# The Layoff Sourcing Playbook: From WARN Filing to Ranked Queue

*Run a weekly process that turns WARN and 8-K filings into a verified, scoped, ranked queue of reachable laid-off candidates while they are still on the market.*

- Canonical URL: https://www.refolk.ai/guides/layoff-sourcing-playbook
- Pillar: Recruiting and sourcing
- Format: Playbook
- Published: 2026-09-08
- Last reviewed: 2026-09-08
- Reading time: 15 min

A company is cutting headcount and you have a narrow window to turn that into a warm, ranked list of reachable people before your competitors do. This playbook is for in-house recruiters, sourcers, and talent leads who want a repeatable weekly process that starts from verifiable public filings, scopes which team and office were actually hit, and ranks cuts by relevance before anyone spends a minute sourcing. It covers the involuntary-reduction case specifically, with the timing math and verification steps that decide whether the effort pays.

Most public advice on this is a pile of tips plus a bookmark list of tracker sites. What is missing is the end-to-end procedure: how to read a WARN notice against an 8-K, how to reject a phantom cut, and how to sort events so you reach strong candidates in the first two weeks rather than the last. That is what follows.

## Why layoff sourcing beats ordinary cold sourcing

The filing itself is a quality filter you get for free. WARN notices and SEC Form 8-K Item 2.05 only fire on group reductions, never on individual performance exits, which are never filed. That means everyone in a filed cohort left involuntarily and for reasons that have nothing to do with their work, which is the exact opposite of the adverse-selection problem that haunts ordinary cold sourcing.

The second advantage is timing. A WARN notice reaches affected workers at least 60 calendar days before their separation date. Read that as a warning and you do nothing; read it as a sourcing head start and you have a two-month lead on people who are still employed, still reachable, and about to be on the market. Most recruiters ignore it because they wait for the announcement to make the news.

**137 - US software engineers with an open-to-work signal in Refolk's index**

The Germany equivalent is 4 - a 34x supply gap that makes a US-anchored layoff playbook far denser to work.

The catch is that this supply is wildly geo-concentrated. In Refolk's index, 137 US software engineers carry an open-to-work signal against just 4 in Germany. Part of that gap is real market size, but part is that EU notice rules and GDPR suppress open public job-search signals, which is why the same playbook produces a fraction of the reachable names in Europe. Plan your effort where the supply is.

## The three signal streams and their lead times

Three public streams carry the same event at different speeds and with different blind spots. WARN gives the longest runway and the exact site. The 8-K gives the percentage and function fast. Trackers surface the news in near real time but unofficially. You need all three because each one covers a case the others miss.

| Source | Lead time before separation | Update cadence | Coverage limit |
|---|---|---|---|
| WARN notice | Up to 60 calendar days | Per state portal (varies) | Only 100+ employee firms, 50+/site cuts |
| SEC 8-K Item 2.05 | Days after board commits | Within 4 business days of estimate | Public companies with a material charge |
| Tracker (layoffs.fyi) | After public announcement | Continuous/live | Mostly tech, self-reported |
| Challenger report | After the month | Monthly | Aggregate counts, not names |

WARN is the earliest and most precise on location, but it only covers larger employers and larger cuts. The 8-K is fast once a board commits, and if no cost estimate is ready it must follow within four business days after the estimate is formulated, so watch for the amendment. Trackers are near-live but self-reported and skewed to tech. Challenger's monthly report gives you the macro tide, not names, so treat it as context rather than a lead source.

#### How one cut moves across the three streams

1. **Board commits** - 8-K Item 2.05 files within days, naming a percentage and a function
2. **Notice served** - WARN notice hits the state portal up to 60 days before separation, naming the site
3. **News breaks** - Tracker logs the event in near real time, unofficially and self-reported
4. **Month closes** - Challenger report aggregates counts for macro context, no names

*The same layoff appears earliest and most precisely in WARN, fastest in the 8-K, and loudest in the tracker.*

## Who has to file, and who never will

Federal WARN covers employers with at least 100 employees excluding part-time workers, or 100 or more employees who together work at least 4,000 hours per week. It requires 60 days' advance notice to affected employees, states, and localities. Below that size, or below the event thresholds, nothing is filed and your WARN feed stays silent even when real cuts happen.

The event triggers, measured across a 90-day window, are the part people misremember, so keep them where you can see them.

> **Rule:** The WARN event thresholds
>
> A single-site closure hitting at least 50 people, a layoff of at least 33% of a single site's workforce affecting a minimum of 50 employees, or any planned layoff affecting 500 or more employees regardless of percentage.

Two consequences follow. First, a 40-person cut at a small firm will never appear in WARN, so an empty WARN feed does not mean no layoffs - it means no layoffs above the floor. Second, 15 states have layoff notification laws that supplement or expand federal WARN, lowering the threshold, so a cut invisible federally may still be filed in a mini-WARN state. Check the state, not just the federal rule.

## Verify and scope: turning two filings into a target

No single documented standard exists for this, so the working method is cross-referencing filings until each supplies what the other lacks. The 8-K carries the headcount and function; WARN carries the site, city, and dates. Neither is sourceable alone.

Take LiveRamp's 8-K, which disclosed a workforce restructuring involving approximately 65 full-time employees, representing approximately 5% of the company. That 5% figure is company-wide. It tells you nothing about which office to source until you match it against the WARN single-site count for a specific city. The 8-K gives the strategic shape; WARN gives the address.

The across-the-board versus single-function read changes how you weight candidate quality. Solexa's 8-K described a workforce reduction of approximately 17% that included positions in most functional areas. That is a broad cut with no strategic signal about any one team. A cut concentrated in one function or one site reads instead as a deliberate exit of that unit, which means a denser, more coherent cohort of one skill in one place. This read matters more as reductions increasingly hit specific automatable functions; in the Challenger data, AI was cited in 116,175 job-cut announcements year to date, about 22% of all cuts.

> **Watch out:** Company-wide percentage is not your site
>
> The 8-K might say 5% overall while your target office was untouched, or hit far harder. Match the 8-K percentage and function to the WARN single-site headcount before you assume your city was affected.

## The window: how long candidates stay reachable

The tightest figure is the one to plan around: reemployed workers took 7 weeks on average to find their new jobs. Broader BLS-based measures run longer, but the seven-week average tells you that a meaningful share of a strong cohort converts fast, so your outreach has to start within days of the filing, not after the news cycle.

| Metric | Value | Source |
|---|---|---|
| Average weeks to reemployment (survey) | 7 weeks | ZipRecruiter research |
| Median search (BLS-based) | 11.6 weeks | Careerminds |
| Average search (BLS-based) | ~26 weeks | Careerminds |
| Share searching >15 weeks (Oct 2024) | ~40% | CBS News |

The two families of numbers are not in conflict. The seven-week survey figure describes people who got reemployed; the 26-week average and 11.6-week median describe the full population of searchers, including those still looking. For sourcing, the seven-week figure is the operative one because it describes the people you are competing for. Front-load your outreach to the first two weeks of the window.

The market has been lengthening. In late 2024, about 40% of the 7 million out-of-work people, roughly 2.84 million, had been job-hunting more than 15 weeks, up 20% year over year. That cuts in an unexpected direction. When hiring is slow, cohorts sit on the market longer, so ranking by recency matters more, not less, because the freshest cuts hold the people who have not yet grown discouraged or accepted a lesser offer.

> Work the separation date, not the announcement date - the two-month gap is your entire advantage.

## The weekly procedure, start to finish

Run this once a week. The whole loop is roughly two to three hours of coordinated work plus one to two hours of sourcing per event you decide to work. Verify before you rank: some practitioners rank first to triage volume, but the safer order verifies first so you never spend judgment on a phantom cut.

#### The layoff-to-queue loop

1. **Monitor the filings feed** - Pull this week's new WARN notices by state, SEC 8-K Item 2.05 filings, and one live tracker. Roughly 30 minutes; done when you have a dated list with employer, headcount, and site.
2. **Filter to coverage** - Drop events under WARN thresholds unless a mini-WARN state applies. Roughly 15 minutes; done when only cuts large enough to yield a reachable cohort remain.
3. **Verify each cut** - Confirm the event appears in two independent sources and reconcile headcount and dates. Roughly 20 minutes per event; done when no single-source or rumored cuts survive.
4. **Scope function, office, headcount** - Cross-map the 8-K percentage and function to the WARN single-site count. Roughly 20 minutes; done when you know which team and city to source and how many people.
5. **Rank the cuts** - Score by relevance to open reqs, headcount in your function, geography, and days since separation. Roughly 20 minutes; done when you have a ranked queue, best first.
6. **Time the window** - Prioritize events whose separation date is recent, since strong candidates convert in about 7 weeks. Ongoing; done when events are sorted to hit people while still on the market.
7. **Build the candidate queue** - Identify affected individuals via public profiles and open-to-work signals, capturing only job-related data. One to two hours per event; done when you have an outreach-ready list with a lawful basis logged.
8. **Log lawful basis and send notice** - Record the legitimate-interest reasoning and send a privacy notice within one month, honoring any objection or erasure. Same day; done when every contact is documented and compliant.

Step seven is where the identification work happens, and it is the slowest stage by far. Instead of hand-searching profiles for everyone who left a named company in a named window, you can describe the cohort in plain language and get back the reachable, open-to-work subset directly.

I ran this search: `Software engineers laid off from Oracle or Dell in the last 8 weeks who are open to work in the US.` - [see the full result list](https://www.refolk.ai/s/ds6vp06bpf).

*Returns US software engineers tied to those employers who carry a recent open-to-work signal, so you skip the manual profile-by-profile hunt and go straight to a queue.*

[Refolk](/) is worth reaching for at exactly this stage because the friction the paragraph just described - matching a named employer and a time window to public open-to-work signals across many profiles - is the part that eats the hour. In Refolk's index, the top US employers in the open-to-work software cohort included Amazon, Google, Uber, and Oracle, which is the same population these filings put on the market.

## How this goes wrong

The failure modes here are specific and each one has a false positive that looks like success. Treat this section as the checklist you run before you trust your own queue.

| Failure mode | What it looks like | The check |
|---|---|---|
| Sub-threshold blindness | An empty WARN feed read as "no layoffs" | Monitor 8-K and trackers alongside WARN, and check mini-WARN states |
| Phantom cut | A confident headcount no filing supports | Require two independent sources before ranking |
| Stale separation date | Sourcing a cut where strong people are already placed | Rank by days since separation, not announcement |
| Wrong site or function | Assuming your office was hit from a company-wide percentage | Match the 8-K percentage to the WARN single-site count |
| Date confusion | Contacting before people have actually left | Read the layoff effective date, not the notice date |

**Sub-threshold blindness** is the quietest failure because nothing tells you it happened. A 40-person cut never files a WARN notice, so a silent feed reads as calm when it is really just below the floor. The 8-K and trackers catch cases WARN misses.

**Phantom cuts** come from a tracker entry or a rumor with no filing behind it. You end up with a headcount and a company name that no official record supports, which is why verification against a second independent source is a hard gate, not a nicety.

**Announcement-versus-effective-date confusion** is the subtle one. WARN dates are the future separation date, not the date the notice was posted. Contact people before they have actually left and you burn goodwill and the window at once. The separation date is a field on the filing; read it every time.

> **Tip:** Sort your queue on one field
>
> Once cuts are verified and scoped, sort the whole queue by days since the separation date ascending. The freshest cohorts rise to the top automatically, and the seven-week average does the prioritizing for you.

## Sourcing lawfully: notice, basis, and no stockpiling

For EU and UK candidates, the lawful basis for a first cold contact is legitimate interest, not consent, because the candidate did not give you their information directly and none of the other bases will cover it. You cannot treat a public open-to-work signal as consent, and you cannot bulk-save profiles for later.

The stockpiling ban is explicit: building your talent database by adding candidate data in case you need it in the future is not legal under GDPR. You must plan to contact candidates as soon as possible, and you can only hold a candidate's data without informing them for a limited time, typically one month. The notice obligation runs alongside: inform a sourced candidate that you hold their data, typically within 30 days of collecting it and before you use it for any purpose.

> **Rule:** Contact intent is the line between lawful and not
>
> Sourcing a specific cut for a specific, near-term contact is legitimate interest. Saving the same profiles into a general pool with no plan to reach them is stockpiling and is not lawful. Log the intent to contact when you capture the data.

Enforcement is real and large. The Irish Data Protection Commission fined LinkedIn EUR 310 million in October 2024 for relying on invalid consent and unlawful legitimate interest claims for behavioral profiling. US CCPA specifics for candidate data are not established in my sources here, so verify your US obligations against current guidance rather than assuming they mirror GDPR. Use legitimate interest for outreach and reserve consent for any special-category data.

**Legitimate-interest log entry (one per candidate)**

```
Candidate: [name / public profile URL]
Source: [WARN filing state + date] + [second source]
Data captured: name, current employer, role, public open-to-work signal
Purpose: outreach for [named open req], near-term contact
Basis: legitimate interest (involuntary group reduction, candidate reachable and job-searching)
Notice sent: [date, within one month of collection]
Objection/erasure handled: [yes/no + date]
```

*Adapt the fields to your ATS. The point is a dated, defensible record you can produce on request.*

Send the notice inside the one-month window, honor any objection or erasure request on receipt, and keep the log current. That record is what makes the whole playbook repeatable rather than a one-off risk.

#### Before you call the queue done

- [ ] Every event is confirmed by two independent sources
- [ ] Each 8-K percentage is matched to a WARN single-site headcount
- [ ] The queue is sorted by days since the separation date, freshest first
- [ ] No candidate was captured without a logged intent to contact
- [ ] A privacy notice is scheduled or sent within one month of collection
- [ ] Sub-threshold cuts were checked against mini-WARN state laws
- [ ] You read the effective separation date, not the notice date, for every event

## Keeping the process current

The playbook is stable, but its inputs drift. State WARN portals change format and cadence, so re-confirm each state feed you rely on every quarter. Mini-WARN coverage shifts as states pass laws, so re-check the list of states with notification laws rather than trusting a memorized figure. And the reemployment window widens and narrows with the hiring market: when Challenger reports high monthly cut totals, cohorts sit longer and your recency ranking earns more; when totals fall, the window tightens and speed matters more. Re-read the macro figure each month and let it set how aggressively you front-load outreach.

The durable core does not move. Filings only fire on group reductions, so the cohort is always involuntary. WARN always leads separation by up to 60 days, so the head start is always there for whoever works the date. And supply is always concentrated where public job-search signals are open, which the index numbers make plain. Build the weekly loop around those three facts and re-verify the rest on a schedule.

## Frequently asked questions

### Where do I find WARN notices for free?

Each US state runs its own WARN portal, and the US Department of Labor Rapid Response page links coverage rules. Structured aggregators re-publish state filings in a common schema of company, location, headcount, dates, and NAICS code. For public companies, cross-check SEC EDGAR full-text search for 8-K Item 2.05 filings. Pull all three weekly rather than relying on any single feed, because each has a different coverage limit.

### How fast do I have to move after a layoff is announced?

Start within days. Reemployed workers took 7 weeks on average to find new jobs, so a meaningful share of a strong cohort is placed before week seven. Work the separation date on the filing, not the announcement date, and front-load your outreach to the first two weeks of the window. In a soft hiring market cohorts sit longer, which raises the payoff of ranking by recency rather than lowering it.

### What size of layoff actually shows up in a WARN filing?

Federal WARN covers employers with at least 100 employees excluding part-time, or 100 or more working a total of 4,000 hours per week. It triggers on a site closure hitting 50 people, a cut of 33 percent of a single site's workforce affecting at least 50 employees, or any cut of 500 or more. Smaller cuts and sub-threshold employers never file, though 15 states have mini-WARN laws that lower the bar.

### Is it legal to source laid-off candidates without their consent?

For EU and UK candidates the first-contact basis is legitimate interest, not consent, since the person did not give you their data directly. You must inform them you hold their data, typically within 30 days of collecting it and before you use it. Building a just-in-case pool is unlawful, and enforcement is real: the Irish DPC fined LinkedIn EUR 310 million in October 2024. US CCPA specifics for candidate data are not established in my sources.

### How do I tell an across-the-board cut from a targeted team shutdown?

Read the 8-K language. An across-the-board reduction names a percentage and spans functions, like Solexa's 17 percent cut that included positions in most functional areas. A cut concentrated in one function or site signals a strategic exit of that unit and changes how you weight candidate quality. This read matters more as AI-driven cuts increasingly cluster in specific automatable functions.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/guides/layoff-sourcing-playbook*