I have experience designing and developing both relational and non-relational databases.  The relational databases have been in SQL Server using both event level data and json structured records.  The non-relational database was done in Firebase.  Having both front end software development and data analytics experience, I think holistically and believe in structuring the database so that software records the data in a manner that maximizes the meaningfulness that data analytics can process.  Unique ids should exist for every table and every drop down input from the software.  Typically there is benefit to having source tables that simply contain the data and then scripts that process the source tables and create "mastered" tables that apply business rules to simply the needs for analytic reports/processes to leverage.