Accuracy & methodology
A wrong heir costs more
than a missing one.
That single principle decides most of how Heir Lens behaves. A phantom heir dilutes everyone's share, sends you after the wrong person, and quietly makes the whole list untrustworthy. A missing one is a gap you can see and fill. So when the evidence is ambiguous, we flag rather than guess.
Conservative by default
A same-surname relative with no obituary and no recorded parent-child code is genuinely ambiguous — sometimes a real child, sometimes a cousin. Rather than guess, Heir Lens marks them 'possible heir, verify' and gives them no share. You see the candidate; the math doesn't pretend to be sure.
Identity before relationship
Most bad heir lists come from the wrong person with the right name. Before an obituary can confirm a death or a record can add a child, it has to clear identity gates: middle initial, age agreement, location, and whether the family it names overlaps the family we already know.
Same address, same answer
Web search and AI extraction are not reproducible by nature — ask twice, get two answers. Heir Lens freezes the evidence it gathered for a property the first time and replays it on every subsequent run, so a lead doesn't change under you between sessions.
Unsolvable is an answer
Some properties genuinely can't be resolved from public records. Those get an explicit 'needs human deep dive' flag rather than a silent zero, because an empty list that looks like a finished list is the most expensive output we could give you.
Global rules, never one-offs
When we fix a mistake, we fix the rule that caused it and re-test every benchmark property. No per-address patches — a fix that only helps the property it was written for is how systems quietly rot.
Every heir is sourced
Each person on the list carries the reason they're there: the obituary line, the record code, or the inference — plus any caveat. You can audit a result in about a minute, which means you can also disagree with it.
What counts as proof
Not all evidence is equal, and the ranking matters more than any single source. Heir Lens applies a fixed hierarchy so the same conflict always resolves the same way.
| Source | Authoritative for | Never used for |
|---|---|---|
| Obituary | Relationships — who is a child, spouse, sibling, or grandchild | Proving identity on its own; a same-name stranger's obituary proves nothing |
| Public records | Identity, age, vital status, addresses, recorded parent-child codes | Inferring relationships from shared surnames or age gaps alone |
| Parcel / deed | Who legally owns the property and since when | Anything about family structure |
| Memorial sites | Corroborating a death when no obituary is found | Establishing survivors or shares |
| AI extraction | Turning prose into structured survivor lines | Deciding kinship — every AI-proposed relationship passes deterministic gates before it counts |
The practical version: an obituary decides how people are related, a public record decides who a person is, and neither one alone is allowed to invent an heir.
How we measure ourselves
Accuracy claims are only worth the benchmark behind them, so ours is fixed, external, and graded by a human.
A fixed property set
Fifty real properties with heir lists confirmed by a human researcher. The set doesn't change to flatter the software, and ground truth is supplied by the analyst, not by us.
Two different measurements
We grade a property twice: as a brand-new address researched live, and as a repeat run of an address already in the system. First-run accuracy is the honest number for a new customer; repeat-run consistency proves results don't drift.
Every miss classified
Each incorrect result is sorted into a cause — retrieval, identity, relationship logic, or contested ground truth — and only the classes we can fix with a general rule become engineering work.
Placeholder content. The dashes below are blank on purpose, not broken. Heir Lens is graded continuously against 50 real properties whose correct heir lists were confirmed by a human researcher. We haven't published the percentages yet because a brand-new address and a re-run of a known address score differently — and we'd rather publish the honest number than the flattering one. The figures go in as soon as we sign off on which measure to show.
50
Benchmark properties
—
First-run accuracy
—
Repeat-run consistency
What Heir Lens does not claim
- We do not certify heirship. Only a court can do that.
- We do not read wills. Intestate succession assumes no valid will exists — if there is one, it overrides everything Heir Lens computes.
- We are not a title search and do not surface liens, mortgages, or competing claims.
- Coverage depends on public data. Where obituaries were never published or records are thin, the honest answer is a flag, and that's what you'll get.
- Pennsylvania, Texas, Florida, Georgia, North Carolina, Ohio, and Washington have fully modeled succession engines; other states currently use a general per-stirpes model and should be treated as an estimate.
Heir Lens produces research estimates from public records. It is not legal advice and is not a substitute for a title search, probate filing, or an attorney's opinion.
Questions about a specific result? Send us the address — we'll walk you through exactly what the research found and why.
Judge it on your own market
Run addresses you already know the answer to. That's the only benchmark that counts.