Cases the matching rules could not settle arrive here in full context, to be reviewed, decided, and documented.
One review queue
Unresolved cases
Documented decisions
Progress tracking
Accepting every match without review merges distinct entities, and rejecting them all leaves the duplication untouched.
The Stewardship Queue defines where automation stops and human judgment begins, instead of leaving that line to chance.
Gathering cases that need a human decision into a single queue rather than scattering them across processes.
Presenting the similar records and the points where they agree and differ, to support the review decision.
Approving or rejecting a merge, or splitting records that were merged in error.
Routing cases to the data stewards responsible for the relevant data type.
Recording each decision with its owner and date, keeping the reasoning available later.
Following pending and completed case counts and review rates across your organization.
The unresolved case is routed from the matching process into the review queue.
The case is directed to the data steward responsible for that data type.
The records and the points of similarity and difference are examined before a decision is made.
The merge is approved or rejected, and the decision is recorded with its owner and date.
A clear work queue of what needs a decision, instead of hunting for problems in the data.
Visibility into pending cases within their scope and their effect on record quality.
Data that is not merged in error, so a customer does not vanish or fold into another without reason.
Documented evidence that sensitive decisions passed through approved human review.
Two records similar enough to raise a question but not enough to merge automatically.
Trusted sources providing different values for the same attribute, with no clear tiebreaker.
A record missing the fields the matching rule relies on to reach a conclusion.
Data types your organization requires to be reviewed manually regardless of confidence.
They are the person accountable for the quality and integrity of a specific data type within the organization, reviewing cases that need a decision and approving corrections to them.
Because matching works on degrees of likelihood rather than certainty. Cases near the threshold need human judgment informed by business context.
It is a case that did not reach enough confidence to merge automatically, and did not fall low enough to be dismissed, so it waits for a reviewer to decide.
A split becomes necessary to restore the two separate entities. This is why caution in merging is generally preferred over haste.
Because a decision with no recorded reasoning is hard to review or learn a rule from. Documentation turns repeated decisions into better matching rules.
No. A completely empty queue often means rules are too permissive and merging what should not be merged. Cases awaiting review are usually a healthy sign.
A queue holding what the rules cannot settle, with enough context to decide and full documentation of every case.