The second layer you saw in the architecture is the data warehouse.
The Garbage Must be Put Away.
Dirty Data Is Not Suitable For Analysis! Do You Agree?
The Data Source May Be Directly From A Data Warehouse (Rather Than OLTP).
The advantage of this type of data is that it is clear from the beginning and does not need to be cleaned.
Do you agree that data collected by humans for years may not be clean?
And do you agree that analysis of dirty data makes the output of the final model unclean?
So they first need to be cleaned up using specialized ETL tools (including SSIS, which is similar to SSAS, another Microsoft Business Intelligence Service) and put in place called a data warehouse (DW).