Hiring is a decision under uncertainty. Bayes' theorem is how serious people make those decisions.
Every hire is a prediction — whether someone you barely know will succeed in a role that doesn't yet exist, under conditions that will change. That is the textbook definition of a Bayesian problem: you start with the best estimate of the odds, then update those odds in proportion to the strength of each new piece of evidence.
The math is exactly what the CIA uses to evaluate intelligence, what the Navy used to find a lost submarine, and what every modern AI uses to update its predictions in real time. It works for hiring too — and the same math runs in both directions. The company is trying to predict whether the candidate will produce a Win-Win hire at the one-year anniversary. The candidate is trying to predict whether the job will turn out to be a Win-Win career move. Both sides are doing Bayesian reasoning, whether they know it or not.
The strength of evidence in Bayesian terms is captured by one number: the Diagnostic Ratio (DR). A DR of 1 means a tool tells you nothing. A DR above 5 is strong evidence. A keyword screen has a DR below 1, which means it makes your decisions worse than chance. A trusted referral has a DR of about 6. A well-prepared Performance-based Job Description combined with a Performance-based Interview has a DR of about 5. Hiring outcomes are driven by a small number of high-leverage decisions — tipping points — that each carry a measurable DR.
Starting odds · the sourcing channel
Where the candidate comes from sets the odds before any evaluation begins.
Read these three rows as starting positions. The amount of working evidence about the person — and reciprocally, the candidate's working evidence about the company — collapses by an order of magnitude at each degree of separation. Everything in §02 stacks on top of whichever row you start in.
| Starting position | DR likelihood ratio | Company odds Win-Win hire | Candidate odds Win-Win job |
|---|---|---|---|
|
1stKnown performer · direct evidence i
Someone on the hiring team has worked with this person, seen them perform in context, over time. The evidence is already in the room before the interview starts. This is the only sourcing channel where the starting odds are high enough that methodology improves an already-strong number rather than rescuing a bad one.
|
DR ≈ 8.5 | 40–50% | 40–50% |
|
2ndTrusted referral or warm introduction i
Someone with direct working evidence about the person is willing to vouch — and the hiring team weights their judgment. The Diagnostic Ratio of a trusted referral is about 6. The number collapses when the referrer is a casual connection rather than someone with real working evidence. Candidate odds run higher than company odds because the referrer has usually known the company longer than they've known the candidate.
|
DR ≈ 6 | 30% | 35–40% |
|
3rdStranger · responded to a posting i
No direct evidence on either side. A perfectly qualified candidate facing 200 applicants and an AI screen has roughly a 1–2% chance of being interviewed at all. The sourcing channel itself is a neutral baseline (DR ≈ 1) — the catastrophic odds come from this baseline combined with the absence of every other update. Most right candidates are invisible to the company. Most right jobs are invisible to the candidate.
|
DR ≈ 1 | 5–10% | < 1% |
The updates · six methodology tipping points
Each is a discrete intervention that updates the starting odds.
Shifts shown apply each tipping point to a 3rd-degree starting position — where most hiring happens and where the largest absolute gains exist. Applied to a 1st or 2nd-degree pool, the same interventions still move the numbers, but they improve an already-strong starting odds rather than rescuing catastrophic ones. Hover any row for the mechanism.
| Tipping point | DR likelihood ratio | Company odds shift | Candidate odds shift |
|---|---|---|---|
|
Define the work, not the person i
A Performance-based Job Description replaces the skills-and-credentials posting. The role is described as the outcomes someone has to produce in the first year — not the resume that's allowed to apply. The only update that raises the base rate itself. Every other update gets multiplied against this.
|
DR ≈ 5 | 8%→20% | <1%→5% |
|
Require accomplishments up front i
The two-step screen. Candidates submit examples of major accomplishments at application — not a resume. The screen evaluates trajectory and outcomes, not credentials. The single largest screening-stage intervention available — surfaces strong performers who fail credential screens but have the accomplishments to prove fit.
|
DR ≈ 5 | 8%→25% | <1%→10% |
|
Source against the Achiever Pattern i
Outreach targets candidates whose career history shows accelerating scope, stretch, and accomplishment — not candidates whose resume matches a keyword list. Changes who enters the funnel, not how they're evaluated. Achiever-pattern candidates select roles deliberately and are harder to mismatch.
|
DR ≈ 3 | 8%→25% | <1%→15% |
|
Control first impression bias i
The primary cause of most hiring mistakes — Type I and Type II both. The interviewer's snap judgment in the first four minutes turns the rest of the interview into confirmation theater. Bias control isn't a small additive lift — it's a threshold condition. Without it, the effective DR of every other evaluation tool collapses, because the interviewer has already decided and is now just gathering confirmation. The discipline that lets every other interview method actually function.
|
DR ≈ 3 | 8%→25% | <1%→10% |
|
Run a Performance-based Interview i
Structured fact-finding on past accomplishments and problem-solving instead of a conversation. The interviewer collects evidence; the decision is made afterward, against the PBJD, by the team. Structured behavioral interviews are one of the strongest predictors in the meta-analytic literature. The interview stops being where you decide and starts being where you collect what you'll decide on.
|
DR ≈ 5 | 8%→20% | <1%→8% |
|
Frame the offer as a career move i
The 30% Solution. The job is pitched as a meaningful non-monetary career step — stretch, growth, scope, satisfaction — not as a compensation transaction. The career-move conversation isn't just closing — it's the final evidence-gathering step. What the candidate values, how they think about scope and stretch, whether their stated motivations match their work history. Determines whether the right person stays long enough for the hire to register as Win-Win at the anniversary.
|
DR ≈ 3 | 8%→20% | <1%→15% |
|
All six, stacked · the full PbH method i
PBJD + Achiever-pattern sourcing + accomplishment screen + first-impression-bias control + Performance-based Interview + 30% career-move offer. Applied end to end. The PBJD and the PbI act as a paired update with a combined DR of about 5; the rest of the stack adds further independent lift, with discounting for overlap. Applied to a 3rd-degree pool, gets close to the ceiling of what selection methods can produce. Applied on top of a 1st or 2nd-degree source, both sides converge near the upper asymptote. Convergence between company and candidate odds is the signature of the methodology working.
|
DR ≈ 40+ | 8%→70–80% | <1%→55–65% |
The Bayesian read-out
Three things the table is saying — that the conventional hiring conversation hides.
01 · Sourcing dominates
The gap between a 1st-degree and 3rd-degree starting position is larger than the gap between doing nothing and doing everything. Most companies spend their methodology budget on the wrong end of the table — fixing screens and interviews while leaving sourcing untouched.
02 · The miss is bigger than the hit
Type II errors are strong candidates filtered out because their backgrounds don't fit the template. They are the majority of the qualified pool, not the exception. The two-step screen, Achiever-pattern sourcing, and networking into 1st and 2nd-degree connections each act directly on the Type II problem — three of the seven tipping points are about finding people the traditional process misses.
03 · Symmetry is the signature
The two columns converge only when both sides are evaluating the same evidence against the same defined outcome — the PBJD on the company side, the career-move conversation on the candidate side. That convergence is what Performance-based Hiring is. The company isn't trying to hire a person; both parties are trying to verify a match.
METHODOLOGY: The Diagnostic Ratio (DR) is the positive likelihood ratio of a hiring tool — how much more often it produces a true signal of strong performance than a false one. A DR of 1 means the tool tells you nothing; a DR above 5 is strong evidence. Full explanation here. Percentages are directional estimates calibrated against the Sackett et al. (2022) selection validity meta-analysis, the Harvard/Accenture Hidden Workers study (2021), the Adler Group practitioner dataset, and the Performance-based Hiring framework. Stacked DR figures reflect overlap discounting between updates; rank order and magnitude of shift carry the argument.