How ProfileFinder Confidence Scores Work
A plain-English explanation of how confidence percentages are calculated in ProfileFinder reports.
Confidence scores combine three independent signals:
1. Face-match distance — how geometrically similar the uploaded face is to the matched profile photo using our underlying vision model. Higher similarity = higher confidence.
2. Username pattern overlap — whether the username or display name on the matched profile shares tokens with other known accounts for the same person.
3. Location and bio consistency — whether the matched profile's location, employer, or bio language is consistent with other profiles in the result set.
Each signal is weighted and blended into a single 0–100% confidence value. Scores above 85% indicate a very strong visual match. Scores below 60% are shown as low-confidence hints rather than confirmed matches.
We report false positives at under 2% on clear, frontal, well-lit photos.