For data stewards
You decide the cases the engine could not decide alone, and keep mastered data correct.
Start here
Where most of this role's day starts.
- Entities, crosswalks, and golden records
- Matching and merging — what the tiers mean
- Review potential matches
Your daily work
| Task | Page |
|---|---|
| Work the review queue | Review potential matches |
| Investigate a record | Read the 360 view |
| Find records | Search for entities |
| Resolve ambiguous links | Resolve relationships by natural key |
| Check quality | Validate and score data quality |
Judging a pair well
Read the breakdown, not just the score. The composite weight tells you how much evidence there is; the per-attribute breakdown tells you what kind, which is what you actually need.
- A pair carried almost entirely by one attribute deserves suspicion, however high the total.
- Several independent attributes each contributing modestly is usually a genuine match.
- A disagreeing identifier deserves more attention than a high total — identifiers do not usually disagree by accident.
Your rejections are valuable
A rejection is a labelled non-match, and hard negatives are the scarcest input to threshold tuning. Rejecting carefully improves the engine.
When the queue never empties
That is a configuration signal, not a workload problem. Persistent volume usually means the clerical review threshold is too low, or an attribute is contributing evidence it should not. Raise it with whoever owns the profile.
Next
Last verified against commit 944761c7 (2026-08-03)