Skip to content
Monomer

Link models

← Model

Use Links to Model when one model needs a field from another. The stored relationship uses a shared entity name. For example, an orders metric can use customer attributes when orders and customers describe the same customer identity.

Model the relationship

Section titled “Model the relationship”
  1. Confirm what each source row represents and check that the proposed customer key is unique in the customer table.
  2. On the customer model page, select Edit. In Columns, use Row key for the customer identifier. If the model already has a different Row key, use the field’s advanced Use as unique key option for this identifier instead.
  3. Edit the orders model. Keep its own Row key, choose Links to Model for its customer column, and select the customer model’s key in Links to / details. The stored entity name is set to the target key’s name; the source column expression stays unchanged.
  4. Save both models. The shared name must describe the same business identity even if the physical columns have different names. If several models use the same eligible key name, the selector and saved definition show all matching models; the DSL does not store a separate target-model ID.
  5. Read Joins and the grouped Slice by fields on the model page. Select Open map in Calculation path to inspect the relationship. In Explore → Metrics, choose the order metric and inspect the customer group-bys under by.
  6. Run a small comparison and compare the total to an ungrouped run over the same period.

Catalog foreign keys can recommend a relationship when a matching model exists. Accept the recommendation only after confirming its business meaning. If no model has the matching key, the row says No model is keyed by … yet. Create the target model or choose another supported key.

Do not mark a repeating customer key in orders as the Row key just to force a join. A wrong uniqueness claim can change the meaning of an aggregation. If the desired grouping is absent, verify entity names, types, each model’s source table on the selected connector, and whether the relationship supports the requested path.

If grouped totals differ unexpectedly, first check duplicate keys, missing related rows, and filters. Resolve the data meaning before reusing the model in a dashboard.