Must hold
Every rating discussed is backed by evidence against the written level expectations, not by likeability or recent events. The meeting changes ratings only with a reason that is written down.
Inputs
- Rating scale and what each point means: [paste]
- Proposed ratings: [paste rows: employee ID, team, level, manager, proposed rating, two lines of evidence]
- Guidance on the expected spread, if any: [e.g. no forced curve, but review any team with all top ratings]
- Managers attending and the facilitator: [paste roles]
- Time available: [e.g. two hours]
Output format
- Pre-read: rating spread by team and by level, and the cases to discuss first (outliers, promotions, lowest ratings, ratings that changed sharply from last cycle).
- Timed agenda, with time per case.
- Discussion rules for the facilitator.
- Bias prompts to read out: recency, halo, similarity, leniency, central tendency, and visibility of remote or quiet employees.
- Decisions log: | Employee ID | Proposed | Final | Reason | Follow-up owner |
- After the meeting: what each manager tells their team members, and by when.
Rules
- Use employee IDs, not names, in the pre-read.
- Do not suggest a final rating for anyone; list the questions that would decide it.






