How do you guarantee you won't have any bad rows maculating your source of truth?

Your answer is here:
thedatasitter.substack.com/p/netflixs-w...

Nevertheless, the spirit of the division is the same. You'll have your go-to production data at the end - the gold or analytics layer.

These are supposed to be the most trusted, up-to-date, clean, and good looking tables.

AWS names them “raw, staging, and analytics.”

Databricks likes the medallion convention of “bronze, silver, and gold.”

In some scenarios, you would add a layer for events. Some call them “transient,” others “landing.”

I tried and failed to study math for ML for 3 years now, and finally now, after treatment started, I can concentrate on these complex topics and actually learn lol