Corrections log

Every value Statbook has ever published wrongly is recorded here, permanently, with a numbered advisory — and preserved in the figure's history with a corrected status. So is every published assurance claim that turned out to overstate its own evidence, even where the number itself was right — those carry no corrected marker, because nothing in the data was wrong. This page is a feature, not an embarrassment: a registry that claims it never errs is lying about something.

The protocol

Source divergence is auto-flagged disputed within 24 hours, before human review.
Confirmed errors are corrected within 1 business day via a patch release.
Each correction gets a numbered advisory (SB-C-NNNN) listing the wrong value, the right value, the affected period, and how it happened.
The wrong value stays in the record's history, marked corrected — history is never rewritten.
Business-tier subscribers receive correction notices directly.

Advisories

SB-C-0001 — assurance status corrected (no value was wrong)

Found 2026-07-26 · published in release 3.0.0 · affected releases 1.0.1, 2.0.0

Figures. employment.unfair_dismissal.compensatory_cap, employment.unfair_dismissal.min_basic_award

What was published. Both figures were published with the assurance status “corroborated” — Statbook’s claim that two independent official sources state the value. The second source was the gov.uk guidance page https://www.gov.uk/dismissal/compensation-and-tribunals, and that page had already been removed.

What is right. “single_source_A”: the value is stated by primary legislation only. Both figures are set by the Schedule to SI 2026/310, and that citation was, and remains, correct. The numbers themselves — £123,543.00 for the compensatory cap and £9,157.00 for the minimum basic award, both from 6 April 2026 — were right in every release and are unchanged.

Affected period. From release 1.0.1 (compiled 2026-07-09) to release 3.0.0. The overstatement was found on 2026-07-26 and corrected in the source data that day, but 2.0.0 — the release that was live — carried it until 3.0.0 replaced it.

How it happened. The corroboration gate checked the shape of the sources, not what they said: two source roles present, the guidance source on an official domain. It never asked whether the archived bytes of that guidance page actually stated the figure. The page had been retired — the cited URL redirects to https://www.gov.uk/dismissal, a guide that states neither figure and carries no money amount at all, and every archived copy Statbook holds of the original URL is a navigation shell that never stated either value. Nightly re-verification did notice something, every night, but filed it as “changed-value-unconfirmed”: the deliberately silent finding meaning “this matcher has never matched this page”, which cannot tell a wording gap from a citation that is simply dead. So a published two-source claim rested on a page that stated neither number, and nothing said so.

The fix. From 3.0.0 both figures publish as “single_source_A” and the dead guidance source is dropped; no value changes. Two guards were added so the same overstatement cannot ship again: the build now refuses a “corroborated” status whose own archived source bytes do not state the value (Gate 2), and the nightly escalates a cited page that states nothing of the value’s kind — no currency or percentage figure anywhere on it — into an alert instead of a silent note. No value entry is marked “corrected”, because no value was wrong: the protocol’s “corrected” marker is for values.