We have some of the coolest terms in the data space, which can be intimidating. Fortunately, many of them describe what I consider fairly basic things, and “medallion architecture” is one of those terms.
What Is It?
If you haven’t heard of the “medallion architecture” specifically yet, you may have heard others talking about their data in the bronze, silver, gold (and sometimes even platinum!) layer. These are all part of the medallion architecture. The philosophy is you want to put your raw data in the bronze layer, your curated data in the silver layer, and your modeled data in the gold layer. At some point, I may have lost you if you spend the majority of your time in Excel (like I once did!).
If I had to describe it beyond that, it looks something like this:
| Layer | Data Form | Layer Description |
| Bronze | Raw | Unmodified and as is, from the data source |
| Silver | Curated | Validated, errors/nulls resolved, columns renamed, and just generally “cleaned up” |
| Gold | Modeled | Data modified to fit your organization’s modeling method, business logic added, and ready for reporting |
Again, But in English Please!
Back when I was an accountant, when spreadsheets were as fancy as I got with technology, none of this would’ve made sense to me! If that’s you now, this is how I would explain it to you (and my former self).
It’s the end of the month again, so Sally in procurement will be emailing you her spreadsheet with everything we’ve purchased so far that hasn’t been received yet. You save this file to the shared drive from her email. That’s your RAW DATA in the BRONZE LAYER.
She’s super helpful and will usually look up the item codes for you and put them in the sheet too, but sometimes they’re new and don’t exist or it’s called something else in the system or she just forgot to do it this month (it happens, we’re all human). You always make sure to look the file over and validate all the item codes exist because you can’t have any blanks.
You thank Sally again this month and save the file with all your changes into a different file in the same folder in case you made a mistake and don’t want Sally blamed for it. You want to be able to go back to Sally’s file and see it wasn’t there – just in yours. That’s your CURATED DATA in the SILVER LAYER.
Now that you’re sure the data is all correct, it’s time to book your monthly journal entry. You poke and prod and move the data around to fit it into the right format to upload to your favourite accounting system (aka, always the one your company uses), making sure all the headers are named correctly, the columns are in the right order and all your calculations are pristine. You save that copy in a different folder so your manager can review your journal entry and approve it with all the other journal entries you’ve made this month. That’s your MODELED DATA in the GOLD LAYER.
Final Thoughts
And that’s it! We all work with data in one capacity or another, and we like to have copies of our data along the way. It just makes sense to keep data at different checkpoints, in case we ever need to go back and reuse the same data for something different. The medallion architecture is one philosophy on where those checkpoints should be.
I should note that “raw” is generally pretty straightforward, but different data practitioners and organizations can have different ideas of what “curated” and “modeled” mean to them. There is not a clear cut boundary between the two and something one organization considers part of curation, another might consider part of modeling.
At the end of the day, to me, medallion architecture is a fancy term that really just describes a fairly common-sense way of organizing your data.
If you have any questions or want to share your experiences, please feel free to drop a comment or connect with me on LinkedIn. Always happy to chat with other data practitioners!