the gap where hiring breaks
every hiring process follows the same flow. a job gets posted. thousands of applications pour in. the ATS filters them by keywords. a small fraction pass through. the hiring team interviews them. somewhere between those two steps, everything falls apart.
where hiring breaks
| job posted | applications |
|---|---|
| 2,000 | |
| ATS keyword filter (95% threshold) | 1,900 rejected |
| 100 pass through | |
| human interviews begin | |
| most are the wrong people |
that gap between the ATS and the human interview is where hundreds of thousands of dollars in wasted interview spend accumulates every year. it is where the wrong candidates get through and the right ones get lost. there is no evaluation, no conversation, no intelligence in that gap. just a handoff from a keyword matcher to a human who has no signal.
the ATS does not evaluate people. it matches words.
every ATS works the same way. it takes the words on a resume and matches them to the words in a job description. it scores the match. it sets a threshold. everyone below that line is rejected. automatically. before a human ever sees them.
| statistic | value |
|---|---|
| applications per job posting at scale | 10,000+ |
| keyword match threshold | 95% |
| qualified candidates filtered out before any human review | 1 in 4 |
| average time to fill a role | 42 days |
this is not a failure of implementation. it is a failure of architecture. the ATS was designed to match vocabulary. it was never designed to understand people.
two failures that happen on every job posting
failure 1: false negatives
the best candidates are eliminated before any human review
a backend engineer with 10 years of distributed systems experience writes "system architecture" instead of "microservices." the ATS sees a keyword gap. it rejects the candidate. the strongest person in the pool is gone in milliseconds.
silently. at scale. on every single posting.
failure 2: false positives
the weakest candidates pass through and waste everyone's time
a candidate with 6 months of experience uses ChatGPT to mirror the job description word for word. every keyword matches. ATS scores it at 98%. the candidate passes. the recruiter spends hours on screens and interviews. the candidate can't answer basic questions.
interview time, money, and team energy. all wasted.
72% of hiring managers say they have made a bad hire due to insufficient signal. the ATS is not giving them signal. it is giving them keyword matches.
why this is getting worse, not better
four structural shifts have converged at the same time. none of them are reversing.
four shifts that broke the system
AI broke the resume
ChatGPT made keyword perfect resumes trivial to generate. the signal that ATS relied on, that a resume reflects real experience, is gone. any candidate can now game any ATS.volume broke the recruiter
remote hiring and one click apply tools permanently increased applications per role. recruiters did not get more capacity. they got more noise.burnout broke the team
turnover in recruiting roles is at historic levels. the humans responsible for finding talent are drowning in the noise that ATS was supposed to manage.the input layer is compromised
mass apply tools submit applications to hundreds of jobs automatically. the ratio of signal to noise in every applicant pool has collapsed. no amount of ATS tuning will restore it.
ATS was built for a world where resumes were human written and applications were intentional. that world is over.
what this costs
| cost type | amount |
|---|---|
| per wrong interview | $800 |
| per panel round | $3,000 |
| per bad hire | 3× annual salary. gone. |
| recruiter burnout | ∞ |
the hidden cost nobody tracks
the cost compounds at every step. a wrong shortlist means five, ten, fifteen expensive interviews with the wrong people. multiply that by every open role and the waste becomes structural. it is not a line item anyone budgets for, but it shows up in every quarter.
the 2024 Hacker News thread "who's getting their job applications rejected?" surfaced the candidate side pain vividly. hundreds of applications, final round rejections, systematic noise. the employer side has the same problem in reverse. both sides are broken by the same system.
what needs to change
the fix is not a better ATS. the fix is not a better keyword filter. the fix is filling the gap with something that can actually evaluate people, not match words.
evaluate behavior, not keywords
the signal that matters is how someone thinks, communicates, and solves problems. not whether they wrote 'microservices' on their resume.
compare candidates to each other, not to a rubric
hiring is comparative. the question is never 'is this person good?' it is 'is this person better than the others for this role?' the tools should reflect that.
screen at scale without losing signal
the gap exists because no tool could evaluate 200 candidates behaviorally in the time it takes an ATS to filter them by keywords. that is no longer true.
this is what we built aperture to do. aperture sits in the gap between the ATS and the human interview. every candidate takes a 15 minute adaptive behavioral interview. scored by lambda CORE, our Bayesian model, which ranks candidates against each other within the pool. the hiring team gets a ranked shortlist backed by data, with confidence intervals, within 48 hours.