Bring rating spreads from every manager into one worksheet, resolve outliers against evidence, and lock a consistent rating curve before feedback goes out.
Quick answer
Inconsistent rating scales destroy trust in performance reviews first — collect ratings and evidence before the meeting, compare team-by-level distributions, flag outliers, and archive final decisions with reasons.
Performance cycles break when 'a 4' means something different on every team. This worksheet collects proposed ratings and their justification per manager before the calibration meeting, shows the distributions side by side, and turns the discussion into a documented, comparable final rating with an AI summary of each decision.
| Template | Performance Calibration |
|---|---|
| Category | HR & People Ops |
| Format | Scored assessment with personalized AI PDF report |
| Time to complete | About 10 minutes per response |
| What you get | A personalized AI report (printable PDF) for every response, plus the raw answers |
| Best for | HR business partners · People managers · Department heads |
| Cost | Free to copy and customize on FormLM |
Sample form and AI report preview (fillable once the app is published).
Everything you need to run this template.
Every manager submits ratings with evidence before the meeting.
Compare teams and levels to find inconsistent bars.
Focus the discussion where raters and evidence diverge.
Final ratings and rationale are recorded, not remembered.
From copy to deliverable in four steps.
Copy it and set the scale to your review cycle.
Ratings plus evidence before anyone discusses.
Work the flags: outliers, bars and borderline cases.
Record final ratings; the AI summary documents why.
Pick your role — see how it fits your workflow.
Keep rating bars comparable across teams.
Defend your ratings with written evidence.
See calibration drift before it becomes grievance.
What makes this template work as a professional deliverable.
A preview of real questions — the full version adds scoring and a personalized report per respondent.
