Most scorecards fail in the debrief, not in the interview. Everyone filled one in, the numbers disagree, and nobody can say why.
Anchor every level in an observable behaviour
A level called "strong" means nothing on its own. A level that says "described a trade-off they made and what it cost" can be checked against notes.
- Three to five criteria per interview, not twelve.
- One question per criterion, asked of every candidate.
- Notes captured before the score, never after.
The point is not to remove judgement. It is to make the judgement legible to the next person who reads it.
