Understanding the confidence score

What the number is

Every Wealth Recon dossier ships with a single confidence number from zero to one hundred at the top of the At a Glance page. The number drives one decision: at or above eighty, the credit consumes and the dossier ships normally; below eighty, the credit refunds automatically, the dossier still ships, and a Low Signal banner names the gap.

What goes into the number

The composite is a weighted average of four signals.

• Coverage (35 percent). What fraction of the fifteen dossier sections carry substantive verified content. A section counts as substantive when it carries at least one verified claim that meets the substantive threshold (a specific named fact or a definitive negative finding). Sections marked Not Applicable are excluded from both numerator and denominator.
• Citation density (25 percent). Average number of distinct verified sources per scored section. The deny-list filter strips press-release aggregators and search-engine-optimized profile-data sites; only the cleaner sources count toward density.
• Identification certainty (25 percent). How strong the disambiguation outcome was, plus contradiction checks across foundational facts (current or most recent employer, age within five years, marital status, primary city of residence).
• Net worth band tightness (15 percent). How precise the net-worth band is relative to its midpoint. A range of fifty million to seventy million dollars has a tightness ratio of 0.33; tighter is better.

A wrong-subject dossier scores at fifty or below regardless of every other signal because every claim verifies against a real source on a real person, just the wrong one. The weighting refuses to let that case slip past the eighty threshold.

What the score does NOT mean

The score is not a guarantee. A dossier at confidence ninety-five is not promising perfection; it is reporting that the four signals all landed strongly on this subject. The advisor still owns the judgment call on every claim.

The score is not a comparison across subjects. A confidence-ninety-two dossier on a thin-footprint subject and a confidence-ninety-two dossier on a dense-footprint subject convey different things. The thin-footprint subject's score reflects strong identification certainty plus tight band on a small public record; the dense subject's score reflects coverage and citation density on a larger public record. Compare each dossier against itself, not across the corpus.

What the Low Signal banner means

When the score lands below eighty, the credit refunds automatically. The dossier still ships, and the Low Signal banner appears at the top of the At a Glance page:

Low Signal Dossier. This subject's public footprint produced a confidence score below our eighty threshold. We have shipped what we could verify and refunded the credit. The sections below show what the agent found and what it could not. The Source Manifest at the back lists every Uniform Resource Locator behind every claim. Use this dossier as a starting point, not a finished picture.

Low Signal does not mean "we failed." Some subjects produce Low Signal dossiers no matter how the engine is tuned: thin-footprint subjects, dynasty-trust beneficiaries, non-United States nationals, sealed estates. The refund acknowledges the constraint; the dossier still ships because partial verified information is more useful than no information.

What you can do to improve a Low Signal dossier

Three actions help most often.

• Add optional fields and re-run. Running the same name with a LinkedIn Uniform Resource Locator, an alma mater, and a likely nickname narrows the disambiguation pass and often lifts coverage by surfacing a different candidate or by tightening the search.
• Edit any incorrect facts inline. Your edit informs every future dossier in your firm on the same prospect. Over time, advisor-corrected facts compound into a higher confidence on repeat-subject dossiers.
• Wait for an incremental refresh. A subject with a thin footprint today may surface a triggering event in the next thirty days. Re-running after the cache window expires picks up the new signal.

[CTA: Read the full methodology]

End of help article.