I don't get this - my database is going to be normalised to whatever is optimal (3rd normal form generally, denormalised for higher load/sharded/caching)

The indexing is where the main optimisations take place - hashmap indexes, or clustering indexes for priority queues.

What am I missing?