Written by the pipeline that builds the data, regenerated with every refresh (2026-09-10).
The free files are the whole dataset — every notice back to 1988, no delay, no login, no paywalled years — and they stay free. None of it is secret: the state portals are public and biglocalnews/warn-scraper is open source. We measured the gap on 2026-09-03, when the archive held 45,755 rows (it holds 57,975 in today's build — this paragraph deliberately keeps the numbers from the day the measurement was taken): only 1,188 of them (2.6%) were older than what the state portals still serve today. That measurement is why the history stopped being a $199 product on 2026-09-07: it is not scarce data, it is 46 portals already scraped, de-duplicated into one schema with stable ids, as a 7 MB CSV (plus JSON) you can load in one line — so you can have it, and the only thing we charge for is the work that has to happen every morning.
The two exceptions are Connecticut and Michigan. CT DOL took its per-year WARN listings offline in 2025 and the replacement document library starts at September 2025; we rebuilt 2010–2025 from Internet Archive captures. Of the 872 Connecticut notices in the archive, 844 are dated before that cut-off and are not served by any state portal today. Michigan's current page lists notices from November 2024 only; its former per-year listings on milmi.org (2000–2025) were retired in 2025 and we rebuilt them from Internet Archive captures, checking each year against the total the agency printed. Of the 2,202 Michigan notices in the archive, 2,060 predate the current page.
What you can pay for is maintenance you would otherwise run. Portals change layout, drop fields and go quiet; keeping 46 of them normalized every day is the work, and we do it whether you pay or not. WARN Watch is that same work pointed at your list: a year of daily checks against your employers and states that you never have to keep running, delivered to Slack, Discord, Teams or RSS. If that is not worth more to you than an afternoon and a cron job, take the free files with our blessing — every row is there, and that is what they are for.
What "maintenance" means, in today's build:
Example: the raw files carry 126 different name strings for Kmart (store numbers, "dba" trade names, punctuation); all resolve to one company_canonical value, so a count per employer is one line of code instead of a regex you maintain. Same schema in every state, stable ids, both dates ISO. That is the weekend of work, done every morning.
Then you already get the layoff signal — a flag on a company record, inside the tool your team lives in. Keep it. What a sales-intelligence seat does not hand you is the filing itself as data: the date the notice was filed, the date the cuts take effect, the worker count as filed with the state, the state’s own notice text, and every earlier filing by that same employer — 57,975 of them back to 1988, as a CSV you can join to your own tables under CC BY 4.0. Those are two different products: one tells a rep to call today; the other lets you count, chart, audit and cite. We sell the second, and we do not sell seats.
Who that second product is for: anyone who has to answer a question with the record instead of a notification — WARN-compliance and employment-law research (how many notices in a state ran under the 60-day line, and which employers filed them), labor economists and reporters (a state-by-state series since 1988, stable ids, one schema), workforce boards and outplacement firms planning where the next cohort lands, and engineers who would rather have a file than a per-seat login. See the notice-period analysis → · How this compares to the paid trackers ↓
What this free archive has that a scraper run today does not: 844 Connecticut notices (2010–2025) that no state portal serves any more — CT DOL took its per-year listings offline in 2025 and we rebuilt them from Internet Archive captures; the whole 2020 peak (15,629 notices, 1,962,468 workers) in the same schema as this morning's rows; and stable ids with one company_canonical name per employer across 46 states, so a count per employer is one line, not a regex. Everything else in it is public record that we keep normalized for you — free, in the daily release, with no account.
Who this is actually for:
