Relationships determine how Datafor joins tables when a query uses fields from more than one table. An incorrect key or cardinality can duplicate rows, remove valid rows, or produce misleading totals.

Create a relationship

1. Open a table menu and select Join to.
2. Select the other table.
3. Choose the join type.
4. Select the matching field on each side.
5. For a composite key, click Field and add every required field pair.
6. Enable Key values are unique on each side only when that side's selected key combination is unique.
7. Confirm the relationship, then review the edge on the canvas.

Choose the join type

The available join types depend on the current data source. When supported by the connection, the designer can offer:

Join	Use when
Inner join	Only rows with matches on both sides should be returned.
Left join	Every row from the left table must remain, even without a right-side match.
Right join	Every row from the right table must remain, even without a left-side match.
Full join	Rows from both sides must remain when either side has no match.

Choose the table order and join type from the required query result, not only from the physical foreign-key direction.

Set cardinality correctly

The relationship edge derives its 1 and N labels from the two uniqueness settings:

• Key values are unique enabled: the side is labeled 1.
• Key values are unique disabled: the side is labeled N.

For example, a unique customer key joined to repeated customer keys in a sales table is a 1:N relationship.

The designer can check whether a selected field combination is unique in the current source data. Treat a warning or inconclusive result as evidence to investigate; it does not always prevent you from confirming the relationship.

Composite keys

Use multiple field pairs when the business key is only unique as a combination. Keep the fields aligned in the same logical order on both sides. Do not omit tenant, organization, date, or version fields when they are part of the real key.

A pair of table aliases has one relationship record. Put all required field pairs in that relationship instead of creating separate relationships for the same alias pair. Each pair is drawn as an edge on the canvas; deleting one edge removes that field pair, and deleting the final pair removes the relationship.

Edit or delete a relationship

Select an existing relationship to edit its join type, keys, or uniqueness settings. Use the relationship context menu to delete it.

Deleting a relationship takes effect immediately and does not display a confirmation dialog.

After any relationship change, open Model diagnostics and check for:

• Missing tables or fields.
• Disconnected Dimensions or Measure Groups.
• Unexpected cycles. In multi-fact models, inspect competing join paths manually even when no cycle warning is reported.
• Measure Groups that no longer have a relationship path to their Dimensions.
• Calculated measures that reference missing Measures.

Practical checks

• Join fields must represent the same business value and use compatible source types.
• Validate both matched and unmatched records when using an outer join.
• Avoid direct N:N relationships unless they are intentional and tested. Use a bridge when the association needs its own grain or attributes.
• Compare representative totals with a trusted query before publishing the model.

Related topics

• Working with Tables and the Canvas
• Advanced Relationship Modeling
• Tidy Model into a Star Schema
• Model Diagnostics