Performance Calibration: A Practical Meeting Protocol

Short answer

Performance calibration is a meeting for comparing how managers applied shared criteria, not for ranking people against one another. Participants should review comparable roles, inspect the evidence and counterexamples behind proposed outcomes, record why a decision changes or stays, and route pay or promotion decisions through accountable human processes. A rating distribution may prompt a process question, but it should not dictate an individual's result.

Define the meeting's authority before inviting anyone

Calibration can mean a consistency check, a recommendation panel or a final approval step. State which one applies. Write down:

• which decisions are in scope;
• who owns the proposed and final outcome;
• which roles and periods are comparable;
• which criteria and rating definitions were communicated in advance;
• what evidence the panel may see;
• how conflicts of interest are handled;
• what the employee can review or challenge;
• whether pay, promotion or formal performance action follows a separate process;
• what record the meeting must produce.

Do not give a meeting broader authority because the software can display more data. Access should follow the decision and participants' roles.

Calibrate application of criteria, not people

The central question is: did managers apply the same stated standard consistently to sufficiently comparable work?

Performance calibration cannot make unlike work comparable by putting it on one scale. If two roles require different evidence, review how each role's standard was applied and avoid cross-role ranking.

The US Office of Personnel Management's guidance on performance review boards describes consistency, accuracy and fairness as goals while keeping formal roles and employee response rights visible. It applies to the US Senior Executive Service, not every employer, but the governance lesson is useful: define authority, evidence and conflicts before deliberation.

Prepare a case card for each decision reviewed

Use a consistent pre-read so meeting time is spent testing evidence rather than reconstructing it.

Remove unnecessary personal detail. Make reasonable adjustments to the process where required. The card is an aid to discussion, not a machine-readable case for an automatic rating.

Use this 60-minute meeting protocol

Adapt timing to the number and complexity of cases. A small, prepared sample is better than a rushed review of every employee.

1. Reconfirm scope and roles: 5 minutes

The chair states the decision stage, criteria, confidentiality rules, conflicts and who has final authority. Participants confirm that proposed outcomes remain provisional where policy says so.

2. Test one anchor case: 10 minutes

Choose a well-documented case for a single criterion. Ask what evidence meets, exceeds or falls short of the stated standard. Record the reasoning as a reference, without turning one employee into the benchmark for everyone else.

3. Compare selected cases: 25 minutes

For each case, follow the same question order:

1. Is the criterion relevant to this role and period?
2. What source supports the proposed conclusion?
3. Does the evidence cover enough of the period?
4. What counterexample or missing context could change it?
5. How was the same standard applied in the comparable case?
6. Is more evidence needed before deciding?
7. Who owns the final judgment?

4. Review decision changes: 10 minutes

For every retained or changed outcome, record the criterion, evidence, counterexample considered, rationale and owner. “Panel consensus” is not a rationale.

5. Separate process findings: 5 minutes

Move recurring issues into a process log: unclear standards, missing periods, inconsistent evidence requests, group differences or frequent late criterion changes. Do not repair a system-wide problem by changing one person's rating without evidence.

6. Confirm communication and follow-up: 5 minutes

Assign who will update the record, tell the employee, handle any response and revise the process before the next cycle.

Fictional example: same criterion, different evidence rules

This example is invented. Two support-team leads are reviewed on the same responsibility: ensure priority escalations have an owner, next action and review time by the end of the shift.

Manager A proposes “exceeds” for Devon, citing three high-profile escalations that senior leaders praised. Manager B proposes “meets” for Lee, citing the complete case log for the period. The panel does not compare Devon and Lee as people or decide that only one can exceed.

It first tests the criterion. The case cards show:

• Devon's three cited cases were complete and involved unusual complexity.
• Devon's full-period sample also contains two routine cases missing a review time.
• Lee's full-period sample contains no missing fields, but the manager provided no evidence about complex cases.
• Both leads handled a similar number of priority escalations, but complexity was not defined in the standard.
• Neither employee had been told that senior visibility or case complexity changed the performance level.

The panel cannot fairly apply an undeclared complexity rule after the period. It asks both managers to evaluate the communicated completeness standard over the same sampling window and records complexity as a criterion-design question for the next cycle. Each manager remains responsible for a supported outcome under the organisation's process. No distribution or head-to-head ranking determines the answer.

Do not force a distribution

A distribution can highlight a question: one manager may use a rating differently, one group may receive fewer positive outcomes, or the standards may be unclear. It cannot prove which individual decision is wrong.

The UK government's Senior Civil Service performance framework for 2026 to 2027 describes moderation as a consistency checkpoint. It explicitly says its expected distribution is not a requirement or target and should not be used to force managers to amend ratings. That framework is specific to the Senior Civil Service, but its distinction is broadly useful: inspect a distribution to audit the process, then return to role-relevant evidence for individual judgments.

Avoid these shortcuts:

• changing an outcome solely to fit a curve;
• requiring every team to produce a top or bottom category;
• ranking unlike roles on a single list;
• treating the loudest manager as the best-informed reviewer;
• using an AI recommendation as a deciding vote;
• hiding a decision change behind a recalculated score.

Add bias checks without diagnosing intent

At the end of the meeting, review process patterns separately:

• Which managers changed outcomes, and for what documented reasons?
• Did comparable cases use the same evidence window?
• Were counterexamples requested consistently?
• Did remote, part-time, disabled or less-visible employees have comparable evidence opportunities?
• Do outcomes differ across relevant groups, with denominators and role mix shown?
• Were standards or support changed during the period?
• Were employees able to correct factual errors?

A pattern can justify investigation. It does not by itself explain the cause. For a case-level method, use the performance review bias audit. The EEOC's performance-evaluation guidance emphasises communicated standards, consistent application and relevant factual detail in the US context.

Keep the decision log useful

Record one line per case reviewed:

Keep the original proposal and final rationale. Do not overwrite history in a way that makes the change untraceable. Follow retention, access and appeal rules for the organisation and jurisdiction.

Connect calibration to the rest of the cycle

Calibration is late if it is the first time managers discover that standards differ. Sample criteria at the start of the period, revisit changed responsibilities during check-ins and audit a few evidence records before final reviews.

Acas performance-management guidance treats formal reviews as part of a wider cycle that includes regular feedback, support and development. Use the objective performance review worksheet to improve criteria and the effective performance conversation guide to discuss evidence and disagreement with the employee.

Where Lontra can support preparation

Lontra can prepare a focused employee conversation and a manager brief when the review needs context about a work example, constraint or prior commitment. Managers do not receive raw employee responses. The product is not a ready-made 360-degree review or calibration system, does not rank employees and does not decide performance, pay or promotion.

The Lontra product overview explains the conversation and brief workflow. Use any relevant context as one source for human review, then keep criteria, ratings, calibration decisions and challenge routes in the accountable performance process.