We should consider sharding when:
I find it interesting that “we start a new project with MongoDB.” is considered an incorrect answer for the quiz. If you know your data set is going to increase very quickly, or you are migrating a large amount of data from another DB or resource why would you not consider sharding from the start? Choosing the correct shard key for the cluster is incredibly important and should be considered when modeling the data, this may be painful if it was not considered during the project began. Also, moving from a replica set to a sharded cluster when the data has grown very large can create network and server slowdowns as the data balances on the shards.