Performance Calibration

How a calibration meeting works

A typical cycle runs like this:

1. Managers draft ratings for their direct reports, with evidence: goals met, metrics, examples of behavior.
2. HR prepares the data: the rating distribution by manager and team, prior-year ratings, and sometimes a nine-box grid view of performance and potential.
3. The calibration meeting: a group of peer managers, led by their common leader and facilitated by an HR business partner, reviews each rating. Managers explain outliers, especially top and bottom ratings, and peers challenge with evidence.
4. Adjustments are agreed and the rationale recorded.
5. Final ratings feed into the performance review conversation, merit increases and talent decisions.

Meetings usually cover 30 to 80 employees in two to three hours. Larger groups cascade: team leads calibrate first, then managers, then directors.

Worked example

A contact center has six team leads, each with about 15 agents. Before calibration, Team Lead A has rated 9 of 15 agents "exceeds expectations"; Team Lead D has rated just 1 of 14. Yet the teams' quality scores and first-contact resolution are within 2 points of each other.

In the meeting, the operations manager asks each lead to show the evidence behind top ratings. A's "exceeds" agents include several who simply met targets; D held back high performers because "nobody's perfect." After discussion against the written rating definitions, A's team ends with 4 "exceeds" and D's with 3. Merit budget now follows performance instead of the manager's grading habits.

Calibration vs forced ranking

Forced ranking (or forced distribution) requires a fixed share of employees in each rating band, for example 20% top, 70% middle, 10% bottom, the "vitality curve" model made famous at General Electric. Many large companies have abandoned or softened it because it pits colleagues against each other, discourages collaboration and forces managers to label someone a low performer even in a strong team.

Calibration aims at a consistent standard, not a fixed shape. A guideline distribution can be a useful check ("we'd expect roughly 15% to 25% in the top rating"), but the goal is that ratings reflect evidence. If a team is genuinely strong, its ratings can be higher.

Biases calibration is meant to catch

• Leniency and severity: managers who rate everyone high or low.
• [Central tendency bias](/glossary/central-tendency-bias/): rating nearly everyone in the middle to avoid hard conversations.
• Recency: overweighting the last month.
• Halo and horns: one trait coloring the whole rating.
• Similarity bias: higher ratings for people like the manager.

Calibration can also introduce bias. The most vocal manager in the room may win more upgrades for their team, and employees whose work is visible to senior leaders get discussed more favorably. A good facilitator watches for this and keeps the conversation on evidence.

Making calibration fair and defensible

• Written rating definitions with behavioral examples for each level.
• Evidence required for every rating at the top and bottom.
• A trained facilitator, usually HR, who tracks time and challenges vague claims ("she's a rock star").
• Demographic review: after calibration, HR checks rating distributions by gender, race/ethnicity, age band, disability status where known, and part-time vs full-time. Differences need an evidence-based explanation, because ratings that drive pay and layoffs can become evidence in discrimination claims.
• Notes on changes, with the reason.
• Confidentiality: individual discussions stay in the room.

For frontline roles with good metrics, calibration can be lighter and faster. For roles judged mostly on behavior and judgment, invest more time.

Common mistakes

• Calibrating without data, so the meeting becomes a debate of opinions.
• Changing ratings after managers have already told employees.
• Using calibration to hit a budget rather than to apply a standard.
• Never telling managers how their ratings compared, so the same skew repeats next year.