Medallion Architecture Disdain
Medallion Architecture Disdain argues that Bronze/Silver/Gold layer naming describes processing stage, not content — call it the Paint Department instead, a name that tells you what happens there.
Nobody at Home Depot organizes by "Bronze aisle." They organize by Paint, Tools, and Lumber — because customers need to know what's inside, not how refined it is.
The problem hiding behind a tidy diagram
Bronze, Silver, and Gold sounds sophisticated on an architecture slide. It falls apart the moment a new analyst asks where customer data lives and gets sent to the Gold layer, where fourteen different tables have "customer" somewhere in the name and three different colleagues give three different answers about which one is actually correct. It turns out the real customer table lives in Silver, because Gold was never finished — a detail nobody wrote down anywhere, because the naming convention never asked anyone to.
Metal-tier names describe processing stage: how refined the data is. They say nothing about content: what the data actually is. That distinction sounds academic until you watch a new hire lose their first two weeks just learning which layer holds what, purely because the labels were optimized to look organized in a diagram rather than to help a human find something.
Name it like a hardware store, not a metallurgy chart
Replace processing-tier names with content-based ones. Raw_SourceSystemName for data exactly as it arrived — Raw_Salesforce, Raw_Mainframe. Curated_Domain for data that's cleaned and usable but still tied to one system — Curated_Claims, Curated_Calls. Integrated_Topic for data that's cross-system and business-ready — Integrated_Customer, Integrated_Lending. Every name now answers "what's in here" on its own, without a meeting.
The test for whether this is working: ask five analysts where they'd look for a specific kind of data, with no documentation open. If they can guess correctly from the name alone, the naming succeeded. If they need to ask someone, it failed — no matter how elegant the underlying architecture is.
It's a rename, not a rebuild
This is the quietly appealing part: the underlying architecture usually does not need to change at all. Same pipelines, same processing logic, same infrastructure — just names that tell a person what they will find instead of how far along the pipeline it traveled to get there. Teams that make this switch report onboarding time and "where's this data" support tickets both dropping substantially, without touching a single ETL job.
Where to start
Audit whatever your current layer or zone names are, and be honest about whether a new hire could guess what's inside each one. If the answer is no, pick a content-based convention, apply it to your most-used tables first, and treat any newly built table that doesn't follow it as a code review rejection.
Read this in the book
The Auction Block →Where do you already do this — and where don’t you?
The Analytics Leadership Circle assessment maps your pattern across the same dimensions these frameworks live in. Free, 5 minutes, no account required.
Take the Assessment →