A hiring scorecard is a one-page document defining the role's mission, three to five measurable first-year outcomes, and the rated competencies that predict them, written before sourcing and scored independently by every interviewer. Structured evaluation like this consistently outperforms unstructured gut-feel interviews, and it's the cheapest process upgrade in hiring.
The template
| Section | What goes in it | Example (AI engineer) |
|---|---|---|
| Mission | One sentence: why the role exists | Ship reliable LLM features that move activation |
| Outcomes (3-5) | Measurable first-year results with dates | Eval suite gating releases by end of Q1 |
| Competencies (5-7) | Behaviors that predict the outcomes | LLM systems depth; eval discipline; ownership; communication |
| Rating scale | Defined 1-4 anchors per competency | 4 = has done it repeatedly at this level, with evidence |
| Stage map | Which interview tests which competency | Take-home → systems depth; debrief → judgment; references → ownership |
| Decision rule | Agreed bar before interviews start | No hire below 3 on any core competency; disqualifiers absolute |
How to use it well
- 1Kickoff: hiring manager drafts, interview panel challenges, everyone signs off, before sourcing starts.
- 2Assign each competency to exactly one stage owner; unassigned competencies never get tested.
- 3Score independently within 24 hours of each interview, evidence quotes required, not adjectives.
- 4Debrief reads scores first, then discusses divergence; nobody updates a score because a louder colleague disagreed.
- 5Log the decision. Six months later, compare scorecard to reality, that's how the next scorecard gets better.
The mistakes that break it
- Writing it after meeting a candidate you like, that's rationalization with a template.
- Ten-plus competencies: everything measured means nothing weighted.
- Undefined scales: without anchors, a '3' means whatever each interviewer's mood says.
- Skipping it for senior or referred candidates, the hires with the highest blast radius get the least structure.
- Never revisiting accuracy: a scorecard that's never audited against outcomes stops being evidence and becomes ritual.
