In analytic workloads MongoDB gets absolutely eviscerated on performance, it is absolutely not optimized for OLAP. Obviously, you've got to go with what you know and it's great that it works for you (though at your data sizes, I'm pretty sure you could also perform quite well by just loading all the data into a vector in memory in your application, or maybe even put it in Excel as another comment mentions :)), but generally speaking this article is aimed at people whose data is already in Postgres or are familiar with it and want to leverage it for OLAP. If that's the case, putting that data in MongoDB and performing worse on basically all fronts isn't a very compelling idea.