The Employer Layoff Exposure Standard, Seven Records to Go, Probe, or Avoid

This standard is for a candidate about to spend a week on an application loop, or sign an offer, at an established employer, who wants to know from public records whether the site or team they would join is about to be cut. It grades the employer on seven criteria, each backed by a record you can cite, and ends in a Go, Probe, or Avoid verdict with one evidence row per criterion that you log next to the application. It is built so that two people grading the same employer on the same day reach the same verdict.

What does the layoff exposure standard grade?

It grades whether public records point to a cut at the specific legal entity, site, and function in your posting. It does not grade whether the company as a whole is healthy. A profitable company can close the one office you would work at, and a struggling one can still be growing the team you would join.

Five terms carry the whole standard:

• WARN notice: the written advance notice that the federal Worker Adjustment and Retraining Notification Act, and the state versions of it, require before a qualifying plant closing or mass layoff.
• Item 2.05: the section of an SEC Form 8-K that a public company files when it commits to an exit or restructuring plan that will carry material charges.
• Requisition: an approved opening the employer is recruiting against. The careers page shows the posted ones.
• Backfill vs net-new: a backfill replaces someone who left. A net-new role adds headcount that did not exist before.
• Legal entity: the company that actually employs you. It is often a subsidiary with a different name from the brand on the careers page.

No single record covers all of these. Each one sees a single layer of the organisation, which is why the standard uses seven criteria and not one.

The seven criteria, and how each one lies

Each criterion is graded Clear, Flag, Unknown, or N/A, and each grade must point to a dated record. Below, each criterion says what it proves, what it looks like when it lies, and the grading rule.

1. WARN filings by legal entity and site

• Proves: a qualifying layoff or closing was planned at a named address, with a count and a date.
• Lies when: you search the brand and the notice was filed by a subsidiary, or the cut fell below the threshold, or the state does not publish notices.
• Grade: Flag if any notice in the trailing 24 months names one of your entity names or your site address. Otherwise Unknown. This criterion never grades Clear, because an empty search cannot prove that no cut is coming.

2. SEC Item 2.05 and 5.02 filings

• Proves: the board or an authorized officer committed the company to an exit plan with material costs.
• Lies when: the filing covers a segment that was sold or that has nothing to do with your function.
• Grade: Flag if an Item 2.05 filing appears in the trailing 24 months. Clear if a 24-month search of a public company finds none. N/A for a private employer. Item 5.02 covers officer and director departures. I have not verified its filing rules, so log any 5.02 hits as context and never grade on them alone.

3. Careers-page requisition trend

• Proves: whether the employer is still recruiting in your function.
• Lies when: the drop happened after a cut was already announced, so it confirms the cut rather than predicting it, or when the page cannot be filtered by function.
• Grade: Flag if the later of two dated snapshots in your function is under half the earlier one. Clear otherwise. Unknown if you have only one snapshot or the page cannot be filtered. The half-cut is a grading convention so that two graders agree. It is not a tested predictive threshold.

4. Guidance changes

• Proves: management lowered its own outlook. Ranking pages put financial signals like this first among layoff warnings.
• Lies when: the guidance cut sits in a division far from yours.
• Grade: Flag if guidance was lowered or withdrawn in an 8-K exhibit in the trailing 12 months. Clear if it was held or raised. N/A for a private employer.

5. Function named in restructuring

• Proves: the cut reaches the team or site you would join.
• Lies when: the filing names no segment at all. That is common, because the required fields are costs.
• Grade: Flag if any Item 2.05 filing, its exhibits, or a WARN notice in the trailing 24 months names your segment, function, or site. Clear if the filings name only other segments, or if no filings exist. Unknown if a filing exists but names no segment.

6. Backfill vs net-new requisition

• Proves: the role has approved headcount and a reason to exist.
• Lies when: the person answering does not work for the employer and is passing on what they were told.
• Grade: Flag if the recruiter says the function has stopped backfilling, or if the role used to be full-time and is now listed as contract. A recruiter-written source puts hiring activity like this first among warning signs. Clear on a direct answer of backfill or net-new against approved headcount. Unknown if you get no answer.

7. Recency of last cut

• Proves: how recently the employer used layoffs at all, at any site.
• Lies when: the last cut happened in a non-publishing state or fell below the threshold, so it left no record.
• Grade: Flag if the most recent recorded cut (WARN or Item 2.05) is under 12 months old. Clear if it is older, or if none appears in 24 months of searchable records. Unknown if the relevant state does not publish.

WARN thresholds by jurisdiction: why "no notice" is not a pass

A missing WARN notice proves nothing on its own, because the law only requires notice above a size threshold at a single site, and three states publish nothing. The thresholds in the table decide whether a cut at your site would ever produce a notice you could find.

Sources: the CRS WARN primer, the DOL employer's guide, and state-law summaries. State rules change, so re-check the site state's statute before you rely on a row. Nebraska's rule is the newest in the table.

Three counting rules that hide cuts from outsiders

1. Sites are never added together. Each trigger is limited to one site. Ten sites each losing 40 people is 400 jobs and produces no federal notice.
2. The newest hire is notified but not counted. Anyone averaging under 20 hours a week, or employed fewer than 6 of the last 12 months, counts as part-time. Part-time employees are owed notice, but they do not count toward the threshold. A cut aimed at recent hires therefore needs more people to reach 50.
3. Government employers are exempt. Federal, state, and local government employers fall outside the federal rule entirely.

The federal law also has a list of exceptions, such as unforeseeable business circumstances. New Jersey drops some of them. I have not checked the federal exceptions list against a primary source, so treat any claim that a specific exception applies as unverified.

Where the notices are published

There is no federal WARN database. States may publish notices but are not required to. One coverage dataset counts 48 publishing jurisdictions out of 51. Arkansas, New Hampshire, and Wyoming publish nothing, so a "none found" result there means nothing.

Formats vary by state:

• New York: a dashboard plus a notice table.
• Texas: a downloadable spreadsheet.
• Illinois: a list organised year by year.
• California (EDD): a periodic report.

The posting lag between filing and publication has not been established from official pages. One third-party monitor reports updates arriving anywhere from daily to in weekly batches, with no feed. One public aggregator holds 45,462 notices from 29 states, matched to 1,277 public companies. That is useful for a first pass, but 29 states is well short of 48, so always search the official list for the site state and the HQ state.

How to read an Item 2.05 restructuring filing against your posting

Read an Item 2.05 filing for its date, its cost, and the segment it names, because the rule requires cost disclosure and nothing else. Any headcount, percentage, or site detail is something the company chose to add.

What the rule requires

• Trigger: the board, a board committee, or an authorized officer commits the company to an exit plan or a plan of termination that will carry material charges.
• Deadline: within four business days of that commitment.
• Required fields: an estimate for each major type of cost, the total, and the cash portion.
• Late estimates: if the costs are not yet known, the company must file an amendment within four business days of working them out.
• Not required: headcount, percentage of workforce, sites, or functions.

That last point is why criterion 5 often ends up Unknown for public companies. The filing tells you money is being spent on an exit. It does not tell you whose exit.

How to search for it

Use EDGAR full-text search. It covers filings from May 4, 2001, and filters by filing category, date range, and executive-office location. Results are capped at 10,000.

The search backend ignores an item-number parameter. If you rely on an item filter you get every 8-K the company filed. Put the item in quotes as a literal phrase instead.

Search phrase:  "Item 2.05" "Example Holdings Inc"
Filing category: 8-K (include 8-K/A to catch amended cost estimates)
Date range:      trailing 24 months from today
Log per hit:     filing date | total cost | cash portion | segment named | headcount if volunteered | 8-K/A follow-up date

Reading it against the posting

1. Find the segment, division, or geography the filing names, if it names one.
2. Find the segment your posting sits in. Use the job family, the reporting line in the posting, and the site address.
3. If they match, criterion 5 is a Flag. If the filing names a different segment, including one that is being sold, criterion 5 is Clear for that filing. If it names nothing, criterion 5 is Unknown.
4. Check for an 8-K/A. The first filing may say the costs are not yet determined, and the amendment may carry more detail.

For a private employer, none of this exists. State WARN rules still apply to private employers. Other private-company sources, such as Form D, court and bankruptcy dockets, and contract award end dates, were not researched for this standard, so I do not grade on them.

Posting drawdowns and tenure: what the evidence supports

The evidence shows that postings can collapse before a cut and that short-tenure workers look over-represented among the displaced. It does not establish a reliable lead time or odds by tenure, and the standard is set conservatively for that reason.

Requisition drawdowns around cuts

Sources: a job-listings analytics firm's analysis for Oracle and Coinbase, and a posting-data analysis for Amazon. The change column is derived.

What this supports:

• In the two documented pre-cut cases, listings fell about 80%. Coinbase's fall came before a cut of about 700 people, roughly 14% of staff.
• Only the Oracle case is dated, at roughly a month before terminations. No systematic study of lead time was found.
• The Amazon drop came after the announcement. It confirms the cut rather than predicting it, which is why criterion 3 needs dates you can check against the 8-K.

That is too little to set a predictive threshold. The half-cut rule in criterion 3 exists so that graders agree with each other. It is not a forecast.

Tenure among displaced workers

Sources: the BLS displaced worker and employee tenure releases. The 41%, 59%, and 34.3% figures are derived.

BLS does not publish displacement odds by tenure, so a true risk multiple is not established. One more figure gives context: 22% of wage and salary workers had a year or less with their current employer at the last tenure survey. Being the newest person on a team is common. Being the newest person on a team that is about to be cut is what this standard exists to catch.

The grading procedure, from legal entity to verdict

The procedure takes about three hours of desk time spread over at least a week, because the requisition trend needs two snapshots. The time estimates are my own and are not sourced. Run it before a multi-round loop starts, and again before you sign.

Log the result in a fixed format so that a second grader can audit it line by line.

Employer (brand):
Legal entities searched:
Site address / state:
HQ state:
Graded on (date):
Grader:

# | Criterion                 | Evidence (source, date, figure)              | Grade
1 | WARN by entity and site   |                                              | Flag / Unknown
2 | Item 2.05 (5.02 context)  |                                              | Clear / Flag / N/A
3 | Requisition trend         | snapshot A: date, count; snapshot B: date, count | Clear / Flag / Unknown
4 | Guidance change           |                                              | Clear / Flag / N/A
5 | Function named            |                                              | Clear / Flag / Unknown
6 | Backfill vs net-new       | answer verbatim; answerer's employer         | Clear / Flag / Unknown
7 | Recency of last cut       | date of last recorded cut                    | Clear / Flag / Unknown

Flags: _  Unknowns: _  Verdict: Go / Probe / Avoid
Re-grade before signing on (date):

The log is only useful if it sits next to the application it governs. I keep it in the same record as the tailored resume, so that when Refolk scores how well I fit a posting, the exposure verdict is in the same row and a strong fit at an Avoid employer is visible before I spend the week.

Verdict rules: Go, Probe, or Avoid

The verdict comes from counting Flags and Unknowns under fixed rules, not from overall impression. These rules are a policy choice made for consistency. They are not calibrated against outcomes, because the public evidence to calibrate them does not exist.

• Avoid: criterion 5 is a Flag (a recorded cut names your function or site) and criterion 3 or criterion 7 is also a Flag. Or four or more criteria are Flags.
• Probe: at least one Flag that does not meet the Avoid rule, or three or more Unknowns.
• Go: zero Flags, and no more than two Unknowns, each with a logged reason.

Because criterion 1 can only grade Flag or Unknown, a Go almost always uses one of its two allowed Unknowns on a WARN-silent row. That is intended. For a private employer, mark criteria 2 and 4 N/A, which does not count as Unknown, and apply the same rules to the remaining five.

What a Probe asks

A Probe verdict means you keep going but ask specific questions before you invest the week or sign. Send them in writing so that the answer becomes an evidence row.

Thanks for the time so far. Before I commit to the full loop, a few questions about the role itself:
1. Is this role a backfill for someone who left, or new headcount added this year?
2. Is the headcount approved, or pending a planning cycle?
3. Is the team still backfilling other departures right now?
4. Was this role previously posted as full-time and is it now contract, or the reverse?
5. Was this team or site affected by the restructuring the company announced, and if not, which part of the business was?
6. Are you employed by the company directly, or recruiting on its behalf?

Question 6 matters more than it looks.

In that sample, the two most common employers were not the household-name companies. If the person answering your backfill question does not work for the employer, their answer is secondhand. Log who they work for, and grade a secondhand "net-new" as Unknown until someone at the employer confirms it.

How this standard goes wrong

Most failures in this standard are false passes: a silent record read as a clean one. Each failure mode below has a check, and the checks are built into the procedure. Skipping them is how a Go verdict ends up attached to a site that closes the month after you start.

1. The legal-entity trap

• What happens: you search the brand name, the notice was filed by a subsidiary, and you log a false "no notice".
• Check: search every entity name from step 1, and search the site street address on its own.

2. Silence below the threshold

• What happens: a cut of under 50 at one site, or one spread across several sites, never produces a notice. Recent hires do not count toward the threshold, which makes this more likely for a cut that falls on them.
• Check: criterion 1 never grades Clear. Log it as WARN-silent with a reason.

3. The non-publishing state

• What happens: you search Arkansas, New Hampshire, or Wyoming and log "none".
• Check: confirm that the state publishes before you log any result. If it does not, grade criteria 1 and 7 Unknown.

4. Item 2.05 read at the wrong scope

• What happens: the filing discloses costs, not sites. A filing about a segment being sold gets read as a cut to your function, or a cut to your function hides inside a total with no segment named.
• Check: match the segment named in the filing to the posting. If no segment is named, grade criterion 5 Unknown, not Clear.

5. The item filter that filters nothing

• What happens: you rely on an item-number parameter in the EDGAR search, it is ignored, and you either drown in every 8-K or wrongly conclude that the matching filing was among them.
• Check: search "Item 2.05" as a quoted literal phrase.

6. Index artifacts that read as absence

• What happens: a filtered search returns zero and you take the zero as fact. In Refolk's index, the Senior filter on the same recruiter title returned 0 results in both the US and the UK. That is almost certainly a gap in how seniority is labelled, not a real absence of people.
• Check: before concluding anything from a zero, rerun the search without the narrowing filter. The same discipline applies to careers pages whose function filter silently drops listings.

7. A posting collapse after the fact

• What happens: you see an 80% drawdown and treat it as a warning, but it came after the announcement, as Amazon's did.
• Check: date both snapshots and compare them with the 8-K filing date. A drop after the filing supports criterion 7, not criterion 3.

A public record of silence is not a record of safety. Grade it Unknown and keep asking.

Keeping the verdict current

A verdict is only valid on the date it was graded, so it carries a re-grade date and specific triggers. Item 2.05 filings can land within four business days of a decision, and state lists post on their own uneven schedules, so a Go from the start of a long loop can be stale by offer day.

Re-grade when any of these happens:

• Before you sign. Rerun steps 2, 3, and 5 at minimum. The offer is the last point at which the evidence is cheap to act on.
• An 8-K or 8-K/A appears. An amendment can add the cost detail, and sometimes segment detail, that the first filing left out.
• The posting changes. A switch from full-time to contract, a new site address, or a different hiring entity on the offer letter sends you back to step 1.
• The site's state changes its law. Thresholds and notice periods differ by state and change over time, so check the statute when you start, not once a year.

Once the verdict is logged, the hard part of the decision is done and the week of application work either starts or does not. For a Go or a Probe you decide to pursue, that week is mostly tailoring. Refolk rewrites the resume against each posting and scores the fit, which leaves the exposure log as the one piece you still update by hand, on the dates it says to.