Skip to main content
Help & Documentation

Model several tables together

Import several tables from one connection, then define how they relate in the diagram view - by hand or by auto-detection - and save the model as a shared dataset.

5 min readUpdated October 27, 2026

Real questions rarely live in one table. Golden lets you bring several tables from a connection into one analysis and describe how they join, so you can chart across them without flattening everything into a single export first.

Import several tables

  1. Start an import from a cloud connection - the dialog is titled "Import Schema from" or "Import from" your connection name.
  2. On the "Browse Tables" tab, select the tables you want. Selected tables appear as chips down the left, each showing schema and table name.
  3. Give the import a name in "Import Name", and a description if you want one.
  4. Click "Import Schema" for a live-query connection, or "Import Data" to bring the rows in.
Note: There is no fixed limit on how many tables one import can take. A very large selection shows an advisory warning that loading and analysis may take longer - it still imports - and suggests removing tables you don't need or using Custom SQL to pull in fewer columns.
Tip: If a table or schema is missing from the list, use "Don't see your schema or table?" and click Refresh - the browser is reading a cached catalog.

Define the relationships

Relationships are set in the diagram view under Connections, not in the import wizard. After an import, Golden runs detection over the new tables and draws what it found.

  • Auto-detect relationships - runs detection over the model and adds what it can establish.
  • Add a relationship - opens the "Add Relationship" dialog to define one by hand.
  • Each relationship row has Edit and Delete actions; selecting a link on the canvas and pressing Enter edits it, Delete removes it.

The dialog asks for a Table and Column on each side, an operator, and a cardinality: One to One (1:1), One to Many (1:N), Many to One (N:1), or Many to Many (N:N). An "Auto-detect" button next to cardinality works it out from the data instead of asking you to know.

Note: Hovering a relationship tells you where it came from - "Detected from warehouse FK metadata" or "Added manually by the user". A foreign key your warehouse already declares is stronger evidence than a name that happens to match.

Adding more data to a model

The diagram has an Add button for bringing in more tables, offering "Upload file" and "Google Sheets" for file-based datasets. You can also drop a file straight onto the canvas - it reads "Drop anywhere to add as a new table". A counter shows how many tables the model holds.

Save the model for your team

Once the model is right, "Save dataset" promotes it into a shared dataset - "Save this data model as a shared dataset for your team to reuse". The dialog asks for a Name, a Space to save into, and optional Description and Tags, and shows a Dataset Summary with the row and field counts before you commit.

Tip: Saving the model, rather than each person re-picking tables and re-declaring joins, is what stops two teammates quietly analyzing two different joins of the same data.