The Employer Hiring-Health Score, Before You Spend Applications

You have a list of target companies and a finite number of hours. Some of those employers are hiring hard, some are quietly frozen with the postings left up, and the ones that look busiest are sometimes the ones in the most trouble. This guide is for a job seeker who wants to score each employer on public hiring-health signals and sort it into target, watch, or skip before writing a single tailored application.

Other standards in this library score one posting's liveness or spot red flags once you are already in the loop. This one scores the employer itself, across six dimensions you can observe from outside, and turns the scattered "search for layoffs" advice into one weighted score with cutoffs you can act on the same evening.

Why score the employer, not the posting

Scoring the employer catches the failure mode a single-posting check misses: a whole company that is frozen while its listings stay live. A posting-liveness check tells you whether one ad is dead. An employer-health score tells you whether the machine behind every ad on the list is actually hiring.

The evidence for scoring the employer is blunt. In any given quarter, 18 to 22% of the jobs posted on the Greenhouse platform are classified as ghost jobs, and nearly 70% of the companies on that platform posted at least one ghost job in a single quarter. On the survey side, 81% of recruiters admit to having posted ghost jobs and over a third say up to a quarter of their active listings are fake. The waste is real: the average ghost-job cycle burns about nine hours of a job seeker's time. Spend that on the wrong employer and you have lost a working day.

The counterintuitive part is the motive. Companies post ghost jobs to keep a presence on job boards and to look viable during a freeze. That means ghost jobs cluster at struggling firms, the opposite of what a job seeker infers from a long careers page. A wall of open roles is not proof of growth. It can be the tell of contraction. So the score treats posting volume with suspicion and leans on signals that are harder to fake: headcount trend, posting age, funding, layoff records, and leadership churn.

The six signals and what each one proves

The score runs on six public signals. Each proves something specific, and each has a way it lies to you. Know both before you weight them.

Two of these deserve special care because they invert. Posting volume reads high at exactly the wrong employers, as above. And recruiter presence reads high too: in Refolk's index the top US recruiter employers are staffing agencies such as Experis, Robert Half, and K2 Partnering Solutions, so an outsider counting "recruiters" overstates in-house hiring capacity unless they filter to the target employer's own talent team.

That team is smaller than most people assume. In Refolk's index there are only 487 director-level in-house talent leaders across the entire United States. "They have recruiters" is a weak signal until you confirm the recruiters actually work there.

How the signals rank by timing and certainty

Signals split into leading and lagging, and the split governs how you weight them. Leading signals surface a freeze early but are inferred and noisy. Lagging signals are certain but arrive late, after the door has closed.

A hiring freeze shows up first in budget approvals and hiring behavior, not headlines. Companies stop converting interns, delay or cancel approved roles, and quietly raise experience bars before any layoff is announced. From outside you cannot see budget approvals, so the observable leading signals are posting age, headcount trend, and a collapse in recruiter response. The lagging signals are the public layoff headline and the WARN notice, which are settled but slow.

The practical rule: weight leading signals for timeliness but lagging signals for certainty. A stale-posting-plus-flat-headcount pattern with no layoff on record is a watch, not a skip, because you are seeing the early tell without confirmation. A dated layoff on a named tracker is a skip because it is settled.

Reading posting age and the ghost-job discount

Posting age is the cheapest high-yield filter you have, so run it first among the postings. In an analysis of 100,000 listings across major boards, more than 40% showed no meaningful human interaction in over 30 days despite still appearing open. Thirty days is where "open" and "live" diverge.

Tag each posting fresh, stale, or repost. If an ad was posted six months or a year ago, assume the role was filled and the ad neglected, or that the company keeps it up to collect resumes. A single repost is normal; a role surfacing three or more times warrants caution. Watch for the permapost trap: high-turnover retail and customer-service roles may be always open and always look fresh, so check repost frequency, not just the date.

For the ghost-job discount, do not apply a flat 20% across the board. Sector matters, and some run far above baseline.

Applying the 20% baseline to a construction or legal employer over-scores it badly. Use the sector rate where you have one, and the 20% working discount elsewhere.

The other reason to lead with direct verification: recruiter silence is often capacity, not a dead role. Greenhouse data shows recruiter workload rose to 588 applications in a single quarter, a 26% year-over-year jump. A recruiter buried under 588 applications is not ignoring you on purpose, which is why a direct probe beats waiting.

Refolk can confirm a real in-house hiring team exists before you invest, which is the check that separates a frozen employer from a busy one.

The scoring procedure

Run these steps per employer. As an outsider, posting age and headcount are the only fully observable dimensions, so lead with those even though internal signals fire earlier. Everything below is doable in an evening for a shortlist.

Here is a scoring rubric you can copy and keep in a spreadsheet. Weight it toward the signals that lie least.

Posting age: fresh <30d = +2 | stale 30-45d = 0 | 45d+ or 3x repost = -2
Headcount trend (your dept, 6-12mo): ~20%+ YoY = +3 | flat = 0 | declining = -3
Discounted posting volume: growing across depts = +1 | flat = 0 | many roles + flat headcount = -1
Funding / press: recent raise or expansion = +2 | none = 0 | workforce-optimisation language = -2
Layoff record: clean (cross-checked) = 0 | one dated cut in 12mo = -3 | repeated cuts = skip outright
Leadership churn (your team): stable = 0 | one recent exit = -1 | HM + dept head gone = -2
Direct probe: active reply = +2 | no reply = 0 | evasive = -1
CUTOFFS: score >= +4 = TARGET | 0 to +3 = WATCH | below 0 or any dated layoff = SKIP

The cutoffs are a starting point, not a law. If you have five slots and only two employers clear +4, promote your best watch and probe it directly before committing.

How this score goes wrong

The failure modes below are where the score lies to you, and they matter more than the happy path. Each has a check that catches it.

The headcount count is stale or inflated. Public employee counts track associated members, not payroll, and departed staff who never updated their profiles linger. A new hire appears only when they update their own profile, days or weeks after they start. Check: use the trend over months, never a single snapshot, and remember count accuracy typically runs 70 to 90% depending on company size.

Small-company noise. A ten-person startup may have a handful of associated members and a jumpy count. A 30% "jump" can be two people editing their profiles. Check: require absolute additions across multiple departments before you call it growth.

The ghost-job discount is misapplied by sector. Construction, arts, and legal run far above the 20% baseline. A flat discount over-scores them. Check: use the sector table.

A permapost is mistaken for a live role. High-turnover roles can be always open with a fresh-looking date. Check: repost frequency, not just the posting date.

Layoff-tracker silence is read as safety. Public trackers exclude rumors and unverified reports, so their figures undercount true layoffs. A clean check means only no verified public event. Check: cross-reference your state WARN portal by name.

A WARN absence is misread. WARN requires 60 days notice only from firms with 100-plus employees cutting 50-plus at one site, and exceptions exist for firms seeking funding, unforeseeable causes, or natural disasters. Government entities and cuts under 50 are exempt. Some states go further: New York requires 90 days and covers employers with at least 50 full-time staff. Check: WARN silence is not a clean bill of health for small cuts.

The freeze is invisible from outside. Leading signals like stalled intern conversion and response-rate collapse are internal. When freezes begin, postings often stay up while response rates fall. Check: probe liveness directly, step seven, before you trust a still-live posting.

A wall of open roles at a quiet firm is a contraction tell, not a growth one.

Reading the employer against its market

Score an employer against its own market, not against a universal bar, because the same absolute talent-acquisition headcount means different things in different countries. Benchmarking across borders will mislead you every time.

The gap is large. In Refolk's index the United States returns 112,357 recruiter and talent-acquisition profiles against 7,256 in the United Kingdom, roughly a 15.5x difference.

So a US employer with five in-house recruiters is running a leaner operation, relative to its market, than a UK employer with the same five. Read recruiter presence within-market, and read it against the target's own team rather than the agency-heavy supply that dominates recruiter counts overall.

Two named tools do most of the layoff work here, and both are outsider-friendly. A public tech-layoff tracker maintained by a single curator carries 4,400-plus layoff events across 2,900-plus companies since March 2020, each entry with a source link, and it deliberately excludes rumors, so its figures are a floor, not a ceiling. Your state WARN portal is the legal floor beneath that. Between them you settle the layoff dimension; the freeze dimension you can only infer.

Keeping the score current and acting on it

Re-run the score on your watch list, because hiring health is a moving target and the leading signals shift in weeks, not quarters. A watch that adds headcount across two departments and refreshes its postings has become a target; a target that goes quiet and loses its hiring manager has become a watch.

Three example searches keep the read live. "People who started at a company in the last three months" is a direct read on whether hiring is actually happening now, cutting through the lagging headcount count. "Engineering leaders who left in the last six months" surfaces the churn that precedes freezes. And "former employees laid off now open to work" corroborates a suspected layoff with named people when the trackers are silent. Refolk runs each of these as a plain-language search, which is faster than reconstructing them by hand across public profiles.

Before you commit an application to any employer, confirm the score holds.

The point of the whole exercise is discipline about where your hours go. A tailored application is expensive and a ghost-job cycle wastes about nine hours. Score first, sort second, and let only the employers that clear the bar cost you a real application. Once an employer is a confirmed target, Refolk tailors your resume to the posting, drafts the cover letter, and scores how well you actually fit, so the hours you saved on frozen employers go into the ones that can hire.