't want to put the responsibility for ensuring the numbers add up on the person requesting the report.
It's already bad enough that people, by default, are not aware of the frame of reference in which they see the world, let alone communicate with.
As just a very simple example, unless I am explicitly clear about speaking from a Living Systems world view and a regenerative paradigm, when I talk about permaculture design here in this forum, I end up talking past the vast majority of people here on HN.
But even staying with the same shared technological world views and paradigms -- the people across the Three Tribes of Programmers (https://josephg.com/blog/3-tribes/) will get into flamewars because the frame differs in distinct ways.
Because a schema makes distinctions on information in a particular way, it will always encode a frame in which to view and understand that information.
This is turning into a rather interesting discussion, but surely it would be better to decide on a schema with each other (or perhaps multiple schemas if absolutely necessary), as opposed to just giving people the ability to query whatever they want and present the results?
If the query itself cannot be explained simply then what does its output even communicate? The more freedom you put in the API the more room you leave for confusion, this might be appropriate for research but not when you're reporting on something.
I agree that you have to make decisions somehow. My points were directed at the phrase “properly modeled”, rather than “very few queries that make sense”.
I see the purpose of reporting within an organization is to allow someone, somewhere to evaluate things and make decisions.
You can have an agreed upon report, but it doesn’t mean that the report itself will always lead to wise decisions. It can also become worse when those very decisions lead to forcing things around you to make things easier to report — that is the core thesis described in that essay, and the book, “Seeing Like a State”
Another example — in the realm of strategic decision making, decisions will always be made with imperfect information. A core way of strategy is deliberately manipulating how the opposing force gathers and interprets information, this influencing their actions. (Example, OODA). So what is already messy gets incredibly messy. This is the kind of stuff that falls way beyond reporting, and not something I think can ever be adequately modeled.
