I have never (in twenty years) seen a non-trivial database that is solely accessed by a single application. Given the ORM anti-pattern, many of these databases had no constraints ("we'll do it in the model!") or had only partially defined schema ("it's faster when we don't have to migrate!"), meaning that anyone who just wanted the data had to either load all the application garbage, or reverse engineer the schema.

If you start with your tables, and let the database engine do its job (of course, this is predicated on using a database and not, say, MySQL), querying data becomes a one-banana operation. Of course, there are downsides to this approach, but at scale, all quick and dirty hacks become nightmarish tarasques. DO FEWER HACKS.