As per databricks https://www.databricks.com/glossary/medallion-architecture silver layers typically represent the "enterprise view" with improved quality than bronze (cleansed, deduplicated , augmented ) and mostly has 3NF like normalised data .
The question is, do we really need to make it normalised if my raw data is flat? This will make room for more expensive joins when creating gold tables or materialized views in the gold layer. Analytics is one of the main consumers, and eventually there will be APIs that will serve the data in our use case.Any help will be greatly appreciated ?