Bronze, Silver, Gold. Three layers that make sense on a slide and fall apart the moment someone asks “where does this transformation belong?”
I see this pattern constantly. A team adopts Bronze/Silver/Gold because Databricks recommends it. They load raw data into Bronze. They clean it into Silver. Then they… clean it more into Gold? The boundary between Silver and Gold turns into a judgment call nobody documented and everyone interprets differently.
The architecture itself isn’t the problem. Layered ingestion with increasing structure - that’s been standard practice since before anyone called it medallion. RAW to staging to conformed dimensions. The concept is sound.
What went wrong is adoption without intent. Teams imported the naming convention without defining what each layer means in their context. What “clean” means for your finance data is different from what it means for your product events. Without explicit contracts at each boundary, you’re just moving data between folders.
I’ve started asking a different question: forget the layer names. For each transformation, who requested it, what does it guarantee, and who’s accountable when it breaks? If you can answer that, the architecture works. If you can’t, adding more layers won’t help.
Who defined what each layer means in your specific context?