Discovery is not identity binding.
A page can be real, current, and relevant while still failing to bind to the exact person under review.
Enterprise credibility intelligence
Most of what you can find about a person cannot be proven to be about that person.
The internet is not lying to you. It cannot tell you whose record it is holding.
The research engine is live. Enterprise interfaces shown here are a target experience in development, not a claim of deployed functionality.
Live read
Illustrative read, invented data
AM
Avery Monroe
Placement
Recognized Credibility
Confidence
High
18Sources found
12Sources bound
3Sources not read
Relevancesupported
Demonstrated Expertisecorroborated
Reciprocitynot read
Two independent axesEvery placement travels with its confidence band. There is no numeric score about a person.
Evidence
Identity
Confidence
Permitted use
Identity
Confidence
Permitted use
The measured proofFifty governors. The entire population, not a sample.
50Sitting U.S. governors
3,900Source and subject pairs evaluated
43%Bindable to the right person and record
0Figures inferred
The measured problem
Hiring systems find information. They rarely prove whose information it is.
Can this source be bound to this person, in this office, in this jurisdiction, at a date that can be established?
The index shows what it could read.
Confidence measures how much the index could read, not how much the index approves of someone.
Unknown never becomes zero.
When evidence cannot carry a conclusion, the instrument returns a typed refusal instead of manufacturing precision.
The governor study
Even the most public professional records fail to bind more often than they succeed.
43%Bindable. Fewer than half.
This is not a story about governors. It is the ceiling. Director-level candidates have a fraction of that public record.
1,669Bound to the right person, office, jurisdiction, and establishable date.
2,231Found and real, but not bindable to the subject under review.
$97.41Total cost of the study.
0Figures inferred.
The finding is public. The protected classification mechanism underneath it is not disclosed on this site.
The instrument
Two axes. One defensible read.
Placement and confidence answer different questions. A result is incomplete without both.
Axis one: placement
Where the bindable record places the subject, expressed without a person score.
Established Authorityplacement
Recognized Credibilityplacement
Building Credibilityplacement
Emerging Presenceplacement
Insufficient basis to place. A refusal, not a level.
Not started yet. No read attempted.
Axis two: confidence
How much of the record the index could read. This is a statement about evidence, never about the person.
Low
Medium
High
Most restrictive first. A refusal always beats a placement. No level is computed and then hidden.
There is no numeric score about a person anywhere in this product.
Seven independent signals
Evidence is read in dimensions, not collapsed into a verdict.
Signals are read independently of placement. Placement never reads a signal value, and the two do not borrow each other's confidence.
Social Proof
Reachable recognition
How much of what others published about this person is reachable from a property they own.
Consistency
Controlled identity
Whether the same person shows up the same way across the properties they control.
Relevance
Close to the work
Whether the record is close to the work they are being considered for.
Demonstrated Expertise
Inspectable substance
Whether narrowly defined substantive work can be inspected and responsibly attributed.
Authenticity
Specific positions
Whether this person takes positions specific enough to be wrong about.
Reciprocity
Recognition both ways
Whether recognition runs in both directions or only toward them.
Resonance
Carried by others
Whether what they publish is carried further by identifiable others.
An unqualified instrument publishes not read with a reason, never a number and never a blank.supported · corroborated · uncorroborated · not read
The live run view
The client sees the read happen, in the order the instrument actually evaluates it.
Sources found, sources bound, and sources not read remain separate. The surface shows the outcome without disclosing the protected mechanism underneath it.
Read in progressIllustrative intent · invented subject
6Sources found
0Sources bound
2Sources not read
AAuthorized purpose presentpassed
BSubject confirmation presentnot met
CInsufficient basis to placerefusal returned
DPlacement statesnot reached
The refusal engine
The refusal is not an error state. It is the product working.
A downstream system cannot read a hidden answer because no answer exists.
The customer cannot override it.
No setting, plan tier, or support request turns a refusal into a placement.
The subject can resolve it.
The person can confirm their own record. The candidate is a party to the read, not a target of it.
The refusal is auditable.
Reason, evidence state, and timestamp travel with the result.
placement.read · typed refusal
{
"placement": "INSUFFICIENT_BASIS_TO_PLACE",
"level": null,
"confidence": null,
"score": null,
"sources_found": 6,
"sources_bound": 0,
"sources_not_read": 2,
"permitted_use": ["none"],
"receipt": "rcpt_01K7...e3a"
}
Purpose gating is data.
Enterprise delivery
The answer travels with the decision, not beside it.
Authorization, confidence, evidence state, permitted use, and refusal travel with every result.
Model Context Protocol
A typed response for assistants, internal tooling, and automated workflows. Unknown cannot be quietly rounded into zero.
Inside the hiring system
The read attaches to the requisition the team already works, with authorization and subject confirmation carried on the record.
Risk and governance console
Legal, compliance, and talent leadership receive the evidence trail, authorization history, and every refusal made on their behalf.
EngagementAuthorized use established
SubjectRecord confirmation requested
InstrumentEvidence bound and evaluated
ResultPlacement plus confidence, or refusal
Governance by design
The limits are part of the product, not a footnote.
A
Human review remains in the loop.
Employment use is authorized per engagement and disabled by default, and a documented compliance and reinvestigation process governs every read.
B
No subject confirmation means no placement.
The instrument does not preview, cache, or describe a hidden level.
C
Evidence vocabulary is controlled.
Supported, corroborated, uncorroborated, and not read. The index does not call anyone a liar.
D
Every response carries permitted use.
Purpose gating is data.
Before operational damage$127,500Illustrative enterprise hiring economics from market-paid inputs.
The economics
One avoidable failure can carry more cost than the engagement.
Retained search, 30 percent of first-year compensation on $250K$75,000
Recruit and train the replacement, 21 percent of salary and benefits$52,500
A $100K annual contract would break even at eight tenths of one prevented failure a year. Both inputs are prices actually paid in the market. No operational damage is included.
Enterprise access
Bring greater clarity to consequential hiring decisions.
Access is granted per engagement. Employment use requires signed authorization and the subject's own confirmation.