How Layoff Radar counts
Layoff Radar collects company announcements, reporting and public notices into a source-linked record. It is a research collection, not a complete census of job losses.
What goes into the headline totals
Only confirmed, top-level events with source links contribute to totals. A primary source or corroboration across independent source domains meets the current confirmation rule. “Confirmed” describes the evidence for an announcement; it does not mean every proposed reduction has happened.
Reported events have one qualifying source and are shown separately. Linked WARN filings provide detail for an existing announcement and are not added again. Conflicting counts under review are excluded. When a company publishes no number, the event adds its documented minimum, shown as “at least”: the low end of a reported range, an explicit “at least” figure, sourced figures for separate places added together, or the people in its linked WARN notices, whichever is largest. Each minimum shows where it comes from. Vague wording such as “thousands” is never turned into a number, so an event with no documented figure adds nothing.
What source checks mean
Each research claim carries a source URL, quoted evidence and retrieval time. The gate attempts an exact or approximate quote match. Blocked pages, PDFs and unavailable sources may remain “Quote not rechecked”; a confirmation label is not a promise that every quoted passage was independently retrieved. A source whose page does not exist (the site answers “not found”) is rejected and never counts.
Reported workforce change
For some companies, Layoff Radar also records the headcount they publish themselves: in annual reports, SEC filings and quarterly results. These figures show workforce change beyond announced layoffs, but a decline is not necessarily a layoff. Attrition, hiring freezes, divestitures and changes in how staff are counted also reduce headcount. Headcount figures are therefore never added to layoff totals.
A decline of 2% or more is marked “Headcount decline — no matching event recorded” only after the snapshot date reaches 45 days beyond the later period, with no confirmed or reported event dated in that interval. Linked notices and recorded effective dates also count as evidence. Before then, the notice window is marked pending. Coverage is incomplete: absence of a matching record does not establish that no layoffs were announced, or that the decline represents job cuts. Layoffs are collected from January 2026, so only a decline that starts on or after the end of 2025 can be marked; earlier changes are shown without the mark. Figures are compared only on the same basis and scope, such as total employees worldwide.
How this compares with other counts
Challenger, Gray & Christmas counts job cuts announced by US employers in every industry: 573,195 through September 2026, of which 165,925 were in technology, with fintech (7,806) and telecommunications (7,446) listed separately. Layoff Radar covers tech and tech-adjacent companies worldwide, so all-industry figures are not comparable, and its US total runs below Challenger’s technology count. A cut is counted here only when public sources support it, and a company that publishes no number adds only its documented minimum. Challenger also counts cuts that are never itemized in public.
Scope and gaps
Layoffs at technology and tech-adjacent companies worldwide, plus US WARN notices. Coverage is incomplete; absence does not mean no layoffs.
- The January 16–31 research batch is still outstanding.
Dates are announcement or notice dates, not necessarily employment end dates. On the dashboard, region and country describe where a layoff happened. An event naming several countries, or described as global, appears under “Multiple countries” because its headcount cannot be split between them. The records page filters by company headquarters instead. “AI cited” reflects an attribution in the collected evidence, not an independently established cause.
Corrections and updates
Sources and company identities can change as new evidence arrives. Historical corrections retain an audit record. The data update date reflects curated changes, rather than how recently the site was rebuilt.
Latest curated change: Oct 17, 2026.
Open the dashboard