Well, here is an interesting case for me that makes looking at NoSQL as a possible solution or adaptation.

Our data is large in volume (10+ million a day), in batch (daily log files), highly relational (OLAP) and is used for analytical reporting (big reads versus fast writes).

The traditional RDBMS handles the volume good these days. Computers are a lot faster, SQL Server is pretty beast and storage is becoming cheap.

Yet, when the data starts becoming so big that it can't easily be handled by a single machine, NoSQL starts becoming an option due to the ease of scaling out. While scaling up is possible, NoSQL seems to ease that pain a great deal.

With the expensive license costs of traditional RDBMS, complex logic required to clean and fit data to rock hard models and the inability to be flexible to changing models, NoSQL becomes a likely option.

However, when it comes to our needs, NoSQL is not good for data analysis with adhoc queries and more. It's not all that perfect when comparing to engines like SQL Server that makes it insanely easy to analyze the data once it's in there for analyzation.

For me, I see both NoSQL and RDBMS serving entirely different purposes that are critical to our business. One as a flexible data lake that is the one source of truth for our data where the other is the BI platform that is refining and ensuring data integrity (which is critical) for analytical reporting.