Make many-to-many expansion explicit in evidence tables — what should the operator record?
Some operational tables legitimately contain several references per source object and several objects per reference. A many-to-many join can expose those links, but its expanded row count is easy to misread. Define relationship grain and preserve both parent keys before using the output in another table.
Represent link relationships without presenting expanded rows as additional independent sources.
State that one output row represents a left_key/right_key pair, not a new person, post or independent observation. For link relationships, preserve left_source_key and right_reference_key separately; concatenating their display labels can create collisions or imply identity between fundamentally different objects.
Name the table relationship-view and include the join rule, preventing downstream users from treating its row count as the size of the researched population. A useful heading says returned link relationships, not unique researchers or total accounts; table naming can prevent later summaries from promoting representation size into a population claim.Declare relationship grain
State that one output row represents a left_key/right_key pair, not a new person, post or independent observation. For link relationships, preserve left_source_key and right_reference_key separately; concatenating their display labels can create collisions or imply identity between fundamentally different objects.
Define pair uniqueness
Choose whether repeated identical pairs are distinct observations or duplicate links, documenting any occurrence identifier used to distinguish them. If repeated pair occurrences are meaningful, add an actual observation reference rather than inventing event times or assuming repeated rows represent independent source activity.
Test expansion fixtures
Use a small labeled relationship example with multiple matches on both sides and verify the expected unique pairs and retained parent keys. Include a fixture where unrelated arrays would multiply accidentally, verifying that the declared relationship field limits expansion to evidence-backed pairs rather than every possible combination.
Reconcile parent coverage
Report distinct left keys and right keys separately from relationship rows, inspecting unexpected Cartesian expansion or missing links. The validation artifact should expose pair duplicates and parent coverage together, since deduplicating relationship rows can otherwise conceal missing or incorrectly associated parent objects.
Deliver interpretable output
Name the table relationship-view and include the join rule, preventing downstream users from treating its row count as the size of the researched population. A useful heading says returned link relationships, not unique researchers or total accounts; table naming can prevent later summaries from promoting representation size into a population claim.
Keep the boundaries clear.
- Represent link relationships without presenting expanded rows as additional independent sources.
- Operational record: Declare relationship grain. Context badges do not certify access.
Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.
Primary references
Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.