Core Creek
One threshold, set to the rule

Tighten the match and real deaths slip through, each a cost the book still owes. Loosen it and living people are flagged as dead: a frozen policy, a suspended annuity, a member disenrolled while alive and eligible. One threshold trades one error for the other, and no setting removes both.

Illustrative model·Synthetic data, no real records

The overlap

Two populations, one threshold, two ways to be wrong.

Diagnostic
02The book Regulatory basis RUUPA · state law
match weight, low to high
Two overlapping distributions of match scores. Real deaths mass to the right, coincidental look-alikes to the left. A draggable threshold separates declared matches from declared non-matches, and a shaded band marks the defensible range of thresholds.
Caught deaths Missed deaths Wrongly flagged living Correctly left alone
Shaded: the defensible band. Its loose edge holds the living share of declared deaths within tolerance; its tight edge meets the death-catch duty. The defensible point is the tight edge.
Missed deaths / yr
0
$0
Wrongful determinations / yr
0
$0
Match threshold52 / 100
Drag the line above or this slider. Tighter misses more deaths; looser flags more living people.
Book size1.5M enrollees
Scales the stakes to your book. Deaths track the program's age-weighted mortality for a book this size; look-alikes track an illustrative collision rate. Data quality sets how cleanly the two separate, not how many there are.
Weight on a wrongful flag
Data quality
How cleanly the identifiers separate true deaths from look-alikes. A verified identifier and full date of birth pull them apart; name and birth year alone leave them overlapping. Narrows the overlap, never closes it.
Total annual cost
$0
A U-shaped curve of total annual cost across match thresholds, marking the current threshold, the cost-minimum at the chosen weighting, and the defensible point.
The cost-minimum moves with how heavily a wrongful determination is weighed. Priced at operational cost it sits at the loosest, cheapest match. Priced as the regulation prices it, it moves tighter, until the death-catch duty holds it at the defensible point. The distance between the two is a methodology choice an institution accounts for, not a number a vendor supplies.
Illustrative model, no real records·Same inputs reproduce the same result