Talent Risk

Talent Risk · Corrections

Corrections

How it works

We publish first, then invite correction. The Editor, Talent Risk (editor@talentrisk.com) answers every request. Every label rests on the public record — the company's own statements, filings, and reporting by named outlets — and every fact on a company's page sits beside its source.

If something we have published is wrong, or an employer wants context published alongside it, use the form below. We read every submission. Approved corrections and responses are published in full and verbatim on the company's page, usually within 24 hours, labelled as the company's response when they come from a company address.

This is a public channel. Do not include anything that identifies you or a colleague.

Policy

Errors of fact are corrected within 24 hours of being established. The correction is marked on the page with the date and what changed. The original text is not deleted.

We do not change a label because a company asks. We change it when the evidence changes, and we say so.

Corrections log

  1. 4 November 2026

    What was wrong. Employers that stopped announcing cuts fell from Cutting straight to Quiet after 90 days, skipping the 180-day step-down the method describes. Among them: Renault, British American Tobacco, Commerzbank, FedEx, BP, Centrica and Pixar.

    What changed. The step-down is read from the evidence; these employers show Signs (stepping down) until 180 days have passed (method v3.3).

  2. 4 November 2026

    What was wrong. UPS was labelled Cutting on a quarterly earnings release filed with the SEC, which is not an announcement of job cuts.

    What changed. Only filings that declare Item 2.05 (costs of exit or disposal activities) count as a formal notice of cuts, and earlier filings are re-read (v3.3).

  3. 4 November 2026

    What was wrong. Microsoft's July round showed one site's WARN notice (605 jobs) instead of the 4,800 cuts announced, and several employers' notices at different sites were merged into one figure (Wells Fargo, Citigroup, Intel, FedEx).

    What changed. Each notice is its own event; a notice inside a round already counted is shown, not added.

  4. 4 November 2026

    What was wrong. 320 US WARN notices had been set aside by a check meant for news reports, because their summaries did not repeat the employer's name.

    What changed. Formal notices matched by the employer's legal name are kept; the 320 were restored.

  5. 4 November 2026

    What was wrong. The downloadable files (index.json and index.csv) served labels from a retired weekly table, not the labels on the pages.

    What changed. Both files read the live record and are checked against the pages before every build.

  6. 4 November 2026

    What was wrong. A single trade-press item could decide a label; one outlet's regional editions counted as several outlets; and 'cites AI' was set by commentary.

    What changed. Only the company, formal notices and major outlets decide labels; outlets are counted by publication; AI is flagged only when the company or a major outlet links the cuts to AI (v3.3).

  7. 4 November 2026

    What was wrong. Results, annual-report headcount falls and feature articles about earlier programmes were read as new announcements (Diageo's 2,000, Bosch, Nestlé).

    What changed. They are background; every accepted item is read once more under the rule (v3.4).

Correct the record

0/1500

Last updated: 3 November 2026