Talent Risk · Data
Talent Risk Data
Job cuts at large employers in the UK, the US and Europe, as we find them (sourced, checked and counted once) as data you can build on.
Why it’s different
One event, counted once. Twenty headlines about the same 5,000 jobs are one record with twenty sources, not twenty rows.
The right company. Every employer is a fixed identity (Wikidata). Namesakes, suppliers and sector stories are kept out.
Europe in its own languages. A cut in Germany counts the day the German press reports it, not weeks later in English.
Checked, and measured. A nightly re-read of a random sample: 97.4% held up in full over 30 days. How it works
Plans
Pro
£39 a month
Recruiters, advisers and analysts watching a list
- Same-day events for the employers you choose (up to 50)
- One email a day when anything on your list changes
- CSV and JSON downloads
- Cancel any time
Feed
£249 a month
Teams building on the data
- The full same-day stream: every employer, every event
- API key and webhooks
- Every field, including roles and countries
- History from the start of the record
Licence
From £9,000 a year
Analysts, platforms and research
- The daily archive of what was known, when, kept since 2nd November 2026
- Delivery to your warehouse (CSV, Parquet, S3)
- Redistribution and display rights
- A named contact
Watchlist pilot
Free for two weeks
Outplacement teams following a list of employers
- A daily email: which employers on your list changed, where, and what is confirmed
- What changed since yesterday, with every source
- At the end, tell us what you used and what was noise
- A few places only
Paid plans are not open yet. What is live today is the free data below; the plans show what each will include, and parts of them (alerts, API keys, webhooks, warehouse delivery) are still being built. Register interest and tell us what you need. We never sell to the employers we rate, and nothing paid changes what the free site shows.
Register interest
Email editor@talentrisk.com with your organisation, what you would use the data for, and the markets you need (UK, US, Europe). You will hear when your plan opens; nothing else is sent.
Free, today
| What | Where | Notes |
|---|---|---|
| Ask an AI assistant | /mcp | MCP server for Claude, ChatGPT and others: labels, counts and announcements, read-only. How to add it |
| Every event, last 12 months | /v1/events.json | 7 days behind; same-day is the Feed plan |
| One employer | /v1/employers/{slug}.json, e.g. amazon | Live: label, counts and the latest announcement. Every event and source, 7 days behind |
| Follow an employer | /companies/{slug}/feed.xml | RSS: a new item whenever its record changes |
| Follow a market | UK · US · Europe · All | RSS: company announcements and major outlets |
| The labels, as a table | /data/index.csv · JSON | Every rated employer and every one in the news, one row each: label, rule, 90-day cuts, live |
| Identifiers for every employer | /v1/identifiers.json · CSV | Wikidata, LEI, SEC CIK, Companies House, ISIN, listings: below |
| Every announcement’s id and its dated changes | /v1/ledger.json | 7 days behind, like the events |
| What was known, when | /v1/archive.json | The daily archive’s index: below |
One event, as the API returns it
{
"employer": "amazon",
"name": "Amazon",
"country": "US",
"market": "us",
"type": "cuts_announced",
"label": "Cuts announced",
"first_reported": "2026-11-03",
"last_reported": "2026-11-03",
"jobs": 1250,
"jobs_counted": false,
"ai_linked": false,
"roles_cut": [],
"roles_hired": [],
"countries": [],
"reasons": [],
"method": [],
"complete_by": null,
"sites": [],
"reason_quote": {
"text": "Amazon reportedly laid off \"fewer than 1,000 white-collar workers\" in its stores division.",
"english": "Amazon reportedly laid off \"fewer than 1,000 white-collar workers\" in its stores division.",
"publisher": "tech.yahoo.com",
"url": "https://news.google.com/rss/articles/CBMikAJBVV95cUxQY1dMa2JSQTk3UE9NWFFkaURrMDJVM3hhN2RWZmNkWXAzTWlITFZMQmttenY2T1RGT2ctOXRrY20tNjVoeTBMek1iVFd5MjFUdl9zWUt3eDloQndkajQ1ZkxUc0FXckUyNWdJN2ZHZHFrd1Vub0hXU2owUVRQYlFxU193UU9pZDRuSXBSSUJnVVp1bW1oRFdUY1BnZGRwZDF6ekNPMVozd3kwcV9mUU5JM0pFbldVZHVfRUpDTWpPaFBZR2JpLVZ5aDcxSVhwUGZGWmNaMHhwYUw0MkpkZ1B3RFFkVWFOYjVTYndGVnV5ZENlRVVkODVjdjBFMkYzS1VCWC1NbGo2OXhlSTRpU056Qg?oc=5&hl=en-US&gl=US&ceid=US:en"
},
"moved_to": [],
"hired_from": [],
"best_source_tier": 3,
"outlets": 1,
"id": "amazon-2026-11-02-fa5d90",
"best_source": "Trade press",
"major_outlets": 0,
"count_reason": "Figure given by one report only",
"figure_reports": 1,
"figure_official": false,
"audit": null,
"summary": "Amazon laid off fewer than 1,000 white-collar workers in its stores division affecting the U.S., India and UK, and cut about 250 employees in its devices and services unit, mostly in software development.",
"sources": [
{
"date": "2026-11-03",
"publisher": "tech.yahoo.com",
"url": "https://news.google.com/rss/articles/CBMikAJBVV95cUxQY1dMa2JSQTk3UE9NWFFkaURrMDJVM3hhN2RWZmNkWXAzTWlITFZMQmttenY2T1RGT2ctOXRrY20tNjVoeTBMek1iVFd5MjFUdl9zWUt3eDloQndkajQ1ZkxUc0FXckUyNWdJN2ZHZHFrd1Vub0hXU2owUVRQYlFxU193UU9pZDRuSXBSSUJnVVp1bW1oRFdUY1BnZGRwZDF6ekNPMVozd3kwcV9mUU5JM0pFbldVZHVfRUpDTWpPaFBZR2JpLVZ5aDcxSVhwUGZGWmNaMHhwYUw0MkpkZ1B3RFFkVWFOYjVTYndGVnV5ZENlRVVkODVjdjBFMkYzS1VCWC1NbGo2OXhlSTRpU056Qg?oc=5&hl=en-US&gl=US&ceid=US:en",
"tier": 3,
"audited_on": null,
"audit_held": null
}
],
"url": "https://talentrisk.com/companies/amazon"
}Free data is CC BY 4.0: use it for anything, credit “Talent Risk” with a link. No key, no sign-up.
What we license is our own record: the selection, checking, de-duplication and structure. The facts in it come from public sources, each linked on every item, and we never republish an article. Each source keeps its own terms; the API carries the credit lines they ask for in its attribution field. Sources and licences
Put it on your site
Two lines show an employer’s label and latest cut on any page: a job board, a news story, an intranet. No cookies, no tracking. More on the badge
What’s in a record
- employer
- Fixed identity, name, HQ country, sector, approximate workforce
- type
- Cuts announced, consultation, voluntary exit, hiring freeze, restructuring charge, reported to be weighing cuts, and more
- jobs · jobs_counted
- The figure the reports give, and whether it counts as job cuts (given by the company or a formal notice, or by two reports that agree). For a voluntary exit programme it is the number of people offered — never counted as cuts
- roles · countries
- Which job functions are cut or hired, and where the jobs are
- reasons · method
- Why, as the sources state it (with a verbatim sentence and its source), how the cuts are made, by when, and which sites
- ai_linked
- Whether the company or a major outlet links it to AI
- sources
- Every report, with date, publisher, tier and link;
audited_onandaudit_heldwhere the nightly re-read checked it - id
- The event’s identifier, to refer to it and to find it again in later copies of the record
How sure, on every event
- best_source · best_source_tier
- The strongest kind of source that reports it: Official (1: the company, a filing or a formal notice), Major outlet (2) or Trade press (3)
- outlets · major_outlets
- How many different publishers report it, and how many of them are official sources or major outlets. Twenty copies of one wire story are one outlet
- jobs_counted · count_reason
- Whether the figure is in our totals and, if not, why: figure given by one report only, offered not cut, a sign not a cut, same announcement, same round, several companies in one story. How we count
- figure_reports · figure_official
- How many reports give the figure (within 10%), and whether the company or a formal notice gives it
- audit
- Where the nightly re-read checked any of its reports against the article: how many, how many held up, and when.
nullwhen none has been re-read yet
On every employer
- identifiers
- Its standard codes (below)
- announced_and_reported
- Two facts side by side, from two sources: the job cuts announced and counted over the record’s year, and the workforce the company’s own filings (SEC 10-K, Companies House accounts) or Wikidata report, by year, each with its source. We do not compute one from the other: an announced cut is not a person who has left, and a headcount moves for many reasons.
nullwhen no headcount is reported
What it does not claim
It is not every job cut. It holds what we have found in the sources on How it works: company statements and filings, US WARN notices, the European and UK restructuring and insolvency registers, and named news in 13 languages. Coverage is deepest in the US, where formal notices are public, and thinner in Europe. An employer with no record may simply not have been reported. Figures are what was announced, not people who have left. Each record is the latest version; earlier versions are kept in a daily archive.
Identifiers
Every employer carries the codes other datasets use for it, so the record joins to yours without matching names: its Wikidata id, LEI (the Legal Entity Identifier, checkable at GLEIF), SEC CIK, Companies House number, ISINs and current stock-exchange listings (exchange, its MIC code, ticker). They come from Wikidata, and the CIKs of the employers in the Index are set one by one and checked against SEC records. Each code is checked against its own standard (an LEI’s and an ISIN’s check digits, the shape of a CIK or a company number) before it is kept, and an employer with two different codes of one kind keeps neither. A code we do not hold is null; none is ever derived or guessed.
Of 8,022 employers and subsidiaries: LEI 3,727, SEC CIK 2,195, Companies House 205, listings 4,655.
/v1/identifiers.json and /v1/identifiers.csv list every employer; /v1/employers/{slug}.json carries them under identifiers.
What was known, when
Once a day the whole public record is written to a snapshot file that is never replaced, so what we said on any day can be checked later, including after a correction. /v1/archive.json lists every day (21 so far, from 2026-11-01), with the file’s size and its SHA-256, so a copy can be checked against what we published.
To read the record as it stood on a date: find the date in the index, fetch its file (the files are supplied with the Licence plan, or on request for research), gunzip it and read the JSON, and check its SHA-256 against the index. What each file holds:
- schema 1
- Per employer: slug, name, country, sector, core (in the Index), state (warning | watch | calm: Cutting, Signs, Quiet), jobs90 (counted job cuts first reported in the 90 days before). Per event: type, first and last report dates, headcount (the largest figure any report gives, counted or not), best source tier, and each source's date and link. No rule line, no counted flag and no event id.
- schema 2
- Everything in schema 1, plus per employer the rule line that produced its label, and per event its id, whether its figure was counted and why, and how many outlets and major outlets reported it.
A snapshot holds the labels and events as the record showed them that day, not article text. Files already written are never changed to a newer shape.
Sources and licences
Our own data (the selection, checking, de-duplication, structure, labels and summaries) is CC BY 4.0: credit “Talent Risk” with a link to talentrisk.com. Every source keeps its own terms, below. Where a licence asks for a credit line, the API carries it in attribution, and anything built on the data should carry it too.
| Source | Licence or terms | Used for | Credit line, or what we keep |
|---|---|---|---|
| Google News (and the outlets it finds) | Feeds licensed for personal, non-commercial readers only: switched off 3 Nov 2026 (GOOGLE_NEWS=1 only for a comparison). Items read before then: links and our own summaries only; articles remain their publishers' | Evidence (can change a label) | We keep a link, the publisher's name, our own one-line summary and at most one short verbatim quotation (300 characters or fewer) credited to its publisher; never article text. |
| The GDELT Project | Unlimited and unrestricted use, including commercial, without fee; any use or redistribution cites the GDELT Project with a link | Evidence (can change a label) | News discovered through The GDELT Project (gdeltproject.org). |
| Tavily (and the outlets it finds) | Commercial use, storage of search results and publication of derived records allowed under Tavily's terms; the articles remain their publishers' | Evidence (can change a label) | A news search by employer name, run only when an event rests on one report: we keep the link, the publisher's name, our own summary and at most one short quotation; never article text. |
| The employers themselves | Press releases published for the public; read as the company issues them, robots and text-and-data-mining rules honoured | Evidence (can change a label) | The company's own press release, linked. |
| US state WARN notices | Public records | Evidence (can change a label) | — |
| WARN Feed | CC BY 4.0 | Evidence (can change a label) | US WARN notices: WARN Feed (approjects-warn-act-notices.static.hf.space), CC BY 4.0, compiled from state listings; matched, de-duplicated and summarised by Talent Risk. |
| Big Local News warn-scraper | Apache-2.0 (the code we run to read the state pages; the notices are the states' public records) | Evidence (can change a label) | — |
| SEC EDGAR | Public domain (US government work) | Evidence (can change a label) | Read under the SEC's fair-access policy: a declared User-Agent and no more than 10 requests a second. |
| BODACC (DILA) | Licence Ouverte / Open Licence 2.0 (Etalab) | Evidence (can change a label) | Source: BODACC, Direction de l'information légale et administrative; data as published on the date shown with each notice. |
| The Gazette | Open Government Licence v3.0 | Evidence (can change a label) | Contains public sector information licensed under the Open Government Licence v3.0. |
| Companies House | No reuse conditions ("Companies House imposes no rules or requirements on how the information on the public register is used"); API limit 600 requests per 5 minutes | Evidence (can change a label) | — |
| Wikimedia Commons | Per file (many logos are public domain as simple marks; some are CC BY-SA); trademarks are not licensed by Commons | Which company | Employer logos, shown small beside the name to identify the employer, linked from Commons; never used to suggest endorsement. Terms to confirm per file. |
| Insolvency Service | Open Government Licence v3.0 | Context (never changes a label) | Contains public sector information licensed under the Open Government Licence v3.0. |
| Office for National Statistics | Open Government Licence v3.0 | Context (never changes a label) | Source: Office for National Statistics, licensed under the Open Government Licence v3.0. |
| US Bureau of Labor Statistics | Public domain (US government work) | Context (never changes a label) | Source: US Bureau of Labor Statistics, retrieved daily. BLS.gov cannot vouch for the data or analyses derived from these data after the data have been retrieved from BLS.gov. |
| Eurostat | CC BY 4.0 (Commission reuse policy, Decision 2011/833/EU) | Context (never changes a label) | Source: Eurostat, datasets jvs_q_nace2 and une_rt_m (EU27_2020), retrieved daily; latest value and the value a year earlier shown, unchanged. |
| Wikidata | CC0 1.0 | Which company | Employer identities, headcounts and standard identifiers (LEI, SEC CIK, ISIN, Companies House number, listings). |
| GLEIF | CC0 1.0 | Which company | — |
| SEC EDGAR (Exhibit 21 subsidiary lists) | Public domain (US government work) | Which company | — |
| SEC EDGAR (10-K filings) | Public domain (US government work) | Context (never changes a label) | — |
| Companies House (statutory accounts) | No reuse conditions | Context (never changes a label) | — |
| legislation.gov.uk and GOV.UK | Open Government Licence v3.0 | Context (never changes a label) | Contains public sector information licensed under the Open Government Licence v3.0. |
| UK Parliament | Open Parliament Licence v3.0 | Context (never changes a label) | Contains Parliamentary information licensed under the Open Parliament Licence v3.0. |
| US Federal Register | Public domain (US government work) | Context (never changes a label) | — |
Terms marked “to confirm” are being checked with the publisher.