An ongoing research project by Ryan Vet Back to ryanvet.com
← Back to the study

Methodology

The Hire

This instrument never asks anyone what they value. It shows two candidates and makes a person pick, eight times, with the variables rotated so that every attribute is tested against every other one. Everything below is the exact stimulus and the reasoning behind it, published in full so the study can be checked rather than taken on faith.

What it measures

Four things compete for a hiring decision: the degree, the years of experience, proof of finished work, and who vouched for the candidate. Each is tested against each of the others exactly once (six matchups), plus two rematches at a lower intensity, for eight rounds in total. Counting wins across the first six rounds produces a revealed ranking, which is then compared with the ranking the same person stated at the start. The distance between the two is the headline finding.

The four attributes

The school names

Naming no school would mean every respondent silently substitutes their own, and the stimulus becomes unmeasurable. Naming a single school would import that one school's reputation into the finding. Instead, one institution is drawn at random per respondent from a set of three per tier and held constant for their whole session, including any replay. US institutions only for this first wave.

Top tierHarvard, Duke, Northwestern, Elon University
Flagship tierOhio State, Arizona State, NC State

Every level, verbatim

Each card is assembled from these strings and no others.

E1Degree from {TOP_SCHOOL}
E2Degree from {FLAGSHIP_SCHOOL}
E3No degree, self-taught
E4Certificate from a six-month program
X1Nine years at one company, same role for the last five
X2Twelve years across two companies
X3Five years across two companies
X4Two years, plus a side business they still run
X5Eighteen months, first role in the field
X6Three years, first role in the field
V1Public portfolio with three finished projects
V2Nothing beyond the résumé
R1Referred by a former colleague of yours
R2Submitted by an outside recruiter
R3Applied through the job board

The eight rounds

In every round, the two attributes not under test are identical on both cards, word for word. Row order is fixed on every card: Education, Experience, Evidence, Referral. Card side is randomized per round; rounds one through six are shown in random order; the two rematches are always shown last, in order.

Round 1. Education vs Evidence

Candidate ACandidate B
Education Degree from {TOP_SCHOOL} No degree, self-taught
Experience Five years across two companies Five years across two companies
Evidence Nothing beyond the résumé Public portfolio with three finished projects
Referral Applied through the job board Applied through the job board

Round 2. Education vs Experience

Candidate ACandidate B
Education Degree from {TOP_SCHOOL} No degree, self-taught
Experience Eighteen months, first role in the field Nine years at one company, same role for the last five
Evidence Public portfolio with three finished projects Public portfolio with three finished projects
Referral Applied through the job board Applied through the job board

Round 3. Education vs Referral

Candidate ACandidate B
Education Degree from {TOP_SCHOOL} No degree, self-taught
Experience Five years across two companies Five years across two companies
Evidence Public portfolio with three finished projects Public portfolio with three finished projects
Referral Applied through the job board Referred by a former colleague of yours

Round 4. Experience vs Evidence

Candidate ACandidate B
Education Degree from {FLAGSHIP_SCHOOL} Degree from {FLAGSHIP_SCHOOL}
Experience Nine years at one company, same role for the last five Two years, plus a side business they still run
Evidence Nothing beyond the résumé Public portfolio with three finished projects
Referral Applied through the job board Applied through the job board

Round 5. Experience vs Referral

Candidate ACandidate B
Education Degree from {FLAGSHIP_SCHOOL} Degree from {FLAGSHIP_SCHOOL}
Experience Nine years at one company, same role for the last five Two years, plus a side business they still run
Evidence Nothing beyond the résumé Nothing beyond the résumé
Referral Applied through the job board Referred by a former colleague of yours

Round 6. Evidence vs Referral

Candidate ACandidate B
Education Degree from {FLAGSHIP_SCHOOL} Degree from {FLAGSHIP_SCHOOL}
Experience Five years across two companies Five years across two companies
Evidence Public portfolio with three finished projects Nothing beyond the résumé
Referral Applied through the job board Referred by a former colleague of yours

Round 7. Education vs Evidence, lower intensity

Candidate ACandidate B
Education Degree from {FLAGSHIP_SCHOOL} Certificate from a six-month program
Experience Five years across two companies Five years across two companies
Evidence Nothing beyond the résumé Public portfolio with three finished projects
Referral Applied through the job board Applied through the job board

Round 8. Experience vs Referral, lower intensity

Candidate ACandidate B
Education Degree from {FLAGSHIP_SCHOOL} Degree from {FLAGSHIP_SCHOOL}
Experience Twelve years across two companies Three years, first role in the field
Evidence Nothing beyond the résumé Nothing beyond the résumé
Referral Submitted by an outside recruiter Referred by a former colleague of yours

Scoring

Preference score: rounds one through six only, one point per win, zero to three per attribute. Ranking sorts by wins, and ties are reported as ties, never broken silently. The contradiction score sums the distance between each attribute's stated position and its revealed position. Response time is recorded per round as a quiet data signal and is never shown or scored.

How the sample is handled

Participation is self-selected and not a representative sample. Runs are flagged and excluded from analysis, though still shown a result, when a completion comes in under twenty-five seconds, when every pick lands on the same side, or when a respondent flips on both consistency checks. No finding is published until at least two hundred responses exist in the cell being reported. Replay runs are never mixed with first runs.

Published by Ryan Vet. Independent research, released in full for scrutiny.