1. LandOffer
  2. Blog
  3. All articles
  4. Prepare for Performance Calibration with Evidence Across the Review Period

People Management

Prepare for Performance Calibration with Evidence Across the Review Period

Three completed fictional employee evidence cases and a calibration decision log show whole-period support, shared rating boundaries, reasons and unresolved approval.

10 min readLandOffer team

Prepare for performance calibration by matching each proposed rating to the applicable role expectations and dated evidence from across the review period. Bring the cases where reviewers might reasonably disagree. State which missing fact or different interpretation would change your recommendation. Calibration should make the interpretation inspectable; an orderly-looking distribution does not establish that individual ratings are justified.

This LandOffer.ai guide uses a completed fictional three-employee packet and decision record. The public university guidance was checked October 25, 2026. All employee labels, work records, standards, dates and decisions in the examples are invented. No actual personnel records were accessed, and the examples do not establish your employer's rating rules or approval authority.

Four quarterly evidence rolls share one measuring rule, illustrating review-period coverage and a common standard rather than employee ranking.

Establish the process and the standard before comparing cases

Confirm the appraisal period, applicable job expectations, rating definitions, required inputs, meeting participants and final approval route. Ask whether the meeting can decide a rating, recommend a revision or request additional evidence. A facilitator may manage the discussion without having final approval authority. Keep draft recommendations labeled until the required review is complete.

CU Boulder's published two-step calibration guidance describes supervisor evaluations, reviewer examination and a wider discussion before final evaluations are presented to employees. It also calls for confidentiality, appropriate participants and checks on decision makers. This is one university's documented process, not a sequence every employer uses. Its purpose does not make calibration immune to bias.

Write the applicable rating boundary in concrete terms. For a given role, what distinguishes successful completion of assigned work from the next rating? Which outcomes and behaviors matter, and which goals changed during the cycle? Retain the version and effective date of revised expectations. Do not invent an additional criterion in the meeting because one employee's story is more memorable.

UC Davis's Calibration 101 links proposed ratings with observed results and behaviors and the appraisal period. It separately identifies merit-dollar decisions. Use that distinction to ask what your own meeting is deciding; do not present a calibration recommendation as a promised raise, promotion or award.

Build coverage across the period without making every month look equal

Create a short evidence map before writing the rating argument. Record when work occurred, the expectation it addressed, the employee's contribution, an identifiable source and the relevant limitation. A release record can establish delivery. A stakeholder note may establish a recipient's observation. Neither automatically proves every claim about the employee's impact.

Florida Atlantic University's appraisal guidance recommends documenting examples throughout the period and warns against recent events or isolated incidents dominating the assessment. It also favors job-centered explanations over personality comments. The useful preparation question is whether you can explain the pattern across the period, rather than recall only the latest launch.

Whole-period coverage does not require equal output in every month. Work assignments, approved changes and opportunities differ. Show when a project was canceled, a dependency was unavailable or a responsibility moved. Separate that context from evidence of how the employee responded. A missing record is an evidence gap, not proof that no work happened.

Ask relevant observers for the specific work and period they can describe. Avoid collecting general endorsements such as “everyone trusts them.” Capture observed actions, dates and outcomes instead. Keep review material in the authorized personnel workspace with access limited to the process. Removing names from a real packet does not necessarily anonymize an identifiable project or employee.

Completed packet: three employees and four evidence windows

The following invented packet covers July 1 through June 30, a closed review period. Manager Mina prepares cases for three software engineers at the same fictional level. Director Sora is the final rating approver under this example's stipulated process. Peer manager Tariq challenges the application of standards, while People partner Ruth facilitates and advises on process. Ruth does not select ratings by vote.

The fictional standard says Meets means delivering agreed outcomes with appropriate quality, collaboration and ownership. Above additionally requires a sustained pattern of contribution beyond the assigned scope with demonstrated benefit beyond the immediate work. A single visible success is insufficient under this invented standard. There is no required proportion in either category. These simplified definitions are teaching inputs, not a recommended replacement for your employer's framework.

Employee July–September October–December January–March April–June
A Assigned search-filter change completed; test review records a corrected edge case Second change missed its agreed handoff date; dependency notice came after the checkpoint Revised delivery checkpoints used on the next assigned change; receiver confirms timely handoff High-visibility rollout completed; Mina proposed Above chiefly on this event
B Maintained assigned release checks; receiver notes identify a recurring handover ambiguity Proposed and documented a reusable review method beyond the assigned maintenance work Two receiving teams adopted the method; dated owners confirm their use Third team adopted it; B corrected an unclear step from recipient feedback; Mina initially proposed Meets
C Delivered the first two items under a four-item improvement goal Completed the third item; fourth remained open with an upstream dependency A February planning note appears to remove the fourth item, but the saved copy lacks approval Completed the replacement work; Mina proposed Meets using the apparent scope change

The map makes periods visible, but it is not the entire packet. Each employee needs a concise case with sources and contrary evidence.

A's case. Release entries A1 and A4 support the completed changes in the first and last windows. Handoff note A2 records the missed checkpoint and late dependency notice. Receiver note A3 supports subsequent improvement. Mina's assessment recognizes the successful rollout and stronger follow-through. Her unresolved question is whether A demonstrated the sustained broader contribution required for Above; the current packet does not identify that pattern. A is not described as unreliable because of one miss or exceptional because of one launch.

B's case. Maintenance log B1 supports the assigned work. Proposal B2 identifies B's own contribution to the reusable method, distinct from teammates' implementation work. Recipient records B3 and B4 name three adopting teams and dates. They support adoption and a corrected instruction, not a measured productivity increase. B's low presentation visibility is irrelevant to whether the documented work meets the stated boundary. Mina flags that her initial narrative credited routine maintenance while omitting the later broader contribution.

C's case. Work entries C1 and C2 support three completed original items. Planning note C3 records the proposed February change; replacement acceptance C4 establishes delivery of the new work. Neither C3 nor C4 proves that an authorized person retired the original fourth item. The packet labels that interpretation unresolved. C's effort on the blocked item is context; the unknown decision about the goal remains a separate question.

These cases are comparable because they use the same role boundary and evidence structure. They do not need identical projects, equal presentation time or the same number of artifacts. A consistent packet lets reviewers check the same questions in every case; counting artifacts cannot replace checking their relevance and support.

Test boundary cases without turning the meeting into a ranking exercise

Present the criterion first, then the evidence and your uncertainty. For A, ask: “Which record establishes sustained contribution beyond assigned scope, rather than excellent delivery of the rollout?” For B, ask: “Are we crediting the documented adoption and ownership, or requiring a presentation that the standard never mentioned?” For C, ask who approved the scope change and when.

Check whether the same reasoning works in the opposite direction. If B's quieter work supports Above, A's more visible project should still be assessed on its own evidence. B does not need A's rating to fall. If C's missing approval needs verification, that uncertainty should not automatically become a poor-performance finding.

UNC Charlotte's published calibration guide explicitly describes its university system as neither a quota nor forced distribution. The linked filename carries “rev11.2017,” so treat it as a published institutional example rather than proof of another employer’s current policy. Its distinction illustrates why checking consistency and allocating a predetermined number of ratings are different operations. It does not establish that every employer rejects distributions; confirm the applicable policy instead of assuming it. If a distribution constraint is part of your process, ask the authorized owner how it applies and record it honestly rather than inventing an evidence-based explanation for a numerical adjustment.

Replace character labels with job evidence. “Not leadership material” leaves the standard and behavior unspecified. “The dependency was raised after the agreed checkpoint” identifies something reviewers can examine. Similarly, confidence in a meeting, office visibility or similarity to a manager is not a substitute for the work requirements being assessed. Challenge unsupported favorable and unfavorable interpretations with the same care.

Completed calibration record: two changes and one unresolved decision

In this fictional meeting, the group completes the following record. Sora retains approval authority; discussion alone does not finalize C's rating.

Case Initial recommendation Recorded calibration result Evidence and reason Owner and next step
A Above Sora approves Meets A1–A4 support assigned delivery and recovery after a missed checkpoint; no sustained broader-contribution pattern is established Mina revises the rationale and prepares the employee discussion; successful rollout remains explicitly credited
B Meets Sora approves Above B2–B4 establish a method beyond assigned scope and adoption across three teams over two evidence windows; no presentation requirement applies Mina includes the contribution and its limits; no efficiency metric or pay decision is added
C Meets, relying on proposed goal retirement Approval pending; no final rating recorded C3 lacks the necessary scope-change authorization; C4 proves replacement delivery but cannot resolve the original obligation Mina requests the dated approval from planning owner Oren by July 10; Ruth owns the process clarification for Sora if that evidence is unavailable

The meeting takes place on July 8, after the example's cycle closes. C's record includes the question sent to Oren: “Was the fourth improvement item formally removed in February, who approved that decision, and where is the dated record? Please distinguish the proposal from the final decision.” It contains no request to recreate an approval that did not occur.

By the meeting's close, C remains pending. Ruth will advise Sora on the allowed missing-evidence procedure if no record is found; the group has not promised an extension or predicted the rating. That is a completed decision log with a visible unresolved item, rather than a falsely completed assessment.

A calibration decision sheet carries a retained evidence note while a separate open question remains in the pending tray.

The log also records what did not change: the rating definitions, the review period and the underlying delivery dates. A's change to Meets and B's change to Above are separate corrections to the application of one standard. They are not a trade or an attempt to produce a balanced chart.

Close the loop with the employee and the next cycle

Before the review conversation, confirm the approved version and required communication process. Explain the employee's work, applicable expectation and reasoning without disclosing other employees' packets. Explain the recorded decision without blaming an unnamed committee. If you disagree with the recorded outcome, use the authorized review route with the evidence while keeping the distinction between your recommendation and the approved decision clear.

Preserve important positive and contrary evidence in the final rationale. A's rollout does not disappear because the rating changed. B's method adoption does not prove every engineering skill. C's unresolved scope approval should remain a process question until the evidence and applicable procedure resolve it.

For the next cycle, correct the preparation weakness you found: dated scope decisions, earlier-period notes or recipient evidence for less visible work. Before your calibration meeting, finish the period map, state what would change each boundary recommendation, and mark every rating that still needs approval. Use this LandOffer.ai guide as a publisher reference, then take your packet through your employer's actual confidential review and approval process.

Sources and Further Reading