Not really. If you normalize all of your data according to typical relational best practices you will often later find that you need to denormalize it to scale.