Understanding confidence scores

What the High, Medium, and Low bands mean and how the score is calculated.


Every duplicate pair gets a confidence score from 0 to 100. The score tells you how sure DeDupe is that two records are the same.

Confidence bands

Scores map to three colored bands shown throughout the queue:

| Band | Color | What it means | | --- | --- | --- | | High | Green | Very likely the same record — usually safe to merge. | | Medium | Amber | Probably the same — worth a quick look. | | Low | Orange | Possible match — review carefully before merging. |

The Data Quality page also breaks pairs into finer buckets: Low (under 50%), Medium (50–69%), High (70–89%), and Very high (90–100%).

How the score is calculated

The score blends four weighted factors:

  • Match score — how strongly the matching strategy fired.
  • Method reliability — how trustworthy that strategy is in general.
  • Attribute overlap — how much other data the two records share.
  • Data quality — how complete and clean the underlying records are.

Admins can adjust how much each factor counts toward the final score in the Detection cockpit — see Tune detection. The four weights always add up to 100%.

Using confidence to work efficiently

  • Filter the queue by minimum confidence to focus on the safest merges first.
  • Set an auto-merge threshold so pairs above a score you trust merge without manual review.
  • Drop to the lower bands only when you have time to inspect each pair closely.