NoSQL databases give you the possibility of choosing the best tool for the job. If your data is not relational, or the usage patterns don’t fit the way RDBMS work, you still have plenty to choose from. E.g. wide column Cassandra, document MongoDB, graph OrientDB or key-value Redis.

The amount of unstructured data is growing. Relational databases with their traditional approach of storing normalized data might become a bottleneck for your project and moving to a NoSQL store could be a better approach.

Many NoSQL stores are created with scalability and performance at its roots. Easy sharding and replication gives the ability to easily scale out helping your system with both performance and availability.

Data consistency is achieved through transactions that describe a sequence of operations as a single unit of work. If consistency rules of a database are violated, the entire transaction will be rolled back.

Relational databases have over 50 years of history. The relational model is well-known and understood by software engineers and SQL is well known by data analysts.

Data redundancy is reduced through the process of normalization. This allows to store less data and to have a clearer database schema design.