Clean and combine data (silver)
What you'll accomplish: Create a silver dataset — a curated version of one or more bronze tables.
Before you start
- At least one bronze table in your catalog
When you need silver
- Join two bronze tables (for example customers + orders)
- Drop ID columns or rename fields before modeling
- Filter rows with conditions
If a single bronze file is already clean, you can skip silver and publish gold directly.
Steps
Simple promote (one bronze table)
- Open a bronze table in Data.
- Choose Promote to silver (or equivalent action in the UI).
- Optionally drop or rename columns.
- Save with a clear silver name.
Build with a recipe (joins)
- From Data, start Build silver.
- Add bronze sources and define join keys.
- Use Preview to check rows (limited sample).
- Save the silver dataset.
UI paths
| Action | Where |
|---|---|
| Simple promote | Bronze detail → Promote to silver |
| Join recipe | Add data → Create silver dataset (or Build silver from catalog actions) |
Next steps
- Publish data for modeling
- Lab — Python alternative for profiling and charts
- User journeys
Common questions
Silver vs gold?
Silver is internal curation. Gold is validated for ML/Forecast (row counts, target, time series).
Can I skip silver?
Yes — one clean bronze CSV can go straight to gold publish.