Think about setting up indices, even thought not indexing your data is faster on your write path, it is advantageous to do so to prevent your read path from scanning all documents which is very expensive. On the opposite hand, precomputing the search results for all possible queries would make the read path a lot faster, but this is unbounded and would make the write path extremely expensive requiring infinite time and storage. Therefore, a balance needs to be struck between optimizing for the read and write paths, where work can be shifted between this boundary.
This is what a materialized view does