If you practise on the plaintiff side of employment law, you already know where WARN notices live: California EDD publishes a rolling report, New York DOL publishes an export, and every other state agency publishes something similar. The file is free. Getting it is not the hard part, and any competitor of yours can get it on the same morning.
The hard part is that the file is a notice log, not a lead list. It is designed to tell a workforce board that a layoff is coming so it can stage re-employment services. It is not designed to surface the filings where something went wrong. Three specific properties of the raw data hide exactly the filings you are looking for — and all three are visible once you stop reading the file a row at a time.
This page is about those three. It is written for counsel and risk analysts; there is a page on how the underlying data is built if you want the mechanics first.
1. The signal is the gap, not the notice
A WARN filing carries two dates: the date the employer gave notice, and the date the layoff takes effect. Federal WARN (29 U.S.C. §§2101–2109) requires 60 days between them. Cal-WARN (Labor Code §§1400–1408) also requires 60, and reaches smaller employers than the federal statute does, which is why a great many California filings have no federal counterpart at all.
So the arithmetic is trivial and nobody publishes it. Here is the distribution across all 3,997 California filings on record, every one of which carries both dates:
| Days from notice to layoff | Filings | Share |
|---|---|---|
| Layoff dated before the notice | 153 | 3.8% |
| 0–29 days | 519 | 13.0% |
| 30–59 days | 556 | 13.9% |
| 60–74 days | 2,366 | 59.2% |
| 75–120 days | 349 | 8.7% |
| 121 days or more | 54 | 1.4% |
The median gap is 61 days. The statute is doing visible work — the single largest bucket sits just above the 60-day line, which is what compliance looks like in aggregate. But 1,228 filings, 30.7% of the California series, show fewer than 60 days between the notice date and the layoff date. 216 of them show the same date for both, across 154 distinct employers. The largest negative gap in the file is 305 days: a layoff dated most of a year before the notice that describes it.
What that number is not. It is not a count of violations, and any vendor who sells it to you as one is selling you a problem. A gap under 60 days is a screening signal — a reason to pull the underlying filing — and there are at least four innocent explanations that have to be excluded before it becomes anything more:
- Statutory exceptions. §2102(b) allows reduced notice for a faltering company, unforeseeable business circumstances, or a natural disaster. The employer must give as much notice as is practicable and state the basis for the shortfall — but a short gap on its face is consistent with a lawful notice.
- Amended notices. An employer that revises an earlier notice generates a second record dated much closer to the layoff. The record you are looking at may not be the first one.
- Date semantics. A state file may carry the date the agency received or processed the notice rather than the date the employer served it. Those differ, sometimes by weeks.
- Rolling separations. The layoff date is frequently the first date of a schedule that runs for months, not a single event.
The value of the gap is that it takes a four-thousand-row file and hands you roughly twelve hundred rows worth opening. That is a triage tool. It is not evidence, and it does not become evidence until someone reads the notice itself.
2. The layoffs that are invisible one filing at a time
This is the one that is genuinely hard to see by hand, and it is the one most likely to be sitting unworked in your own jurisdiction.
Both statutes aggregate. Federal WARN, at §2102(d), treats employment losses occurring within any 90-day period as a single event where they would separately fall below the threshold, unless the employer shows that the groups result from separate and distinct causes. Cal-WARN's mass layoff trigger counts 50 or more employees at a covered establishment within a 30-day period. Either way, the legal unit is a period and an employer — not a row.
The state file is organised the other way round. Employers file per site, per building, per store, often on the same day. Read a row at a time, each one is a small layoff nobody would look twice at.
Counting the California series by resolved employer and county, there are 45 employer-county pairs in which no single filing reaches 50 affected workers, but a 90-day window of those filings sums to 50 or more. The largest:
| Employer | County | Filings in window | Workers, aggregated |
|---|---|---|---|
| PULAU Corporation | San Bernardino | 8 | 133 |
| De La Pena Eye Clinic | Los Angeles | 7 | 130 |
| Telecare Corporation | Ventura | 6 | 116 |
| Downtown College Preparatory | Santa Clara | 4 | 99 |
| Good Sports Plus Ltd | Imperial | 13 | 97 |
| CommUnify | Santa Barbara | 22 | 97 |
Twenty-two filings from one employer in one county inside a 90-day window, not one of them reaching 50 workers. On the state download that is twenty-two forgettable rows. Grouped, it is a question worth asking.
The fragmentation is often literal. On 11 August 2026 one California employer filed nineteen separate notices on a single day — seventeen in Alameda County, two in Contra Costa — each one recorded under a different street address, with headcounts running from 1 to 20 and totalling 82 workers. Sorted by worker count, all nineteen sit at the bottom of the file.
The caveat that matters here. Aggregation is a legal question about a single site of employment and about the causes of each group. A county is not a site, and this count uses county because that is what the state publishes. It tells you which employers to go look at. It does not tell you the aggregation succeeds — separate causes, separate sites, and the 90-day proviso all have to be worked through on the actual facts.
3. The employer is not one name
Searching a state file for an employer by name is the single most common way to conclude, wrongly, that there is no filing history.
Agencies record whatever string the employer wrote on the form. The same company appears as its legal name, its dba, its subsidiary, and its store or building. Across the California series, 2,648 distinct filed name strings resolve to 1,632 employers — meaning around a thousand name strings are duplicates of an employer already in the file under another spelling.
Some of it is mundane punctuation. Some of it is substantive: the file contains separate entries reading Epic Games Inc., Epic Games Inc. (Remote Employees in Los Angeles), Epic Games Inc. (Remote Employees in El Segundo) and Epic Games Inc. (Remote Employees in San Francisco) — four rows, one employer, and an unresolved question about site of employment for remote workers sitting right there in the employer's own labelling.
Practically: never conclude an employer is a first-time filer from one name search. Search the distinctive token, not the full legal name, and expect the suffixes to be wrong. Kcalbin's employer index groups the resolved names and publishes the underlying strings it combined, so any grouping can be checked rather than trusted — the methodology page states the merge rule and its known failure mode, which is a deliberate bias toward leaving an employer split rather than merging two different ones.
A smaller thing that costs people filings: the sort
Most public WARN views, including Kcalbin's own recent-filings page, sort by effective layoff date. That is the right sort for a workforce board, which cares what is about to happen. It is the wrong sort for monitoring, and it fails in both directions.
A notice filed this morning for a layoff eight months out sinks to the bottom of the page — the California file contains a notice dated 27 January 2026 for a layoff on 18 September 2026, and a New York notice dated 21 July 2026 for a layoff on 31 March 2027. Meanwhile an old notice with an imminent date floats to the top and looks new every day until it passes. If you are checking a layoff-date-sorted page for new filings, you are reading a list that was not built to answer that question. Sort by notice date, or diff yesterday's file against today's.
Why any of this is time-sensitive
WARN contains no federal limitations period of its own. In North Star Steel Co. v. Thomas, 515 U.S. 29 (1995), the Supreme Court held that courts borrow the most closely analogous state statute of limitations instead. The practical consequence for intake is that the clock you are running against is not uniform across jurisdictions, and it is running from conduct that became a public record on the day the notice was filed — typically weeks before any affected employee has reason to call a lawyer.
Damages under §2104 are measured in days of violation, up to 60. The window in which the facts are cheapest to establish — employees still reachable, records still current, the employer still operating — is at its widest on the day the notice appears in the file, and narrows from there.
Doing this without paying anyone
All of the above is reproducible from public records, and if you have a paralegal with a spreadsheet you should reproduce it rather than buy it:
- Download your state's WARN file on a schedule and keep every version. The agency file is a current snapshot; the history is only in your copies. Diffing yesterday's against today's is what turns a snapshot into an alert.
- Compute
layoff_date − notice_dateand sort ascending. Everything under 60 goes in a review pile. - Normalise employer names — strip legal suffixes, dba tails, and site descriptors — then group by employer and county and sum affected workers over rolling 90-day windows.
- Flag any group that crosses a threshold no individual row crosses.
- Open the actual notices before drawing any conclusion. Steps 1–4 produce a review pile, and nothing in them is evidence of anything.
That is a genuine afternoon of work and then a standing weekly task, per state. Whether it is worth doing in-house is an economics question with an honest answer in both directions.
What we sell, stated plainly
Kcalbin does steps 1–4 continuously for California and New York, and adds one thing a spreadsheet cannot: each filing is scored against its county's labour market using BLS Local Area Unemployment Statistics, then ranked against every other live filing in the state with a future layoff date. A hundred-person layoff in a county already absorbing three closures is a different situation from the same layoff in a county that is hiring, and the ranking is what turns a list into a queue.
The data pages themselves are free and public and always will be — county filings, employers, New York, SB 617 notices, and the methodology and its gaps. What is paid is the scored and ranked delivery: $149 for a single filing unlocked in full, or $299/month for every CRITICAL and HIGH filing as it is scored. Flat pricing only — Kcalbin never takes a share of a settlement, a fee, or a case value, and there is no code path that could.
Leads are delivered teaser-then-unlock, so access starts with a request rather than a public buy button. That is deliberate: the unlock link is tied to one specific filing.
WARN Intelligence — pricing and access How the data is built
Kcalbin publishes public records and labour market statistics. It does not provide legal advice, and nothing on this page is a substitute for it. Statutory descriptions are plain summaries; read 29 U.S.C. §§2101–2109 and California Labor Code §§1400–1408 directly. Counts are from Kcalbin's 13 August 2026 snapshot of California EDD and New York DOL records and will move as those sources are refreshed.