Evidence · Accountability
Track Record
How the public record is actually performing: what has been published, how often grades changed, what was corrected or withdrawn, and where reviews are overdue. Every figure is counted from the record itself when this page loads.
The record is just starting.
There is not yet enough published material for these indicators to say anything. The definitions below show what will be measured; values appear as records are published.
Evidence Cards
| Indicator | Value | Definition | Counted from |
|---|---|---|---|
| Evidence Cards in force | No data yet | Public Evidence Cards whose current version has been approved by a named person and that are neither withdrawn nor superseded. | evidence_cards (status published or under reassessment) |
| Cards in force carrying a formal grade | No data yet | Cards in force whose current version carries a grade. A grade can only be issued under an adopted version of the Evidence Standard after at least two independent assessments. | evidence_card_versions.grade of each card's current version |
| Cards under reassessment | No data yet | Cards whose grade is being reconsidered because new evidence or a challenge was accepted for review. | evidence_cards (status under reassessment) |
| Cards past their scheduled review date | No data yet | Cards in force whose next scheduled review date has passed without a new approved version. Shown so that lapses in currency are visible. | evidence_cards.next_review_due compared with today's date (UTC) |
| Cards withdrawn or superseded | No data yet | Cards taken out of force. They remain on the public record with their full history. | evidence_cards (status withdrawn or superseded) |
Claim Checks
| Indicator | Value | Definition | Counted from |
|---|---|---|---|
| Claim Checks published | No data yet | Claim Checks currently published with a verdict and a named reviewer. | claim_checks (status published) |
| Claim Checks withdrawn | No data yet | Claim Checks withdrawn after publication. They remain on the public record, labelled. | claim_checks (status withdrawn) |
Corrections
| Indicator | Value | Definition | Counted from |
|---|---|---|---|
| Corrections published | No data yet | Entries in the public Corrections Ledger, of any severity, across Evidence Cards, Claim Checks, articles and the Evidence Standard. | evidence_corrections (published) |
| Grade changes | No data yet | Ledger entries recording that a card's grade or transfer rating changed between approved versions. Every such change is logged automatically when the new version is approved. | evidence_corrections (severity grade_change) |
| Grades lowered | No data yet | Grade changes where the new grade is weaker than the previous one, ordered as in the Evidence Standard the new version was graded under. | evidence_corrections joined to the previous and new card versions |
| Grades raised | No data yet | Grade changes where the new grade is stronger than the previous one, ordered the same way. | evidence_corrections joined to the previous and new card versions |
| Withdrawals | No data yet | Ledger entries recording that a published record was withdrawn. | evidence_corrections (severity withdrawal) |
Method
| Indicator | Value | Definition | Counted from |
|---|---|---|---|
| Evidence Standard versions adopted | 0 | Versions of the Evidence Standard formally adopted (in force or since retired). Drafts are not counted. | evidence_standards (status published or retired) |
Reliability
How often the record had to change, how challenges ended, how long corrections took, how well calibrated our forecasts are and how often independent reviewers disagreed.
| Indicator | Value | Definition | Counted from |
|---|---|---|---|
| Corrections per published card | No data yet | Corrections Ledger entries on public Evidence Cards (clarification, minor, material or withdrawal — grade changes are counted separately below), divided by the number of public cards ever published. | evidence_corrections (published, card subject) ÷ evidence_cards (public, published) |
| Reversals: grade changes after publication | No data yet | Published cards whose grade or transfer rating later changed, as a share of all public cards ever published. Every grade change is a permanent entry in the Corrections Ledger. | evidence_corrections (severity grade_change) on public cards |
| Challenge outcomes | No data yet | Resolved challenges by outcome. Each was decided by a reviewer independent of the challenged record. Withdrawn and open challenges are not counted. | evidence_challenges (status upheld, partially upheld or rejected) |
| Challenges upheld (in whole or in part) | No data yet | Upheld and partly upheld challenges as a share of all resolved challenges. | evidence_challenges (resolved) |
| Median time to correction | No data yet | Median number of days from a challenge being received to the publication of the correction it led to. Corrections that did not arise from a challenge have no recorded report date and are not included. | evidence_corrections.published_at − evidence_challenges.received_at |
| Forecast calibration (Brier score) | No data yet | Mean of (forecast probability − outcome)² over forecasts written by BUILD IT staff that have resolved yes or no; pending and ambiguous forecasts are excluded. 0 is perfect; always answering 50% scores 0.25; lower is better. Shown only from 20 resolved forecasts. | forecasts (resolution yes or no, written by a staff member) |
| Inter-reviewer disagreement rate | No data yet | Share of public-record reviews with at least two independent submitted assessments (primary and second reviewer, not flagged conflicted) in which the reviewers proposed different grades or different transfer ratings. Shown only from 5 such reviews. | review_assessments (submitted) on evidence_reviews without a client organisation |
How these figures are produced
Each value is a count of rows in the public record, filtered exactly as the definition says. Private records held for clients are excluded. No figure on this page is sampled, estimated, rounded up or entered by hand.
Grade direction uses the order of grades in the Evidence Standard version the new grade was issued under. Every grade change is listed individually in the Corrections Ledger.
Reliability indicators are computed from aggregate counts that identify no person, client or record. Two are withheld until there is enough to measure: forecast calibration until 20 forecasts have resolved, and the reviewer disagreement rate until 5 reviews have two independent assessments. Until then the number of cases so far is shown instead.