Critical Path
Every figure used in the Critical Path job search talk, with its source, date, sample size and evidence tier. Also the widely repeated statistics that did not survive checking, and why they are not used.
Job search advice is unusually prone to numbers that nobody can trace. A figure gets quoted in a blog post, the blog post gets quoted by a career coach, and ten years later it is treated as established fact with no study behind it. This page exists so anyone can check the work. Each claim below is listed with where it came from, when it was published, how big the sample was, and how much weight it can carry.
300+ applications per open role
Applications per hire have tripled since 2021. Candidates are roughly 50% less likely to reach the interview stage than they were five years ago, though offer conversion rates have risen above 2021 levels.
115 → 244 applications per job
Applications per job roughly doubled between 2022 and 2025, a rise of about 111%. Over the same window applications per recruiter rose 411.8% and recruiting teams shrank 55.6%, to around five recruiters per company.
The 115 figure is reported slightly differently across secondary write-ups, some giving 116. The talk says about 115 for that reason.
5.6% unemployment, 42% underemployment
Unemployment among recent college graduates aged 22 to 27 held at 5.6% through the first and second quarters of 2026, with underemployment near 42%. Recent graduates now face higher unemployment than the workforce overall, which reverses a pattern that held from 1990 until about 2021.
3.1% hires rate
The national hires rate fell to 3.1% in February 2026, its lowest reading since April 2020, then moved between 3.2% and 3.5% through mid-year while layoffs stayed near 1.1%. This is the pattern usually described as a low hire, low fire market.
Referrals: 2% of applicants, 11% of hires
Referrals turn into hires at about 5 times their share of applicants (11% of hires from 2% of applicants). Company career pages produced 13% of applicants and 26% of hires, making a career page applicant about 3 times more likely to be hired than a job board applicant (26/13 = 2.0 vs. 42/61 = 0.69). Job boards produced 61% of applications and 42% of hires.
CareerPlug's client base skews toward small and mid-sized businesses, so the channel mix at a large enterprise will differ. The direction of the effect is corroborated by Gem's independent dataset of 165M applications, which found sourced candidates roughly 8 times more likely to be hired than inbound applicants.
Moderately weak ties drive the most job mobility
Randomized experiments on LinkedIn's People You May Know algorithm covering 20 million users over five years, spanning about 2 billion new connections and 600,000 job changes. The relationship between tie strength and job mobility is an inverted U. Moderately weak ties produced the most job moves and the strongest ties the least. Weak ties helped more in digital industries and strong ties more in less digital ones.
This is the strongest single piece of evidence behind the networking advice, because the experiment randomized who received which connection recommendations rather than observing what people already did. A published critique by Neal (2022) argues the effect may reflect network bridging rather than tie weakness itself.
Intensity does not predict job quality
Across 378 independent samples covering 165,933 people, job search intensity predicted the number of interviews (corrected r = .23), job offers (.14) and employment status (.19). Its relationship with the quality of the job obtained was .06, with a confidence interval spanning zero. Job search quality, by contrast, predicted employment quality at .19, and self-regulation predicted job search quality at .30.
This is the backbone of the quality over quantity argument. Note the effect sizes for intensity are small, which means job search outcomes are shaped heavily by factors outside a candidate's control, including the market and the field.
Job search interventions raise employment odds 2.67×
Across 47 experimentally or quasi-experimentally evaluated programs, participants were 2.67 times more likely to obtain employment than controls. The components that mattered most were teaching job search skills (odds ratio 3.32 with, against 1.62 without), improving self-presentation, building self-efficacy, encouraging proactivity, and goal setting.
9.25% at 11–20 applications, 2.58% at 100+
Interview rate per application, by how many applications a person sent in a quarter: 9.25% at 11 to 20 applications, 6.96% at 21 to 50, 4.65% at 51 to 100, and 2.58% at 100 or more. Tailored resumes converted at 4.23% against 2.07% for untailored, roughly double.
Outcomes here are self-reported by users of a job tracking tool, and Huntr's own disclaimer notes they are almost certainly under-reported. The volume relationship is also open to selection bias, since a person who sends few applications may be a stronger or more focused candidate to begin with. Treat this as directional support for the causal case made by the peer-reviewed work above, not as proof on its own.
85% of effort, 38% of interviews
85% of job seekers said they spend most of their search time on online applications, while 15% spend most of it networking. Among those who landed at least one interview, 38% got it through an employee referral, a recruiter reaching out, or networking.
18% to 22% of postings
An internal review by one hiring platform found that between 18% and 22% of quarterly postings were ghost listings, and that roughly 70% of its client companies had posted at least one in a single quarter.
About 1 in 7 listings
A scrape of 176,268 Indeed listings flagged roughly one in seven as potential ghost postings, using the test of a listing staying live for 30 days or more.
Exposure rises with volume
93% of surveyed job seekers said they had applied to a posting they believed was fake, or suspected they had. Reported exposure scaled with application count: among people who sent 1 to 10 applications, 23% said they had hit ghost postings many times, rising to 42% among those who sent 26 to 50.
LinkedIn's f_TPR parameter counts seconds
Filtering LinkedIn jobs to “Past 24 hours” writes f_TPR=r86400 into the URL. 86,400 is the number of seconds in 24 hours. The parameter accepts any value, so r3600 returns the past hour and r600 the past 10 minutes, tighter than any button the interface offers. Adding sortBy=DD sorts newest first rather than by relevance.
Below roughly 10 minutes the result set usually empties, because most roles are not posted that often. Sources disagree on which value of the work-arrangement parameter means remote, so the talk leaves that one out rather than risk it failing in a live demo.
1 in 5 US job seekers apply only to remote roles
More than one in five US job seekers applied exclusively to remote roles, while only 9% of newly listed US jobs were remote in late 2024. Among still-active jobs that continued attracting applicants, the remote share was 16%. Remote availability peaked at about 27% of applied-to jobs in early 2022 and fell to 16% by late 2024.
hiring.cafe: 3.6M+ jobs, 115,000+ companies
A job search engine that scrapes more than 30,000 company career pages directly rather than accepting employer submissions, using a language model to extract salary, seniority and remote status. Recruiting agencies are excluded by hand, and probable ghost postings are identified by detecting repeated reposts of the same listing. It refreshes roughly three times a day.
Job and company counts are the platform's own self-reported figures and are not independently audited.
Wellfound: 27,000+ startups, 100,000+ hires
Formerly AngelList Talent, spun out as an independent company in November 2022. It carries startup roles only, shows salary and equity ranges on listings, and bans staffing agencies, contingency recruiters and third-party headhunters under its recruiter code of conduct.
Self-reported platform figures. One third-party review puts the applicant-to-interview rate on the platform at 5% to 6%, which is in line with the wider market rather than better than it.
These circulate constantly in job search advice. Each one was checked and set aside. They are listed here because knowing which numbers are unreliable is as useful as knowing which are not.
“75% of resumes are rejected by ATS”
No primary source exists. The figure traces to a 2012 sales pitch by Preptel, a resume optimization company that shut down in 2013, and no methodology, survey or sample was ever published. Interviews with 25 US recruiters in late 2025 found that only 8% configure their applicant tracking system to auto-reject, while 92% rely on human review. The real bottleneck is recruiter overload from application volume.
“Recruiters spend 6 to 7.4 seconds on a resume”
This comes from a real eye-tracking study, but it is routinely over-extended. The 2018 study used about 30 recruiters, did not disclose how they were selected or what roles they were screening for, was produced by a job board, and was never peer reviewed. It also measured the initial sorting pass rather than full evaluation. The directional advice it supports, meaning clean layout and clear titles, is sound. The number itself is fragile.
“Apply within 96 hours for 8× the interview odds”
A single vendor blog analysis of about 1,600 applications from 2017, never peer reviewed and now nearly a decade old. The underlying logic that recruiters work the newest applications first remains plausible, and a 2025 recruiter survey found 52% say applying within 48 to 72 hours meaningfully improves visibility. The specific multiplier should be treated as folklore.
“47% of job listings are fake”
An outlier well above every other credible estimate. The defensible range sits between 18% and 22% from one platform's internal review, and about one in seven from an independent scrape of 176,268 listings. The problem is real and large enough without inflating it.
“X hours until a LinkedIn post hits 100 applicants”
No defensible published figure exists. LinkedIn capped its public display at “Over 100 applicants” at the end of 2023, so the counter stops climbing where the interesting range begins. LinkedIn also counts a click through to an external application as an application, which inflates the number against completed submissions. Anecdotes are plentiful and data is not.
The first is that vendor datasets, however large, describe the customers of one platform. Ashby, Greenhouse, CareerPlug, Gem and Huntr each see a slice of hiring shaped by who buys their software. Where those slices agree, as they do on referrals and on application volume, the finding is more trustworthy than any one of them alone.
The second is that correlation is doing a lot of work in the platform data. People who tailor resumes and send fewer applications may simply be better prepared candidates, and the tailoring may not be what causes the interview. The peer-reviewed experiments carry the causal argument. The platform numbers show the same pattern at a scale the experiments cannot reach, which is why both appear here.
Citations decay. Hiring data from 2024 and 2025 is already describing a different market than the one in front of you. This page is reviewed quarterly.