Hey, Jenny!
It really depends on what you're trying to accomplish with your cost model. For us, our main focus right now is on achieving true application TCO, so we're concentrated on getting those application ID tags into our cloud data. The syntax is a bit different between Azure and AWS (AWS tags come in as separate columns in the data while Azure tags are all combined into a single JSON-formatted column), but a few formulas in your transform pipeline can work that out easily enough.
As far as actually getting the tags in the data, we essentially had to bribe and cajole our cloud teams to cooperate. We finally got them to agree that every time they work on an account in AWS or Azure, they make sure the necessary tags are populated. This means our top applications/accounts are tagged right away, but it also means some accounts/applications will take a very long time to be tagged—or will never be tagged. From a cost perspective, however, those accounts that drive the most spend are the ones we work on frequently. So, that's not a huge issue.
Once the data starts flowing in with tags, we just appended the appropriate tables to the cloud master table, and voila! We have cloud costs broken out for the application layer.