It is finding people who look like great people.
Every company using an ATS shares the same assumption: among the hundreds of people who apply, some great candidates are hiding. The filter's job is to find them.
"The Tao that can be screened is not the eternal Tao. A great candidate who matches your keywords is not necessarily a great hire. A great candidate who does not match your keywords is not necessarily a poor one."
— paraphrasing Lao Tzu, Chapter 1 · adapted for the hiring contextThis is the fundamental error of traditional resume screening. It confuses the symbol — a job title, a credential, a keyword — with the thing itself: the ability to actually deliver results in the role.
The consequences are two classes of error that compound each other, quietly, invisibly, at industrial scale.
A candidate who matches the filter profile — years of experience, degree, title — but cannot actually perform at the level required. The hire looks right on paper. The result is not right in the role.
A candidate who could deliver exceptional results — but whose resume doesn't match the filter. They are screened out before a single conversation. Their performance record never enters your awareness.
Type 1 errors are painful and visible — bad hires get noticed. Type 2 errors are invisible and compounding. No one knows who was screened out. No one is blamed. The cost accumulates silently as a workforce full of adequate hires instead of great ones.
Bayesian inference is the mathematics of updating beliefs as evidence arrives. Applied to hiring, it asks a precise question: how much does passing your screen actually increase your confidence that a candidate is strong?
Before seeing any resume, what fraction of applicants from this channel are genuinely strong? For a typical job posting this is roughly 15–20%. This is your starting belief — your prior probability.
For any screening signal, you need two numbers: how often do strong candidates show this signal, and how often do weak candidates show it too? The ratio of these two numbers is the signal's diagnostic power. A ratio near 1.0 means the signal tells you almost nothing.
Combine the prior with the diagnostic ratio to get your posterior — your updated confidence that a candidate who passed the screen is actually strong. If your screen has a diagnostic ratio near 1.0, your posterior barely differs from your prior. You did screening work and learned almost nothing.
Each independent signal becomes the prior for the next. Signals with genuine diagnostic power compound dramatically — moving from 20% confidence to 90%+ across three well-chosen signals. This is why the choice of screening criteria is not administrative. It is the most consequential decision in your entire hiring process.
The critical insight: standard resume keywords have a diagnostic ratio of approximately 1.1x. Job title match, years of experience, and degree all appear at nearly the same rate in strong and weak candidates' resumes. Running these filters feels like rigor. The math reveals it as noise.
"Performing an action that appears decisive but produces no information is not wisdom. It is the illusion of wisdom. The Tao of Hiring requires that we measure what actually predicts — not what merely looks predictive."
Starting from a 20% prior — the typical base rate for a job board application pool — here is what each type of resume signal does to your belief that a candidate is genuinely strong.
| Signal | P(signal|strong) | P(signal|weak) | Diagnostic ratio | Posterior belief |
|---|---|---|---|---|
|
Generic keywords
Title match, years of experience, degree field
|
55% | 50% | 1.1x | |
|
Buzzwords only
"Results-driven", "strategic thinker", "team player"
|
35% | 65% | 0.5x | |
|
Achiever language
"Grew revenue 40%", "reduced churn by half", quantified results
|
78% | 14% | 5.6x | |
|
Rate of growth
Upward title progression, expanding scope, increasing team size
|
72% | 20% | 3.6x | |
|
2–3 critical keywords
Role-specific, KPO-linked — not generic, tied to actual work
|
80% | 32% | 2.5x |
Posteriors shown for single signals at 20% prior. Combined achiever screening (all three) → 91%.
The divergence is stark. Buzzwords actively reduce your confidence — a candidate whose resume is dense with generic language is less likely to be strong than the base rate suggests. Generic keywords barely move the needle. Meanwhile, quantified achievements, demonstrated growth, and role-specific performance signals compound to give you genuine predictive confidence before a single conversation.
Set your applicant pool quality and see exactly how each screening approach affects your Type 2 error rate and interview precision.
The Tao of Hiring is not about adding more filters or more technology. It is about choosing signals with genuine diagnostic power — signals that appear far more often in strong candidates than in weak ones, and compounding them sequentially.
Three signals, applied in order, can move your confidence from a 20% prior to a 91% posterior before a single interview is scheduled. The key is understanding why each one works.
Quantified accomplishments — "grew revenue by 40%," "reduced churn from 18% to 9%" — are highly diagnostic because weak candidates rarely write this way. It requires having actually done something measurable, and the willingness to be held to a standard.
Upward trajectory — expanding titles, growing team sizes, increasing scope of responsibility — signals that previous organizations bet on this person repeatedly. Organizations are imperfect judges, but the pattern of repeated advancement is genuinely predictive.
Not generic buzzwords — but 2 or 3 terms directly linked to the role's key performance objectives. A candidate who mentions the specific technical approach, methodology, or domain challenge described in a performance-based job description has read it and recognized their own experience in it.
"The master recruiter does not seek the candidate who matches the description. They seek the candidate whose track record describes the outcomes. This is the whole of it."
— Lou Adler · Performance-based HiringThe Performance-based Job Description (PBJD) is the instrument that makes this possible at scale. By describing the work — the 4–6 key performance objectives a person must achieve in the first year — rather than the credentials, it attracts candidates who self-select on capability rather than profile match. The filter becomes the job itself, not a proxy for it.
Whether you want a diagnostic of your current system or direct access to the tools that make this approach practical, both paths start here.
A scored diagnostic of your hiring system — 0 to 100 — across six dimensions. Identifies exactly which stages are generating your Type 2 errors. Free. Delivered in 48 hours.
Tools to write PBJDs, convert them to ads, score candidates on performance criteria, and run structured interviews — all built on the methodology described here.
"A journey of a thousand hires begins with a single performance objective."