Engineering debriefs tend to end with everyone remembering a different conversation. The decision takes about a week to reach. By the end of that week, the reasoning behind it rarely survives intact. Whoever spoke last in the room often shaped the outcome more than the interview itself did.

Four interviews, four accounts

Brannek’s hiring loop was four interviews, each with a clear purpose. The purposes were clear to the people who designed them. They were rather less clear to the people running them six months later. New interviewers learned the loop mostly by watching whoever trained them.

What arrived at the debrief was four separate impressions of a candidate. Each was anchored in a different part of the conversation. None of them had been written down before the room started talking. A recruiter taking notes could only capture what got said out loud.

The loudest account in the room tended to win, whether or not it was the strongest evidence. A quieter interviewer with a sharper read often said nothing at all. The calibration problem got worse over time, not better, because new interviewers copied whatever habit was loudest in the room.

One scorecard per role

Brannek now defines the scorecard in Yardstick Scorecards before the role opens. Every interviewer fills in the same one. Recruiters and hiring managers arrive at the debrief looking at the same evidence, rather than reconstructing it from memory.

The scorecard breaks a role into a small set of named criteria. It covers the systems a candidate has actually run, how they reason through a design problem, and how they work with people outside their own team. Each criterion gets its own score and its own line of evidence, not a single overall impression.

Each interviewer submits a written score before the debrief starts, rather than filling it in live. That single ordering change removed most of the drift. Nobody is calibrating to the loudest person in the room, because nobody has heard them speak yet.

Disagreement is the useful part

Where two interviewers score the same criterion differently, the scorecard makes the disagreement visible and specific, rather than smoothing it over. That is a better meeting than one where everybody agrees for reasons they cannot quite name. The debrief now spends its time on the one criterion people actually disagree about. The scorecard does not choose who advances; a hiring manager still makes that call, and can overrule any score on the record.

Results at the debrief

Debrief meetings are now 35% shorter on average. The argument about criteria happens on paper beforehand, rather than out loud in the room while a decision is still being made.

100% of scorecards are submitted before the debrief starts. That was the whole point of moving the ordering, and the rule has held since the first full quarter of use. A late scorecard now delays the debrief, rather than getting filled in during it.

The loop itself has not shrunk. 4 interviewers still read every candidate against the shared rubric, and overrides remain rare: most debriefs now confirm the scorecard rather than argue with it.

What they are doing next

Brannek is drafting a shorter version of the scorecard for early-stage phone screens, where a full rubric currently takes longer to fill in than the call itself. A trial with two teams starts next month, and the recruiting team will decide afterwards whether to roll it out further.