GIVE.asia recently migrated from MongoDB to Postgresql. We didn't really use any special feature from MongoDB, so it was basically a poor decision.
One of the main problem we have with MongoDB was that there was no schema. When we changed schema (e.g. add column, delete column), we needed to write a migration script to update all the data inside that schema. It was burdensome, so sometimes we didn't do it. This led to a situation where different rows might have different columns.
Postgresql (or mysql and others), on another hand, forces us to handle schema change immediately because the schema change (e.g. add column, remove column) applies to all rows immediately. That leads to cleaner data.
Two situations you might need MongoDB (or other nosql): 1. having an extremely large data that it's impossible to modify all of them when changing schema, or 2. you actually don't care about schema that much (e.g. logs)