OBJECTIVE: Check how a proposed record link becomes a decision. Use a worksheet or authorized test environment with entirely fictional records. Do not submit invented identities to a live service or copy real identity documents into a public discussion.
1. Define the Expected Distinctions
Create four records and a private answer key for the exercise. Records A and B share a name but describe different fictional people. Record C represents the same fictional person as A, with a corrected spelling. Record D has the same name but insufficient supporting information.
The answer key establishes the synthetic scenario; it is not information the matching process may assume it knows. Document the available fields and the review rule separately.
2. Preserve Candidate Status
Run the comparison and record which pairs are retrieved as candidates. Check whether a downstream screen can distinguish a candidate from a confirmed relationship. An unresolved D should not become confirmed merely because no conflicting value was supplied.
Inspect how blank fields are displayed and explained. A missing value must not be presented to a reviewer as an observed agreement.
3. Follow One Consequence
In the simulation, attach an appointment flag to B. Check whether that flag reaches A before the relationship is confirmed. Record the decision owner and the safeguard that should prevent an unsupported restriction.
Do not assume that separating every record is safe either. A test should also reveal when a genuine correction, such as C, cannot reach the record it belongs to.
4. Dispute and Correct the Link
Mark the A-B relationship disputed. Follow the status through the simulated search view, combined history and decision view. Preserve both source records, remove the unsupported relationship when the review warrants it, and identify any dependent action requiring reconsideration.
Completion criterion: candidate status remains visible, the documented rules handle the synthetic cases as intended, and a disputed link has an accountable correction path. Passing four examples is not an estimate of accuracy across a real population.
Continue with A Match Is Not an Identity and the fictional Shared Name.