RedHub AI
How the roll-up gate works
How the Entity Identity & Functional-Key gate works
This tool grades candidate duplicate-record pairs against five weighted identity signals. It reads 8 pairs and rolls them into one verdict — the worst, never the average — then a gate that can only make it worse.
The 8 pairs
A disqualifier - two different values for an identifier that's unique by definition, or a fact the two records can't both hold - only fires on a pair the evidence already reads SAME ENTITY at 60 or above. But when it fires, that pair becomes IDENTITY CONFLICT and blocks the whole sweep, no matter how high the mean evidence score runs.
Why: Three pairs read DISTINCT and three read SAME ENTITY - work waiting to happen, not a problem. But P-04 and P-01 both clear the 60-point evidence threshold and each fires a disqualifier: a functional key that's unique by definition, and a fact the two records can't both assert. Mean evidence across all eight pairs sits at 68/100, but the sweep reads the worst pair, not the mean, so it's IDENTITY CONFLICTED.
No single pair is the crisis — the clustering is. The roll-up takes the worst, the gate escalates when trouble compounds, and it names the one thing to fix first. Grades the record pairs you paste in. Not legal or compliance advice, and it merges nothing on its own.