The Recruiter Chairs the Debrief and Guards the Evidence

Who should chair, and who should decide?

The recruiter chairs and the hiring manager decides. That split works because the two jobs need opposite things. Chairing needs the widest view of the evidence and no stake in the outcome. Deciding needs accountability for the result and the standing to be wrong about it in public. The recruiter has seen the screen, the async round, the exercise and every card. The manager will live with the hire.

What the chair actually holds is small and specific:

• The order. Cards read in submission order, most junior voice first, discussion only where two cards conflict.
• The sourcing rule. One question, asked over and over: where did you see that?
• The provenance statement. A sentence at the start naming every stage where nobody observed anything directly.
• The record. What was decided, on which claims, and what nobody could resolve.

What the chair does not hold is a view. A chair who wants an outcome runs the meeting differently, because every procedural choice becomes a move toward it, and the room reads that faster than anyone expects.

Why shouldn't the chair vote?

Because a recruiter's metrics are graded on whether the role closes, which is a real conflict with a no-hire call, and everyone in the room knows it. A second reason survives argument better: facilitation and advocacy cannot be held by one person, and the moment the chair has a preference, the order, the framing and the summary all quietly serve it.

Seniority is not the qualification it gets treated as either. Reviewing the evidence on selection judgment, Highhouse writes that although it is commonly accepted that some employment interviewers are better than others, research on variance in interviewer validity suggests the differences are due entirely to sampling error, and reproduces a 1943 admissions result in which high school rank plus an aptitude test correlated .45 with academic achievement while the same two predictors plus counselors' intuitive judgment correlated .35 ¹. That is an admissions study of 1939 entrants rather than a hiring study, and the sampling-error line summarises a paper reported at second hand.

It is enough to retire one habit. The practice of letting the most experienced person's read outrank the written evidence is exactly what an unearned vote becomes in month three, and the 1943 result is a reminder of the direction that goes: a judgment layered on top of two solid predictors made the prediction worse rather than better. None of which says interviewers add nothing. They are good at gathering information and poor at combining it, and the chair's whole design follows from taking that seriously.

Run the debrief in this order

Cards in before the meeting, read in submission order with the most junior voice first, discussion only where two cards disagree about the same claim, then the decision. The chair offers no view at any point. What the chair does is refuse to let an unsourced claim into the discussion. That sounds severe, and it stops being severe quickly: a team that has watched a few claims dropped starts sourcing them on the way in.

A thirty-minute agenda for a four-person loop:

1. Two minutes. The chair names the stages, and says which produced a human observation and which did not. 2. Eight minutes. Cards read aloud in the order they were filed. No commentary yet. 3. Fifteen minutes. Only the claims where cards conflict. Each speaker names the round and the moment. 4. Five minutes. The manager decides, or names the single observation still missing and what round would get it.

Cards have to exist before any of this works, and filing them in the right order is a separate discipline covered in submitting scorecards before the debrief and the transcript. What happens when step three does not converge belongs to the rule when the panel disagrees and whether anyone holds a veto.

What does the chair owe the evidence?

One sentence at the start of every debrief naming the stages where nobody watched anything. If a screening tool produced a value, an async round was assessed from a recording, or a summary stood in for a round, the chair says so before the discussion opens and says what that stage can and cannot support. Nothing else in the meeting supplies that, and nobody else in the room has seen all of it.

Adding people does not do this job. Across five studies of demographic similarity between interviewer and candidate, the standard review of structured interviewing reports small or no effects in one-on-one interviews and in two-person, three-person and four-person structured panels alike, and it reports no panel-against-individual validity comparison at all ². Most of those were field studies with large applicant samples and rigorously designed structured interviews, which is the condition under which such an effect would be hardest to find, so the honest reading is that structure appears to be doing the work and headcount neither added nor removed a similarity effect on top of it. A panel covers more claims. It does not cancel a bad one.

The automated stages need naming most, because the literature has not caught up with them. The authors of the 2022 revision of the standard validity estimates, whose own table ranks structured interviews first, note that as administration and scoring move toward automation the interview becomes more amenable to running at volume, and that validity needs to be evaluated under those changed circumstances ³. They are flagging an open question rather than settling it. The chair's opening sentence is what carries that into a room which would otherwise assume the automated stage watched something.

On Monday, name a chair for the next debrief who did not interview the candidate, and hand them one line of authority: nothing counts until somebody says where they saw it. On a team too small to find that person, the fallback is the smallest hiring process a two-person company can defend. One chair, one order, one question asked until it stops being needed.

See how it works >