as a rule of thumb .... in projects which don't need permanent consistency (for example a bank wants to see all transactions happening at the same time so the data they query is correct at the time of the query - a job for SQL)
projects that use complex mixed documents which can contain different data. -> tracking, invoices, basically everything you would use a deep table for in SQL which you recompose to an object in the application.
We can go for CAP as well ... but that goes to far think practical first -> mongoDB is a document-based with eventual-consistency database so obviously document-based applications are the ones you should use it for.
concrete things would be logging, SaaS system where you allow the customer to create custom data sets, e-commerce, tracking, searchable documents -> word to mongodb, online word applications .... such things.
as soon as you need distributed systems with sharding and scale we get to the consistency levels but this to much for now.
does this help? I didn't want to go to much into details