Six years on Firestore, then two on MongoDB, and the same thing drove me off both: a schema change in a document store is a running mechanism with no record, no way to trace it, and no way to reverse it. Postgres because it is the developer's database, migrations and deployment built in. This is the reasoning, including the reactive layer you lose and how it was rebuilt in days.
In the second week of the project, before there was much of a product to measure, I built the analytics. Not a feature, a primitive. Page views, request timings and errors look like separate systems and are one shape, so they land in one contract and one table. A rented dashboard can drive a website. It cannot drive a platform.
The first commit was a repository with no product in it, just code rules, shell tooling, testing standards, and an architecture document. Here is all of it verbatim: the six non-negotiables, the anatomy of a file, seven documentation domains and the one left deliberately empty, the four quality gates, and the shell-script template every project shares. Why the way a project builds can be written down before there is anything to build.
A marketing agency is almost pure labor, and payroll is half to two-thirds of what it takes in. Automating the production work, rather than selling another tool, is where AI reaches the industry first. The case in dollars, with sources.