WARN Filing to a Ranked Queue of Reachable Displaced Candidates

A just-filed mass-layoff notice names a company and a site. It names zero people. This guide is the conversion step that public pages skip: how an in-house recruiter or sourcer takes one WARN or mini-WARN filing and produces a role-matched, reachability-ranked outreach queue of named former employees, timed to reach them before the open-market flood begins. You get the resolution method (company plus site to people), the timing math (which date you count from), and the compliance log that keeps the queue lawful.

A WARN notice is a Worker Adjustment and Retraining Notification filing, the advance notice a large employer must give before a plant closing or mass layoff. Reading one is easy. Turning it into named candidates, timed correctly, is where the work lives.

What a WARN filing actually gives you, and what it withholds

A WARN notice carries the employer and the site, not a single named individual. Federal WARN requires employers with 100 or more employees to give at least 60 calendar days advance written notice of a plant closing or mass layoff affecting 50 or more workers at a single site. The filing you can read tells you who is cutting, where, how many, and when. It tells you nobody's name.

A standard notice carries seven fields worth capturing every time:

• Company name (often a subsidiary, not the brand you know)
• Site location, by city or county
• Number of workers affected (an employer estimate, sometimes blank)
• Type of action (layoff, closure, or relocation; permanent or temporary)
• Effective date (the last day of work)
• Notice date (when the filing was made)
• Industry or NAICS code, in many states

That is the raw material. The conversion job is to take those fields and generate a list of named people the filing itself never contains. Everything downstream depends on two of those fields: the worker count, which sizes your cohort, and the effective date, which times your outreach.

The data is messy by design. Each state runs its own system: some post PDFs, others spreadsheets, some have searchable databases and many do not, formats change without notice, and archives disappear. One scraped state had an effective date filled on only 303 of 310 notices; the seven blanks were filings where the state printed no full date. Those seven cannot be timed, which matters because timing is the product.

Why the effective date is the only clock that counts

The lead time you own is the gap between the notice date and the effective date, usually about 60 days and 90 in New York, New Jersey, and Maine. During that window the person is identifiable from the filing but has not yet broadcast availability. Once the effective date passes, people flip Open to Work, which the platform rewards with 40 percent more recruiter InMails, and your advantage dissolves into the open-market flood.

There is an anchor problem, and it is the single most common way this job goes wrong. Some writers time from the notice date; the sources disagree. Count from the effective date. The effective date is the last day of work, the moment active search begins. Practitioner guidance tells laid-off people to wait at least 24 hours before flipping Open to Work, update the role end date the day after the last day, and reach full active-search mode by about week three. Your reach-by date should land before that week-three surge, which means before and shortly after the effective date, not before the notice date.

The states with stricter laws hand you more to work with. Lower thresholds push more filings into the public record, so California, New York, and New Jersey yield denser pools than federal-floor states. New Jersey alone logged 141 notices covering 17,913 workers in 2026 year to date, already above its full 2025 total of 90 notices.

The thresholds and notice periods that set your clock

Here is the eligibility and timing table. The notice-days column is what drives your reach-by math: subtract it from the effective date to find when the person became identifiable.

A mass layoff under federal WARN is 50 to 499 workers if they are at least one-third of the site workforce, or 500 or more regardless of the fraction. Separate actions aggregate: if a company terminates 30 on day one and 29 on day 31, the mass-layoff threshold can be reached through separate actions within a 90-day period. That aggregation rule is why a single site can generate several related filings in a short span, which you will need to dedupe later.

The count of mini-WARN states is genuinely disputed. Sources put it at 14, 15, 16, 18, 20, or 23, and I will not pretend to a number the record does not support. What is not disputed is the shape: stricter states file more, and California, New York, and New Jersey are the densest workable jurisdictions. If you source nationally, weight your attention there.

New Jersey's one-week-per-year severance is worth noting for your outreach framing. People leaving with meaningful severance are often in a more deliberate search than people cut without it, which changes the tone of a good first message.

The procedure: from one filing to a ranked queue

This is the end-to-end method, in order, with owner and rough time per stage. Run it on a single filing and you finish with a dated, ranked, compliant queue.

The two stages that carry the most risk are resolving the employer entity and isolating the cohort. The filed legal name is often a subsidiary, so search parent brands too; a mismatch here leaves your pool empty or wrong. On the cohort, remember that not everyone at a site is cut, and current-versus-past status on profiles is unreliable mid-layoff, so you confirm individuals by end-date changes near the effective date rather than trusting a raw export.

The resolution stage is where Refolk removes the most friction. Steps three through five are a manual chain of entity matching, Boolean filtering, and end-date confirmation that falls apart when the filed name is a subsidiary or the site filter is noisy. Asking for the people in plain English collapses that chain into one query and returns named people with a contact channel already attached.

How this goes wrong: failure modes and false positives

Most WARN-sourcing attempts fail in one of eight predictable ways. Each has a tell and a check. Work through them before you call a queue finished.

Timing and entity errors

Wrong timing anchor. Counting from the notice date instead of the effective date compresses your window. The false positive is believing you have three weeks when the effective date is eight weeks out. Recompute from the effective date, and flag records with a missing effective date because they cannot be timed at all.

Subsidiary mismatch. The filed legal name differs from the LinkedIn brand, so your pool comes back empty or wrong. The company name is often a subsidiary. Confirm the parent or brand before pulling the pool.

Amendment or withdrawal blindness. Notices can be revised or withdrawn; filings are snapshots. The check is to look for an "(Amended)" tag or a refiling within 45 days at the same location before you invest outreach in the cohort.

Cluster double-counting. Related notices in one week often equal one coordinated restructuring. The false positive is treating one event as several and inflating your pipeline. Dedupe by employer plus location plus effective date.

Cohort and outreach errors

Over-inclusion of the uncut. Not everyone at a site is laid off, so ranking a full-site list wastes outreach. Size against the filed worker count and prioritize end-date changes near the effective date.

Headcount taken as fact. The number is a conservative employer estimate, and totals run low. Treat it as a sizing bound, not a roster.

Generic blast. Most outreach is ignored, and the false positive is a high open rate with near-zero replies. Use a named role and a specific reason, and cap at three touches.

No logged lawful basis. Sourcing from a public filing without a legitimate-interest assessment and an Article 14 notice is the compliance gap. Store the source and the assessment per candidate before first contact.

Ranking by reachability and role match

A queue is only useful if it is ranked by two things at once: fit against your open reqs, and whether you can actually reach the person. A perfect role match with no contactable channel is a dead row. Rank reachability by channel, because channel choice drives response.

Personalization roughly doubles InMail response, which is why a WARN-derived queue of named people with a specific, known reason to reach them is structurally advantaged. Average InMail response runs 10 to 25 percent; personalized outreach can reach 35 to 50 percent. The reason is already in your hand: you know the company, the site, and the approximate last day.

Small beats large. Campaigns under 200 recipients returned 15 to 20 percent reply rates against 8 percent for campaigns of 500 to 1,000. A tight WARN-derived queue is exactly the kind of small, researched campaign that wins. Pair that with sequence discipline.

When you score, use a simple rubric so the ranking is defensible and repeatable.

Cohort confidence: 2 = end-date near effective date, 1 = likely, 0 = uncertain
Role match: 2 = open req match, 1 = adjacent req, 0 = no current req
Reachability: 2 = direct channel (email/GitHub/SMS consent), 1 = InMail only, 0 = none found
Timing: 2 = inside notice-to-effective window, 1 = within 3 weeks after, 0 = past week 3
Seniority fit: 2 = target band, 1 = one band off, 0 = two or more off

The contact channel column is what separates a ranked queue from a list. Asking Refolk for the cohort ranked by how reachable they are by email or GitHub returns that column populated, so you are not hand-hunting for a way to reach each name.

Scoping your realistic pool size

Before you run a WARN filing, know how large the addressable pool is for the roles you source, so you can judge whether a given filing is worth the effort. Refolk's index is one way to size a slice. In Refolk's index of professional profiles, US profiles with a title containing "Software Engineer" at the entry-level seniority band number 353,689.

Those profiles concentrate in predictable places. From a 25-profile sample of that slice:

Read this table as indicative, not precise. The 353,689 base count is a real figure from Refolk's index; the region shares are derived from a 25-profile sample and should be treated as a rough signal of where the pool sits, not a census. The practical point is that engineering talent clusters in the same metros where large employers file WARN notices, so a California or Seattle filing sits on top of a dense addressable pool. I will flag an honest gap: planned comparison slices at the senior band and for product managers returned zero resolvable matches at query time, so this sizing reference holds only for the entry-level software-engineer slice.

Keeping the queue lawful and current

Compliance is a sourcing step, not an afterthought. The standard lawful basis for sourcing and first contact under GDPR is legitimate interest, not consent, supported by a documented three-part legitimate-interest assessment and an Article 14 "data obtained from other sources" notice on first contact. Active sourcing is compatible with GDPR when you use only necessary professional data for a defined role, document the assessment, and tell the person what is happening. It is not permission to build hidden lists or use any channel.

Regulators enforce this directly. The Irish Data Protection Commission fined LinkedIn EUR 310 million in October 2024 for invalid consent and unlawful legitimate-interest claims for behavioral profiling, against a GDPR ceiling of up to EUR 20 million or 4 percent of global annual turnover. For California residents, CCPA requires updating privacy notices and creating a process to honor deletion requests.

Your Article 14 line can be short and reusable.

I found your profile through public professional records after a WARN layoff notice at [company]. I am reaching out about a [named role] on my team because your background fits it. I hold your professional details under legitimate interest for recruiting; you can ask me to delete them at any time.

Run this verification before you call any WARN-sourced queue finished.

Keeping it current

WARN data decays fast and formats shift without warning. Re-pull filings on a fixed cadence rather than treating any snapshot as durable, and re-check each live cohort for an amendment within the 45-day window after the original filing. When a state changes its format or an archive disappears, fall back to a deduplicated aggregator that links amendments, so you are not blindsided by a withdrawn notice you already built a queue against. The discipline that keeps this work honest is simple: the filing is a plan, the profile is the evidence, and the effective date is the clock.