Check my comment history, I've given several use cases. Basically, Mongo is nice if you don't need a lot of relational stuff but do have lots of arbitrary data to store. A good example is time series data where the format changes over time -- often it's not a good idea to go back and convert old data (sometimes it's not even possible). Mongo makes it really easy to support multiple schemas if your business requires it, rather than having to maintain arbitrary numbers of different tables.