Layoffs are usually reported as a single headline. In reality they unfold as a sequence of public signals: hiring slows, open roles disappear, a regulatory notice is filed, an announcement is made, and months later the effective dates arrive. Teams that track workforce changes well do not wait for the headline. They watch the whole sequence and assemble it into events they can trust.
This guide covers where those signals come from, how to turn them into a reliable event feed, and the boundaries that keep the work focused on companies rather than people. The broader practice of measuring labour markets from public data is covered in job-board data and labour-market intelligence.
Who needs this, and why “real time” matters
Talent teams use layoff data to find experienced candidates the week they become available. Market intelligence teams read it as a signal about a competitor or a sector. Investors, suppliers and lenders watch it as an early indicator of trouble. For all of them the value decays quickly: a workforce change learned about a month late is history, not intelligence.
The signals, from strongest to weakest
| Signal | What it tells you | Strength |
|---|---|---|
| Government layoff notices | Company, location, headcount, effective date | Strong where published |
| Public company filings | Restructuring and associated costs | Strong for listed companies |
| Company announcements | Scope and rationale, as the company frames it | Strong but selective |
| Reputable news coverage | Early reporting, sometimes before any notice | Medium, needs corroboration |
| Job posting withdrawals | Open roles pulled, hiring frozen | Leading but noisy |
| Careers page and office changes | Teams, locations or roles removed | Supporting |
| Community-run layoff trackers | Crowd-sourced event lists | Useful for discovery, needs verification |
Government notices
In the United States, federal WARN rules require employers with 100 or more employees to give 60 days’ notice of plant closings and mass layoffs. Several states have their own rules with lower thresholds or longer notice periods, and many states publish the notices they receive. These notices are the strongest single source available: they name the employer, the location, the number of affected workers and the effective date.
They also come in every format imaginable. Some states publish HTML tables, some spreadsheets, some scanned PDFs, and some change layout without warning. Collection has to handle all of them and detect when a source changes shape.
Outside the United States, collective redundancy notifications to authorities are often not published, so the other signals carry more weight.
Filings and announcements
Listed companies frequently disclose restructuring plans and their costs in public filings. Company announcements give scope and rationale, though framed as the company chooses. Both are authoritative for what they say and silent on what they leave out.
Hiring signals
The earliest signal is often a change in hiring. Open roles disappearing faster than usual, whole teams’ postings withdrawn at once, or a company that normally posts steadily going quiet. Postings are noisy, since roles get filled and reorganised, so treat a hiring drop as a reason to look closer rather than as an event. Collecting postings consistently enough to see a change is covered in job-board data and labour-market intelligence, and hiring as a buying signal in scraping intent signals to power sales intelligence.
Build an event model, not a list of articles
The core design decision is to track workforce events, each assembled from many observations, rather than storing every article and notice separately.
A useful event record includes:
- Company, resolved to a canonical entity so subsidiaries and brand names roll up correctly
- Event type: layoff, office closure, hiring freeze, restructuring
- Locations affected
- Headcount, with its source, since figures often differ between announcements and notices
- Key dates: first signal, announcement, notice filed, effective date
- Sources, each with its confidence tier and the date it was observed
- Status: rumoured, reported, confirmed
When a news report, a WARN notice and a company statement describe the same layoff, they should update one event, not create three. Matching on company, location and a date window catches most duplicates; the rest need review. Company identity resolution works the same way as described in building talent-mapping and org-chart data.
Confidence, not certainty
Different sources disagree, and early reports are often wrong on scale. Give every event a status that moves as evidence arrives.
- Rumoured: a single low-confidence source, such as a community post.
- Reported: reputable news, or a strong hiring signal with corroboration.
- Confirmed: a government notice, filing or official announcement.
Show the status to whoever consumes the feed. A talent team can act on a reported event; a report to leadership should usually wait for confirmation.
Collection that keeps up
Each source type needs its own cadence.
| Source | Cadence |
|---|---|
| State notice pages | Daily, with change detection |
| Filings | As published |
| News | Continuous or several times a day |
| Job postings | Daily for watched companies |
| Careers pages | Weekly, with change detection |
News sites and job boards are the sources most likely to rate-limit, geo-filter or serve regional editions, so collect each region from that region. With the Shifter gateway, the market goes in the credentials against p.shifter.io:443, and holding a session keeps a paginated listing consistent:
customer-USERNAME-country-us-sid-warn-watch-12-ttl-600:PASSWORDKeep request rates ordinary and back off on errors, as covered in rate limiting and request throttling. Store raw responses before parsing, so a notice page that changes format can be re-parsed without re-collecting; the pattern is in moving web scraping API data into SQL.
Track companies, not people
Layoff data sits close to individuals, and the boundary should be explicit.
Collect company-level facts: who, where, how many, when. Do not build lists of named affected employees from social media posts, and do not store personal details that appear in notices or articles unless your use genuinely requires it and you have a lawful basis. A talent team that wants to reach affected people should do it through channels those people chose, such as job boards and open-to-work signals they published themselves, not through a list compiled about them.
The same principle, working at the level of roles and organisations rather than individuals, is set out in building talent-mapping and org-chart data. The legal framing is in residential proxies and GDPR compliance.
Measuring the feed
| Metric | What it tells you |
|---|---|
| Lead time | How early the feed detected events compared with confirmation |
| Confirmation rate | Share of reported events later confirmed |
| Duplicate rate | Whether events from different sources are being merged |
| Coverage by jurisdiction | Where the feed is strong and where it relies on weaker signals |
| Source freshness | Whether any source has stopped updating or changed format |
FAQ
What is the most reliable source for layoff data?
In the United States, state-published WARN notices. Elsewhere, company filings and official announcements, supported by news.
Can hiring data predict layoffs?
It can raise an early flag. A sudden withdrawal of open roles is a useful signal, but it has innocent explanations, so treat it as a prompt to investigate rather than an event.
How do we avoid double-counting the same layoff?
Model events, not articles. Match new observations to existing events on company, location and date window before creating anything new.
Is it appropriate to track affected employees?
Keep the feed at company level. Individuals who want to be found will usually say so on the platforms they choose.
The bottom line
Layoffs are a sequence, not a headline. Track the whole sequence: hiring signals that come first, notices and filings that confirm, announcements and news that describe. Assemble them into events with a status, a confidence level and sources attached, collect each source at its own cadence, and keep the work at company level.
The product view is on the recruitment and talent data page, and the related guide to employer reputation is how to monitor employer reviews and ratings across sites.