Your Panel Interview Isn’t Triangulating You. It’s Averaging Toward Whoever Talks First.

Here’s a number that should sit uncomfortably with you: 36.8%. That’s how often people in Solomon Asch’s classic experiments agreed with an obviously wrong group answer, just because everyone else in the room said it first.

A 2023 replication with 210 participants landed in the same neighborhood, around 33% on the standard task and 38% when the question involved opinion instead of a plain line on a card (you can read the full replication here). Sit with that last part. When the judgment gets subjective, conformity goes up, and candidate evaluation is about as subjective as judgment gets. So when you walk into a panel interview imagining four independent minds cross-checking each other, understand what’s usually happening instead: the room is quietly averaging toward whoever spoke first.

The 37 percent nobody warns candidates about

Asch’s original studies used all-male American college students in the 1950s, so the exact figure deserves an asterisk. But the effect has been replicated for decades, and the direction never changes: put a person in a group and their stated judgment drifts toward the group’s.

The uncomfortable upgrade from the 2023 work is that conformity was higher (38%) on opinion items than on the perceptual line task. A hiring panel isn’t measuring a line on a card. It’s arguing about whether you’re ‘a culture fit’ or ‘ready for the scope,’ which is exactly the kind of soft, contestable call where social pressure does its best work.

• Perceptual task: about 33% conformity in the modern replication.
• Opinion task: about 38% conformity, higher when the answer is debatable.
• The takeaway: the fuzzier the judgment, the more people fold toward the group.

Panels cascade, they don’t cross-check

The comforting story is that four interviewers each form an honest view, then compare notes and converge on truth. The messier reality is a sequence. Someone speaks first, usually with confidence, often with rank, and the rest of the room calibrates against that read instead of against the candidate.

This isn’t a fringe complaint from academics. Truffle, a hiring-tech firm with a commercial interest in panels working, writes plainly that ‘the actual effect in most rooms is consensus drift toward the most senior interviewer’s read.’ They go further, warning that panels ‘redistribute the bias from the individual to the group, and the group’s bias is harder to detect because everyone agrees.’ You can read their full breakdown here.

• The first strong read anchors the room. Later evaluators adjust toward it, not away from it.
• Rank amplifies the anchor. When the hiring manager outranks the panel and speaks early, others tend to nudge their stated scores toward that view.
• Agreement feels like validation. A unanimous panel looks rigorous, even when the unanimity is just an echo.

The validity number panel defenders skip

Here’s where the data gets sharp. Conway, Jako and Goodman’s meta-analysis of 111 interrater reliability coefficients found that unstructured interviews, the format most real panels actually use, have an upper validity ceiling around r=.34. Highly structured interviews reach roughly r=.67. You can see the meta-analysis here.

Translate that: the typical unstructured panel operates at less than half the predictive power its defenders cite when they point to the impressive structured-panel research. They’re borrowing credibility from a format they aren’t running.

• Unstructured panel: ceiling around r=.34.
• Highly structured interview: around r=.67.
• The gap: most ‘rigorous’ panels are running the weaker version and calling it the strong one.

The debrief is where it collapses

The same meta-analysis found something even blunter about the group discussion afterward. Combining panel ratings subjectively, meaning the debrief conversation, showed ‘no evidence of usefulness.’ The reliability gains come from mechanical combination: independent scores, averaged, without the deliberation.

Read that twice. The part of the process that feels most valuable (the group talking it through) is the part carrying the least statistical weight. And it’s the exact stage where anchoring takes over. Research on anchoring consistently shows final group estimates stay within 11 to 20% of the original number even when that number is obviously wrong. Whoever floats the first ‘I’d say a 6’ has already set the range everyone else negotiates inside.

• Subjective combination: no measurable benefit to reliability.
• Mechanical combination: independent, averaged scores are what actually helps.
• Anchoring drag: group estimates typically settle within 11 to 20% of the first number on the table.

Interview Guys Take: The cruel irony for you as a candidate is that the strongest interview of the day can get quietly overwritten in a debrief you never see. If the person who liked you scored you independently but the senior voice opens the room with a lukewarm take, the anchoring math starts working against a number you earned fairly. The evidence didn’t change. The order of speaking did.

When panels actually earn their reputation

None of this means panels are useless. It means the good numbers belong to a version most companies don’t run. Wingate and colleagues’ 2025 meta-analysis, covering 37 studies and 30,646 participants, found highly structured panels reaching .78 interrater reliability and r=.57 predictive validity, figures that rival cognitive ability tests. You can see that summary here.

SHRM-cited research points the same way, reporting multi-evaluator formats producing meaningfully higher predictive validity than single-evaluator ones, but only when the observations are genuinely independent. The format isn’t the problem. The near-universal failure to enforce the structure is.

• Independent pre-scoring: each interviewer commits a score before anyone talks.
• Simultaneous reveal: everyone shows their read at once, so nobody anchors.
• Structured questions: same core questions, scored against the same rubric, for every candidate.

One dissenter breaks the spell

The most useful finding in the whole Asch line is also the most hopeful. When even a single confederate broke unanimity and gave the correct answer, conformity dropped sharply. The pressure isn’t magic. It’s fragile, and it depends on everyone appearing to agree.

In a hiring room, that means one panelist willing to disagree, or a formal ‘devil’s advocate’ role, can disrupt the whole senior-anchor cascade. The problem here is organizational, not biological. Companies that rotate who speaks first, or assign an explicit dissent role, neutralize the exact dynamic this article is about. It’s worth being honest, too, that the classic Asch framing gets overstated: a large share of subjects stayed independent, and conformity varies across cultures and eras.

Interview Guys Take: The companies interviewing you fall into two camps, and you can often tell which one from the process itself. One camp has read this research and built guardrails: separate scorecards, no debrief crosstalk before scores lock, junior voices first. The other camp calls four people in a room a panel and lets rank do the deciding. The first camp is rarer than it should be, and it’s usually the better place to work for reasons far bigger than their interview design.

What this actually changes for you

You can’t fix a company’s debrief process from the candidate’s chair. But knowing the room averages toward the first strong read changes where you spend your energy.

First contact carries outsized weight, so treat early touchpoints as anchor-setting, not warmups. That’s true from the first phone screen onward, because the read someone forms early tends to get repeated later. When the panel opens with ‘tell me about yourself,’ you’re not making small talk, you’re planting the frame the rest of the room will calibrate against, which is exactly why that question is trickier than it looks.

• Identify the senior voice fast. In an unstructured panel, that person’s read is likely to become the room’s read, so make sure your strongest evidence reaches them clearly.
• Make your answers portable. Structure stories with the SOAR method (Situation, Obstacle, Action, Result) so a supportive panelist can repeat your win accurately in a debrief you’ll never attend. That discipline matters most in senior and leadership rounds.
• Give every seat a reason to advocate. In cross-functional panels, like a typical account manager loop, tie your value to each person’s world so no single voice owns the verdict alone.

The pitch for panel interviews is triangulation: more eyes, more angles, less bias. The evidence says that promise only holds when the structure is enforced, and the structure almost never is. Left to run on default social wiring, a panel doesn’t triangulate you. It anchors on a first read, drifts toward rank, and calls the resulting agreement ‘consensus.’

So when you hear ‘don’t worry, it’s a panel, so it’s fair,’ hold that claim loosely. Fairness lives in the process, in the pre-scoring and the reveal order and the willingness to let one person disagree out loud. When those guardrails are missing, the room isn’t checking four opinions against the truth. It’s rounding four opinions toward whoever spoke first.