Configure search and resolution

Use an ADMIN identity to configure resolution. Start with a labeled test set containing known duplicates, incomplete records, and similar subjects that must remain separate. Keep automatic stewardship off while evaluating changes.

Configure in dependency order

Save the resources and select them on the entity. Preserve its source, destination and presentation references. Choose DUPLICATES for manual resolution; AUTO_DUPLICATES also needs an attached steward and automatic mode for scheduled stewardship.

Apply and verify a configuration change

1. Inspect resource validation and dependencies before saving.
2. Synchronize using loadMask:NONE, indexClassificationMask:CHANGES, and sinkMask:NONE when only matching configuration changed.
3. Follow any returned run; no run can mean the configuration was already applied.
4. Search known records and inspect candidate membership.
5. Read classification and pairwise evaluation for known matches and non-matches.
6. Preview a merge and compare its values with the approved precedence policy.
7. Test a separation and confirm a later merge refuses to cross it without an explicit override.

Use the Aurelia decision exercise for a complete partial-selection example. Counts alone do not establish correct matching: more candidates can mean more false positives.

Introduce automatic actions

Test the steward’s filters, condition units and rule order. Inspect its test report before a live run; IGNORE is an actual state change, not a dry-run setting. Start with a small coffeeBreak, follow one run, and review the changed records before broadening the policy. See the steward rollout procedure.

Recover from an incorrect policy

Stop future automatic runs, inspect active work, and restore the reviewed resource definition. Synchronize and repeat the labeled cases. Restoring configuration does not undo decisions already applied; merge undo depends on retained history and its recovery conditions.