It depends on what your data looks like. If you've got a collection of JSON documents that gets served verbatim through a JSON API, Mongo is a natural fit. If you've got a domain model with different types of entities and links between them, Mongo is aweful.

The app I'm working on right now evolved from the former to the latter, prompting me to switch from Mongo to Postgres, and it's made the code base much, much simpler. Mongo gets really painful when you have to fetch several documents (serially, because you have to follow the links between them) and join them at the application level before rendering the output.

For that kind of application, SQL with joins and sub-selects is so much better.

The nice thing here is that it's easy enough to write a mongo compatability layer on top of Postgres (hell, it probably already exists...), allowing you the just-get-to-it ease of mongo combined with the substantially superior engineering of Postgres.

https://github.com/JerrySievert/mongolike https://github.com/olegp/pg-mongo/ and my favorite: http://pgre.st/