How to run a hiring debrief, with an agenda

Why most debriefs go wrong

A debrief should combine independent evidence into one decision. Many do the opposite. The first confident voice sets the frame, everyone else adjusts, and the meeting ends with a decision nobody can trace to anything the candidate said.

• Anchoring. Whoever speaks first, usually the most senior person, sets the reference point for every opinion after.
• Conformity. Interviewers who scored a 4 quietly become 3s once the hiring manager says "not convinced".
• Adjectives. "Strong presence." "Not sure about the energy." Impressions are easy to say and impossible to check.
• No decider. When nobody owns the call, the safe move is another round.

That last habit is expensive. In Gem's 2025 benchmarks, interviews per hire rose from 14 to 20 between 2021 and 2024, and time to hire grew from 33 to 41 days. Extra rounds are what a panel buys when it does not trust its own evidence.

The fix starts before the meeting. The US Office of Personnel Management's structured interview guide has panelists rate each candidate independently, "without discussion with other panel members", and only then compare notes and "explore the basis for discrepancies". Independent scoring is part of what makes an interview structured, and structured interviews predict job performance at .42 against .19 for unstructured ones, in the 2022 re-analysis by Sackett and colleagues.

Before the debrief

• Every scorecard is in. No exceptions for seniority. If one is missing at the start time, move the meeting rather than hear that interview from memory.
• The facilitator builds a grid. Competencies down the side, interviewers across the top, scores in the cells. Mark every competency where two ratings differ by two points or more. Those marks are the agenda.
• Everyone reads the evidence first. Share all scorecards and the grid an hour ahead, once everyone has submitted. People should read what was said before they hear what anyone thinks.
• It happens within two working days of the final interview. The 24-hour feedback system is what makes that possible.

Three roles, named in advance

Keep the facilitator and the decider apart. A person who runs the meeting and holds the vote controls both the agenda and the outcome, and the panel will read the room accordingly.

The 30-minute agenda

When every score agrees, the same agenda runs in ten minutes: confirm the evidence, record the decision. A senior or contested role may need forty-five. Extend the walk-through, never the opening.

Five ground rules

1. Scores are submitted before the meeting. Nobody rates a candidate after hearing the room. If someone changes a score during the debrief, the note-taker records the original and the reason.
2. The most junior person speaks first. Seniority is a powerful anchor. Reverse the order and you hear what each person actually saw.
3. Quotes, not adjectives. "He said he would roll back first and investigate after" is evidence. "Solid under pressure" is a conclusion. The facilitator asks for the first whenever they hear the second.
4. Disagreements are settled against the rubric. Read the anchors aloud and ask which one the quoted answer matches. The question is never "who is right?" It is "what does the evidence show against this scale?"
5. The decider is named, and decides. Consensus is welcome; it is not required. One named person makes the call and records why.

How to handle a split panel

A split is information, not a failure. Work through it in this order:

1. Find the competency. Splits are rarely about the whole candidate. Usually one or two competencies carry the disagreement.
2. Check it is the same evidence. Two interviewers often asked different questions or heard different stories. Then you have two data points, not a disagreement.
3. Go to the anchors. Read the quote, read the scale, and ask each side which anchor fits. Many splits close here.
4. Name what would settle it. If the evidence is genuinely missing, write the one question that would resolve it. Ask it in a short follow-up call or a reference check, not a full new round.
5. Let the decider decide. If the split survives, the decider makes the call and writes down the risk they are accepting.

What to record

The record serves two readers: the candidate who asks why, and the team that reviews its own hiring later. Keep it in your ATS, not in a chat thread.

• The decision, the decider and the date.
• The two or three competencies the decision turned on, each with the quote behind it.
• Any dissent, in the dissenter's own words.
• For a hire, the risks accepted. For a rejection, the gap that decided it.
• Next steps: who tells the candidate, by when, and what they will be told.

Anti-patterns to stop

• The HiPPO. The highest-paid person's opinion arrives first and the rest of the meeting agrees with it. Speaking order and scores submitted in advance are the antidote.
• "Culture fit" as a veto. An undefined objection nobody can answer is a vibe with a veto attached. If the concern is real, name the behavior and the evidence; if not, set it aside. OPM lists rating candidates higher because they seem "similar to me" among the common rating errors.
• Re-litigating with new questions. "But can she handle enterprise customers?" may be a fair question. If it was not in the rubric, raising it now moves the bar for one candidate. Add it to the next rubric; judge this candidate on this one.
• Hearsay scores. An absent interviewer's view, relayed second-hand: "he said she was fine". Use the written scorecard, or nothing.
• The résumé re-read. The meeting drifts to the candidate's past employers instead of their interview evidence. The résumé earned the interview; the interview is what you are deciding on.
• Deciding by average. Adding scores up to a threshold skips the conversation about the competencies that matter most.

Debriefs with the evidence on the table

The hardest rule to keep is "quotes, not adjectives", because few interviewers write down exact words while they listen. A recording solves that, with the candidate's consent. Itya joins Google Meet, Microsoft Teams and Zoom as a notetaker, transcribes with speaker labels, and prepares a debrief that cites the transcript moment behind each claim, so anyone in the room can open the exact answer being discussed.

The decision stays where this agenda puts it. By default, Itya's AI recommends and a named person confirms; it never rejects a candidate automatically.