MongoDB is disproportionately liked by the inexperienced. There are things to like about it, and being able to have sparse secondary b-tree indexes on arbitrary data is, overall, pretty rad. For prototyping, being able to just toss some shit in there, especially when your data comes from an external service and you're not in control of your incoming data (very common these days), it works great for simple use cases and CRUD apps (which is a lion's share of new projects in the industry of the Internet). Being able to predict all the incoming data ahead of time isn't always doable, because a lot of services have underdocumented APIs, and every now and then you'll get data that's just ever-so-slightly different than what you expected. With Mongo, you can at least guarantee you're holding onto that, so that you can use this existing data in the future. That's way less contrived than it sounds. E.g., a third-party changes one of the data formats, adds a field or changes a datatype on one, for instance. If you have a strict schema, you might fail to write some of that data until you update everything. With Mongo, you can at least capture that data and get it working later.

I'm using it on one of my projects, and as I said, there are things to like about it, but graph traversals remain problematic. Overall, I'd say I'm pretty unhappy with MongoDB, and I wish I hadn't chosen it; it doesn't work well for my project. A lot of things I'm doing boil down to performing some kind of graph traversal, which is painful to do at runtime in Mongo (the potential solution space is too large to be precomputed). From what I've seen, MongoDB hasn't been working for me very well with a highly-connected data set.