An official website of the Disclosure Foundation
RecordsObservablesOriginsClaimsScoringOnboardDocs

Records

The unit of record

Observables

Detection

Origins

Classification

Claims

Attribution

Scoring

Evaluation

Onboard

Bring your dataset

Docs

Implement the standard

GitHub

Open standards and tools for UAP research.

RecordsObservablesOriginsClaimsScoringOnboardDocsGitHubDonate

© 2026 Disclosure Foundation. All rights reserved.

The Standard · Part 5 of 5 — Evaluation

Scoring

Two separate questions, scored separately: how completely is a case documented, and how compelling is what the evidence shows? Opinion lives here — never in the record.

@disclosureos/scoringDocumentation

Two separate questions

Scoring deliberately answers two questions that are usually conflated, and never lets one masquerade as the other. A thin record of something extraordinary and an exhaustive record of something mundane should both be visible for exactly what they are.

Completeness

How well-documented is this case?

Measures how much of a record is actually filled in, against every field the standard defines. Add a field to the standard and the measure extends automatically.

Compellingness

How anomalous is what the evidence shows?

Blends the technology, biologics, and origin claims on a record into a weighted consensus — with an honest range when evaluators disagree.

Completeness is measured against the standard itself

There is no hand-maintained checklist. The measure is derived from the standard itself: every individual field a record can hold — 226 fields today, 12 of them required — checked for whether it actually carries a value. Add a field to the standard and the measure extends automatically. The fields that come back empty are the point: the missing list is the disclosure gap, named field by field.

  • temporal

    40 fields · 2 required

  • location

    30 fields · 6 required

  • provenance

    27 fields

  • sourceData

    18 fields

  • objectCharacteristics

    12 fields

  • featuredMedia

    11 fields

  • movement

    11 fields

  • aviation

    10 fields

  • identifiers

    10 fields

  • environment

    9 fields

  • investigation

    9 fields

  • relations

    7 fields

  • witnesses

    7 fields

  • responseImpact

    6 fields

  • sensorEvidence

    4 fields

  • createdAt

    1 field · 1 required

  • dataSourceId

    1 field

  • description

    1 field

  • documents

    1 field

  • eventType

    1 field

  • extensions

    1 field

  • id

    1 field · 1 required

  • internalNotes

    1 field

  • media

    1 field

  • physicalEvidence

    1 field

  • schemaVersion

    1 field

  • status

    1 field · 1 required

  • summary

    1 field

  • testimony

    1 field

  • updatedAt

    1 field · 1 required

Compellingness blends three signals

The second axis reads the claims attached to a record — never the raw facts directly — and reduces them to one weighted consensus. Each signal comes from an earlier part of the standard, and the default weights are published numbers, not hidden settings: anyone can apply their own, and every result records which weights produced it.

Technology observables

45%

Weight of the technology-anomaly signal.

Part 2 — Observables

Biologics observables

35%

Weight of the biologics-anomaly signal.

Part 2 — Observables

Origin claims

20%

Weight of the (non-prosaic) origin-confidence signal.

Part 3 — Origins

Observable claims — anomaly strength

Each observable claim contributes its evidentiary tier’s weight, multiplied by the claim’s confidence. The ladder is the same one from Part 2: testimony alone moves the score a quarter as far as independently confirmed sensor data.

  • Not Indicated0
  • Reported0.25
  • Documented0.5
  • Measured0.75
  • Confirmed1

Origin claims — the prosaic gate

An origin claim contributes its confidence in a non-prosaic explanation. A confident classification under the prosaic branch of the Part 3 tree means the case is explained — so it contributes zero anomaly signal, no matter how dramatic the testimony reads.

  • 1.1.1.*Prosaic→ 0

    Ordinary or conventional explanations for observed phenomena.

The prefix set is overridable per call — a stricter policy can treat more of the tree as explained.

Disagreement is reported, not erased

Because a record keeps every claim as a list (Part 4), the scorer never has to pick a winner. It reports the consensus, the honest spread across competing claims, and whether evaluators disagree about whether anything anomalous happened at all — in one result.

  • score

    Overall compellingness (consensus point estimate), 0–1.

  • range

    Spread of the score across competing claims — wide range signals disagreement. Bounds are computed per-signal under the same strongest-anomaly (max) reduction as the point estimate, not a global min/max over every claim.

  • contested

    True when claims disagree on *direction* (some assert anomaly, some assert none).

  • components

    Per-signal consensus contributions before weighting.

  • weights

    Relative weights for the compellingness components. Need not sum to 1 — normalized internally.

  • scoringVersion

    Methodology version that produced this score.

Who gets the most weight?

Records hold facts and attributed claims; scoring is where trust enters. The consensus score accepts an evaluatorWeight setting that controls how much each evaluator’s assessment counts toward the final number (all equal by default). An organization can give official or verified evaluators more weight without changing the method. The range and the contested flag are deliberately untouched by trust: they always report what evaluators actually said. Different organizations can apply their own policy to the same records; the standard ships one transparent default, and every part of it can be adjusted.

One case, scored live

The record below is USS Nimitz “Tic Tac” (2004) — the standard’s golden-path example, carrying the claims from Parts 2 and 3. Every number in this section is computed by @disclosureos/scoring when this page builds. Nothing here is typed in, so if the methodology changes, this section changes with it.

Inputs — the claims on the record

  • Instantaneous AccelerationConfirmed85%

    Radar tracked descent from ~80,000 ft to sea level in seconds.

  • Transmedium TravelDocumented60%

    Subsurface radar returns reported on departure.

  • 1.1.3Extraterrestrial40%

    Performance exceeds known aerospace capability; cannot exclude advanced terrestrial program.

    1.1.1.2.1Governmentalalternative25%

Completeness

10%23 of 226 paths · 100% of required

Even one of the best-known cases on record fills a fraction of the schema. That gap is the point: completeness quantifies what disclosure would actually take.

As published — evaluators agree

not contested

Every claim asserts some anomaly, so the consensus is tight and the range collapses to a point.

0.46range 0.46–0.46
technology
0.85
biologics
0.00
origin
0.40

Now add one dissenting claim

Instantaneous AccelerationNot Indicated70%

Apparent acceleration is consistent with a radar-processing artifact.

With the dissent — evaluators disagree

contested

One evaluator asserts no anomaly. The point drops, the range widens to show both readings, and the case is flagged contested.

0.29range 0.21–0.46
technology
0.47
biologics
0.00
origin
0.40

The standard, end to end

That completes the journey: a record of facts, the observables it may exhibit, the explanations it may have, the claims that attach each assessment to its evidence, and the scoring layer that turns it all into honest numbers. To put it to work, start with the documentation.

Part 4 of 5ClaimsAttributionReady to buildDocumentationImplement the standard