This is what my medallion architecture looks like - 

1) Bronze Layer - append raw data.

2) Silver Layer, reflect current(active) data and I do business logic transformations. The Silver layer should serve as your cleaned and transformed staging area. Here, data from the raw layer is sanitized, column names are standardized, and other transformations needed to prepare the data for analytical processing are applied. This layer often doesn't maintain historical changes—it’s more about reflecting the current, cleaned state of incoming data.

3) Gold layer - Moving to the Gold layer with a focus on dimensional modeling and implementing SCD2, your strategy needs to capture and store historical changes for analysis:

Dimension Tables with SCD2: This involves tracking the full history of data changes. Each dimension record should have metadata columns like start_date, end_date, and an is_current flag to indicate whether the record holds the current state.
Fact Tables: These typically store transactional data and should link to the dimension tables via foreign keys.