MongoDB built its market on 'flexible schema' at a time when relational schemas were rigid. Postgres caught up with JSONB. So when does Mongo still make sense? Narrower than you'd think.
Where Mongo fits
Massive write scale on document-shaped data (>50K writes/sec sustained). Schema that legitimately can't be modeled relationally (sparse, deeply nested, hundreds of optional fields). Sharded multi-region deployments where Mongo Atlas's managed sharding is the operational win.
Where Postgres JSONB wins
Most 'we need a document store' workloads: full ACID, joins available when needed, single operational model. JSONB + GIN index gives you Mongo-equivalent query speed.
Where neither fits
If you find yourself joining across documents constantly: rethink the schema. If you have strict relational invariants (foreign keys, uniqueness): Postgres. If you're at extreme scale needing partition tolerance: Cassandra/DynamoDB.