Transparency
How We Source Our Data
Every layoff event on LayoffTalk comes from one of three sources: official government WARN notices, news reporting, or a firsthand community story. Here is exactly how each works — and where the limits are.
1. Official WARN Notices
The federal WARN Act requires most employers with 100+ employees to give 60 days' written notice before mass layoffs or plant closures. States publish these filings, and we import them daily. WARN notices are our highest-confidence source — they are legal filings, not reports — and they often surface layoffs weeks before news coverage.
We currently ingest WARN filings from 12 states:
California, Colorado, Florida, Georgia, Illinois, Massachusetts, New Jersey, New York, Ohio, Pennsylvania, Texas, Washington.
A WARN notice with a future effective date is listed under upcoming layoffs until that date passes. Notices from states we don't yet cover won't appear, so our totals understate national activity.
2. News Monitoring
We monitor news coverage daily for the thousands of companies we track. When an article reports a layoff, we use AI-assisted extraction to pull the structured facts: company, headcount, location, and dates. Every extraction carries a confidence score — if it's below our threshold (or missing), the event is held for review instead of published.
- When reporting gives a range ("500–600 jobs"), we record the confirmed lower bound.
- Percentages without a stated workforce total are recorded as an unknown headcount, not a guess.
- Implausibly large figures (typically a company's total workforce misread as a layoff count) are rejected automatically.
- Each news-sourced event links to the coverage it came from, so you can check the original reporting.
3. Community Stories
People who experienced a layoff can share their story anonymously. Every story is reviewed by a moderator before it appears on the site. Stories are individual accounts — they add context the filings can't, but they represent one person's experience.
Where AI Is Involved
We use AI in two places, and we label it:
- Extraction: turning news articles into structured events, gated by the confidence rules above.
- Summaries: event summaries marked "AI-assisted" are generated from the cited coverage. They can contain errors — the linked sources are always the authority.
Deduplication & Company Data
The same layoff often appears in a WARN filing and several news stories. We merge duplicates by company, date, and headcount, and we consolidate per-facility WARN filings under the parent company. Company profiles (sector, size, financials) come from public data and may lag reality.
Spot something wrong? Tell us — corrections with a source get fixed fastest.
Limitations
- Our data is not exhaustive: small layoffs, quiet cuts, and non-WARN states are underrepresented.
- Headcounts can be estimates or lower bounds, and filings are sometimes amended after we import them.
- Nothing here is legal, financial, or career advice. Verify with official sources before making decisions.
Who maintains this, and corrections
LayoffTalk is maintained by a small editorial team rather than credited to individual authors. Most entries are compiled automatically from public filings, so attributing them to a person would misrepresent how they were produced. What we do stand behind is the provenance: every event page names the agency or report it came from and links to the original, so you can check it yourself rather than take our word for it.
The editorial team reviews the parsers and summary pipeline, investigates reported errors, and decides what gets removed. Records are corrected or deleted when a filing is amended or withdrawn, when a company is wrongly identified, or when a summary misstates what a source said.
Found something wrong? Tell us and include the page URL. Corrections to the underlying record are applied at the source, so they propagate to every page that uses it.