Layoffs Tracker — WARN Notices & Company Layoffs

A layoffs tracker for agents and GTM teams: every official WARN notice from 13 US states, plus announced company layoffs, as one clean, deduplicated, typed feed.

If you sell to companies, a layoff is one of the loudest signals there is. Budgets get cut, tools get consolidated, a VP who owned your deal leaves, and a new one arrives with a mandate to change things. In the US, employers planning a plant closing or mass layoff must file a WARN notice with the state weeks before it happens. That makes these the earliest hard facts about a downsizing. The catch is that every state publishes them differently: spreadsheets, CSV exports, HTML tables, month-by-month lists, amendments filed as separate rows. Nobody has time to check a dozen state sites every week.

This Actor does it for you. It reads the official WARN filings of 13 states and the layoffs companies announce publicly. It turns them into one row per workforce-reduction event — company, employees affected, notice and effective dates, location, layoff vs closure, and the link to the official notice — newest first. Schedule it, and each run returns only what is new or amended since the last run.

Raw facts only. No AI summaries or scores: your agent decides what a layoff means for your pipeline.

Who it's for

• AI agents and MCP clients that poll for downsizing events on a schedule and act on them. Every field is typed, dates are ISO, and a missing fact is null, never a guess.
• Sales and RevOps teams watching their accounts for budget cuts, tool consolidation and churn risk. Add your account list and get only their events.
• Recruiters and staffing firms who want to reach displaced talent the week a notice is filed, by state, city and headcount.
• Investors and analysts tracking distress and restructuring in a sector or region.

What you get

One row per event:

Field	Example
company	Microsoft
eventType	layoff · closure · furlough · relocation · unknown
employeesAffected	277 (a number only when the filing states one)
noticeDate / effectiveDate	2026-09-22 / 2026-11-21
state / city / county / location	WA / Redmond / — / Redmond, WA
industry	54 Professional Scientific and Technical Services (where the state publishes it)
sourceType	warn-notice or news
noticeUrl	the official notice or listing
isAmended, sources[]	every filing and amendment folded into this event
isNewSinceLastRun, isUpdatedSinceLastRun, firstSeenAt	the incremental feed
canonicalId	a stable ID for the event across runs

It also keeps the source's own wording next to the parsed value: eventTypeRaw ("Mass Layoff - No Recall"), employeesAffectedText ("39 (remote workers in MD)") and effectiveDateText ("9/16/26 - 1/8/27"). You can always check a row against the filing.

{
  "canonicalId": "wa-91c678630f449839",
  "company": "Microsoft",
  "eventType": "layoff",
  "eventTypeRaw": "Layoff Permanent",
  "isTemporary": false,
  "employeesAffected": 277,
  "noticeDate": null,
  "receivedDate": "2026-09-22",
  "effectiveDate": "2026-11-21",
  "filedMonth": "2026-09",
  "state": "WA",
  "city": "Redmond",
  "location": "Redmond, WA",
  "sourceType": "warn-notice",
  "sourcePortal": "WA ESD WARN database",
  "noticeUrl": "https://…",
  "isAmended": false,
  "sourceCount": 3,
  "isNewSinceLastRun": true,
  "isUpdatedSinceLastRun": false
}

Coverage

Official WARN notices: California, Colorado, Florida, Illinois, Indiana, Maryland, New Jersey, New York, Ohio, Pennsylvania, Texas, Washington and Wisconsin.

Announced layoffs: company layoffs reported in the news that never reach a WARN filing, such as tech and remote teams, cuts below the WARN threshold and staff outside the US. Only headlines that name the company and state a cut are used. Speculation ("could cut", "weighs layoffs") and roundups are skipped. When an announcement matches a WARN notice (same company, same place or size, within 30 days), it is attached to the notice as a source instead of listed twice.

Some states publish less than others. New Jersey and Pennsylvania give only the month a notice was posted, so noticeDate is null and filedMonth is set. New Jersey, Texas and Florida publish no layoff/closure type, so eventType is unknown. We never fill a gap with a guess. More states are being added. If a state can't be read on a given run, the run still succeeds with the others and says which one was missing.

How deduplication works

• Re-filed and amended notices become one event. The latest filing's facts win, and every filing is kept in sources[] with isAmended: true.
• Two sites of the same employer stay two events: each WARN notice covers one site.
• Several rows filed the same day for the same site (one per unit or department) become one event with the summed headcount.
• Many articles about the same announcement become one event.

How to use it

1. Run it as it is. You get every notice filed in the last 30 days across all covered states, plus announced layoffs.
2. Narrow it down: pick states, add your companies to watch (names or domains), filter by event type, headcount or effective date, or add keywords such as hospital or warehouse.
3. Make it a feed. Turn on Only new or amended since the last run, give it a Watch ID, and schedule it daily or weekly. The first run is the baseline. After that, each run returns only new notices and notices an amendment changed.

For AI agents: the input is fully optional, and field names are forgiving (states, companies, daysBack, maxResults, …). The run summary in the key-value store (OUTPUT) lists events per state, which states were read, and anything that was skipped.

Pricing

Pay per event: $0.01 per layoff event delivered to your dataset, plus the standard Actor start fee. Duplicates, amendments of an event you already have, filtered-out rows and empty runs are never charged. A typical daily run over all states delivers 5–15 new events a day (about 5–15¢), and a 30-day backfill of every covered state is roughly 250 events.

Good to know

• WARN law covers larger layoffs (typically 50+ workers at a site, with state variations). Smaller cuts show up only if a company announces them.
• employeesAffected is what the employer filed. The final number can differ, and amendments are picked up automatically.
• companyDomain is left empty on purpose: we don't guess domains from names. Pair this Actor with a company-enrichment Actor to add domains and decision-makers to the rows you care about.
• Dates are ISO (YYYY-MM-DD). A date range keeps its first date in the parsed field and the full wording in the …Text field.