# Hr Performance Review

> Structure a fair performance review — self-assessment, evidence-based manager template, and calibration guidance tied to level expectations.

- Skill: `matrixx0070/hr-performance-review` (Agent Skill)
- Install (CLI): `npx skillmds@latest add matrixx0070/hr-performance-review`
- Raw SKILL.md: https://api.skillmd.com/api/skills/matrixx0070/hr-performance-review/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: Matrixx0070 (https://skillmd.com/u/matrixx0070)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/matrixx0070/hr-performance-review

---


## When to use

Reach for this when running a review cycle, coaching a manager to write a fair review, or designing calibration so ratings are consistent and defensible. Keep feedback specific, evidence-based, and tied to expectations for the person's level.

**Not for:** the new-hire ramp (use hr-onboarding), pay decisions that follow a rating (use hr-comp-analysis), or interview scoring (use hr-interview-prep).

## Method

1. **Set the frame** — state the review period, the competencies/goals assessed, and the rating scale with a clear definition for each level.
2. **Self-assessment** — prompt the employee on accomplishments (with impact), goal progress, strengths, growth areas, and support needed. Ask for evidence, not adjectives.
3. **Manager template** — per competency/goal, capture concrete examples, a rating with justification, and forward-looking development actions. Separate performance-in-role from potential.
4. **Gather multi-source input** — incorporate peer/stakeholder feedback where available. Decision point: attribute recurring themes, not one-off gossip, and discount recency or halo effects.
5. **Draft the rating** — map evidence to the scale for the person's level. Decision point: write the summary so it would hold up if read aloud to the employee — if it wouldn't, it needs more evidence or a softer claim.
6. **Calibrate** — compare ratings across a peer group, surface inconsistencies and bias patterns, adjust with documented rationale, and finalize the distribution.

## Example

Employee: Marketing Manager (L4), H1 cycle. Competency "cross-functional influence." Evidence: drove the product-launch content plan across 3 teams, but two stakeholders cited late handoffs. Self-assessment claims "excellent collaboration"; manager rates 3/5 (Meets) with the handoff examples and a development action: run a shared launch calendar next half. Calibration: a peer L4 rated 4/5 on the same competency had cleaner delivery signals — the difference is documented, so the distribution holds up under review.

## Pitfalls

- Ratings built on adjectives ("great teammate") with no observable example behind them.
- Recency bias — weighting the last month over the full period.
- Conflating potential with performance, inflating a rating for someone "on the rise."
- Skipping calibration, so two managers apply the same scale differently and ratings aren't comparable.

## Output format

```
Review frame: period | competencies/goals | rating-scale definitions
Self-assessment template: prompts (accomplishments, goals, strengths, growth, support)
Manager review (per item): examples | rating | rationale | development action
Calibration guide: comparison method | bias checks | adjustment log
Final summary structure + delivery tips
```

## Reference

### The 5-point rating scale (anchored)

Use consistent, level-relative definitions. "Meets" means fully delivering the expectations of the person's *current* level — it is a good rating, not a warning.

| Rating | Label | Definition |
|---|---|---|
| 5 | Exceptional / Greatly exceeds | Consistently delivered well beyond level expectations; outsized, visible impact; raised the bar for others. Rare. |
| 4 | Exceeds | Delivered above expectations on most goals; stretched into next-level scope on some. |
| 3 | Meets / Successful | Fully delivered the expectations of the level. The healthy default for a solid performer. |
| 2 | Partially meets / Developing | Met some expectations, missed others; specific, addressable gaps. New-in-role often lands here. |
| 1 | Below / Does not meet | Consistently missed core expectations; requires a documented improvement plan. |

### Rating-distribution guidance

Most organizations expect a roughly normal spread and use calibration to prevent grade inflation. A common healthy shape for an established team:

| Rating | Typical share |
|---|---|
| 5 Exceptional | 5–10% |
| 4 Exceeds | 20–25% |
| 3 Meets | 50–60% |
| 2 Developing | 10–15% |
| 1 Below | 0–5% |

Treat this as a *reference distribution*, not a forced quota. A forced curve (mandating a fixed % of low ratings) damages trust and is inappropriate for small or genuinely high-performing teams. Use the distribution as a **diagnostic**: if a manager rates 60% of their team "Exceeds," that is a calibration signal — either the team is exceptional (show the evidence) or the manager is inflating. Small teams (<8) should be calibrated against the broader level cohort, not forced to fit a curve internally.

### Calibration rubric

1. **Normalize the scale** — before reviewing people, re-read the level definitions aloud so every manager anchors to the same "Meets."
2. **Present evidence, then rating** — each manager states the case with specifics; the rating is revealed after, to reduce anchoring.
3. **Compare like-for-like** — group by level and role family. Ask: "Would this person's 4 look like a 4 on another team?"
4. **Surface bias patterns** — check for recency, halo/horns, leniency/severity by manager, similarity bias, and any demographic skew in the distribution.
5. **Adjust with documented rationale** — every change up or down gets a one-line evidence-based reason in the log.
6. **Lock the distribution** — finalize; managers deliver only calibrated ratings.

### Bias checklist

- **Recency** — did the last month outweigh the full period? Pull evidence from all quarters.
- **Halo / horns** — is one strong (or weak) trait coloring everything?
- **Leniency / severity** — is this manager systematically high or low vs. peers?
- **Similarity** — are people like the manager rated higher?
- **Idiosyncratic rater effect** — a rating says as much about the rater as the ratee; triangulate with peer input.
- **Potential ≠ performance** — rate what was delivered this period, not future promise.

### Example phrasing (evidence-based, level-tied)

Strong summary lines pair a specific behavior with its impact and a forward action:

- *Meets (3):* "Owned the Q2 migration and delivered on schedule with zero customer-facing incidents. Coordination across the two dependent teams was reactive rather than planned — twice, downstream owners learned of changes late. Next half: publish a shared migration calendar and confirm downstream sign-off before cutover."
- *Exceeds (4):* "Beyond owning the on-call rotation, redesigned the alerting policy, cutting pages 40% quarter-over-quarter; three peers now use the runbook she wrote. This is early next-level (staff) scope."
- *Developing (2):* "Code quality is strong, but three of five committed features slipped past their dates due to under-scoping. This is an estimation gap, not an effort gap. Next half: break work into <3-day tasks and review estimates with the lead at planning."

Avoid: "great teammate," "rockstar," "needs to step up" — none survive being read aloud to the employee with evidence attached. The test for any sentence: could you defend it to the employee's face with a specific example? If not, cut it or add the example.

### SBI feedback frame

For each point, use **Situation → Behavior → Impact**: "In the launch review (S), you presented the plan without the security team's sign-off (B), which meant we caught the gap in the meeting and slipped a week (I)." Concrete, non-judgmental, actionable — the backbone of both written reviews and the verbal delivery.

### Delivery tips

- No surprises: a review should summarize feedback already given, not deliver first-time news.
- Lead with the summary and rating, then the evidence — don't bury the headline.
- Separate the performance conversation from the compensation conversation where possible; comp anxiety drowns out development feedback.
- Give the employee time to respond and record their comments; a review is a dialogue on the record.
- End with 2–3 specific, owned development actions and a check-in date.