Performance Review Skill
Use this skill when you are writing an annual or mid-year review, preparing for a calibration session, or coaching a manager on feedback. The skill provides methodology for performance management, talent calibration, and goal-setting — the organisation-specific data comes from you.
---name: performance-review-skilldescription: Help managers draft fair, specific performance reviews with calibration support and goal-setting guidance. Use when writing annual or mid-year reviews, preparing for a calibration session, or coaching a manager on feedback. Trigger with "write performance review for", "help me with this review", "calibration prep".---
# Performance Review Skill
Design performance management systems, facilitate calibration sessions, conduct talent assessments, and build goal-setting frameworks. Organisation-specific data (ratings, employee records, goal frameworks, competency definitions) comes from you, a connected MCP source, or companyRAG collections.
## Where the Data Comes From
| Source | What it contributes || --- | --- || **HRIS via MCP** (e.g. Workday, BambooHR, Personio) | Employee records, job levels, compensation data, historical ratings || **OKR / goal tools via MCP** | Objectives, key results, goal attainment, KPI history || **companyRAG / file upload** | Competency frameworks, job architecture, calibration guidelines, prior reviews |
> **No connected source?** Provide the data in chat or upload the relevant files — the skill works the same way.
## EU AI Act & GDPR Compliance Boundaries
AI-assisted performance monitoring and workforce management is classified as **high-risk under the EU AI Act (Annex III, Category 4 -- employment and workers' management)**. Per GDPR Article 22, individuals have the right not to be subject to decisions based solely on automated processing that produce legal or similarly significant effects. All outputs are decision-support tools for human practitioners — performance ratings, talent classifications, and development decisions must be made by qualified human practitioners.
## Performance Management Model Selection
Performance management is a DESIGN CHOICE, not a progression from "traditional" to "modern." Each model has legitimate use cases, and the right choice depends on organizational context, not on trends. Present all models as options with trade-offs.
### Model Options
| Model | Core Approach | Best Fit | Key Design Consideration || --- | --- | --- | --- || **Cycle-Based** | Goal-setting → check-in → review → calibration | Structured compensation cycles; regulated industries | Recency bias risk in annual reviews; administrative burden || **Continuous Feedback** | Ongoing 1:1 coaching, real-time recognition, regular check-ins | Fast-moving organizations; roles with rapidly changing priorities | Harder to aggregate for compensation; requires high manager capability; calibration more complex without discrete data points || **OKR-Based** | Quarterly objectives with measurable key results | Product and technology organizations with clear strategic cadence | Separate OKR scores from compensation to prevent sandbagging; not all roles suit quarterly objectives || **Hybrid** | Combining elements (e.g., continuous 1:1s + annual calibration) | Organizations transitioning models or with diverse role types | Higher design complexity; risk of "worst of both worlds" if not intentionally designed |
### Selection Criteria
When helping an organization choose, assess:
| Factor | Points Toward Cycle-Based | Points Toward Continuous | Points Toward OKR-Based || --- | --- | --- | --- || Compensation linkage | Strong -- needs discrete data points | Weak -- harder to aggregate | Moderate -- quarterly data available || Manager capability | Lower bar -- structured process guides managers | Higher bar -- requires coaching skill | Moderate -- objective-setting is learnable || Works council requirements | Natural fit -- formal process documentation | Compatible -- requires structured documentation of ongoing feedback for formal consultation | Depends on jurisdiction || Rate of change | Slower -- annual goals remain stable | Faster -- priorities shift frequently | Moderate -- quarterly adaptation || Organization size | Larger -- needs consistency at scale | Smaller -- relationship-driven | Medium -- strategic alignment matters |
> [Methodology framework] This selection framework is a design tool, not a recommendation. No model is inherently superior. The organization must choose based on their specific context, culture, and constraints.
## Goal-Setting Methodology
Goal-setting frameworks are components of performance management, not systems in themselves. The methodology for designing goal-setting varies by the chosen performance model.
### Goal Framework Options
Each goal framework has a critical design consideration for performance management:
- OKR : Separate OKR achievement scores from performance ratings to prevent sandbagging (setting conservative goals to achieve high scores)- KPI-Based : Distinguish between leading indicators (activities) and lagging indicators (outcomes) -- roles should be measured on indicators they can influence- Competency-Based : Competency assessment requires behavioral evidence, not self-report -- link to the behavioral anchor methodology used in structured hiring
### Goal Quality and Cascading
Apply standard goal quality criteria (SMART or equivalent) to ensure every goal is specific, measurable, and time-bound. The methodology below addresses the non-obvious design decisions:
When goals carry different weights in the performance assessment:
1. Communicate weights at goal-setting time, not at evaluation time2. Ensure weights sum to 100% and are documented3. Cascading: organizational goals → functional goals → team goals → individual goals. Each level should be able to see how their goals connect upward.4. Avoid cascading artifacts: a CEO's goal of "20% revenue growth" should not become every individual's personal goal -- translate to the individual's contribution scope
## Calibration Session Methodology
Calibration methodology is detailed here because structured facilitation training is uncommon in most organizations. Most managers have never been trained on structured calibration facilitation.
### Purpose of Calibration
Calibration ensures rating consistency across managers and teams. Without it, leniency/strictness variation between managers creates inequitable compensation and promotion outcomes and prevents meaningful cross-team comparison.
### Pre-Session Preparation
1. Data compilation : Aggregate ratings by manager, team, function, and job level. Produce distribution charts showing rating spread per group.2. Outlier identification : Flag ratings that deviate significantly from the group distribution for discussion (not for automatic correction)3. Cross-team comparison : Compare distributions across teams at the same job level -- significant differences indicate potential calibration issues4. Manager pre-work : Each manager prepares evidence-based justification for any rating above or below the middle category, referencing specific accomplishments, behaviors, and goal outcomes
### Session Facilitation
The facilitator (typically HR Business Partner or talent management specialist) owns the process; managers own the content.
1. Ground rules : Ratings are based on evidence against job level expectations, not on comparison to peers, personal relationship, or tenure2. Job level anchoring : Begin each discussion segment by restating what the relevant job level expects (from job architecture)3. Review sequence : Start with outlier ratings (highest and lowest), then review the middle distribution4. Evidence standard : For each rating under discussion, the manager presents specific behavioral evidence. "She's a strong performer" is not evidence. "She delivered [ specific outcome ] against [ specific goal ] while navigating [ specific challenge ] " is evidence.5. Challenge protocol : Other managers and the facilitator may ask clarifying questions. Challenges must be evidence-based: "In my team, that outcome at IC3 would be Meets, not Exceeds -- here's why..."6. Adjustment tracking : Document all rating changes with the reason. Changes must be evidence-driven, not pressure-driven.7. Bias checks : At each segment, explicitly check for common biases (see Bias Mitigation section below)8. Time management : Allocate discussion time proportional to the team size and rating divergence, not equally per manager
For agendas, disagreement handling, post-session documentation, and manager communications, extend the facilitation steps above with your organization's HR templates and timelines.
### Calibration Models
Organizations may calibrate using ratings, narratives, or a combination:
**Rating-based calibration**: Aligns numerical or category ratings across managers. Most structured. Requires rating scale (see below).
**Narrative-based calibration**: Managers present written summaries of performance and discuss whether the described performance matches the intended rating category. More nuanced but harder to facilitate at scale.
**Hybrid**: Rating anchored by narrative evidence. The rating provides the structure; the narrative provides the substance.
## Talent Assessment Approaches
Talent assessment is a menu of approaches, not a single tool. Each has legitimate use cases.
**9-Box Grid** (Performance × Potential)
- When it works: large organizations needing visual overview, succession planning conversations, organizations comfortable with explicit potential assessment- Limitations: "potential" is hard to define objectively, can create labeling effects, cultural resistance in some European contexts- Facilitation: define both axes with behavioral descriptors before the session; potential is not "likability" or "visibility"
**Narrative-Based Talent Review**
- When it works: senior leadership review, roles where performance is multidimensional and hard to reduce to scores, organizations that value qualitative discussion- Limitations: time-intensive, harder to aggregate, facilitator skill is critical- Facilitation: structure the narrative with consistent prompts (strengths, development areas, readiness, risk)
**Competency-Based Assessment**
- When it works: technical organizations, roles where competency proficiency is the primary differentiator, organizations with mature competency frameworks- Limitations: requires a maintained competency taxonomy, can miss leadership and behavioral dimensions- Facilitation: link to competency-level matrix from job architecture
**Succession Readiness Models**
- When it works: focused on critical role continuity, organizations with identified key positions- Limitations: narrow scope (succession only), can create entitlement expectations- Facilitation: assess readiness in time horizons (ready now, ready in 1-2 years, ready in 3+ years) with specific development actions for each
### Choosing an Approach
The choice depends on organizational context:
- Organization size: 9-box scales better; narrative scales worse- Cultural context: some European organizations resist explicit potential labeling- Maturity of HR processes: competency-based requires an existing competency framework- Purpose: succession planning favors readiness models; development planning favors competency-based; general talent review can use any
## Rating Scale Design
When the performance management model uses ratings:
### Scale Construction
| Scale Type | Pros | Cons || --- | --- | --- || 3-point | Simple, forces differentiation | May feel too blunt for nuanced roles || 4-point (no midpoint) | Forces a direction (above/below expectations) | Can feel punitive to the "below middle" category || 5-point | Familiar, allows nuance | Central tendency risk (most ratings cluster at 3) || Behavioral labels (no numbers) | Reduces "score" mentality | Harder to aggregate quantitatively |
### Distribution Approaches
**Guided distribution**: Provide expected ranges as guidelines, not quotas. The organization defines the expected distribution shape; managers justify deviations.
**No prescribed distribution**: Let ratings fall where evidence supports them. Calibration catches inconsistency. Requires strong calibration facilitation.
**Forced distribution** (forced ranking): Mandate specific percentages per category. Effective at preventing inflation but controversial -- can damage trust, create competition, and penalize high-performing teams where everyone genuinely meets expectations.
> [Methodology framework] All distribution approaches are design choices with trade-offs. No single approach is inherently correct. Forced distribution is not "outdated" and guided distribution is not "best practice" -- each serves different organizational contexts. Present trade-offs and let the organization choose.
## Underperformance Management Methodology
This section covers the universal process elements of managing underperformance. Employment law implications vary dramatically by jurisdiction -- specific procedural requirements (Abmahnung in Germany, entretien prealable in France, dossieropbouw in the Netherlands) are jurisdiction-specific and require verification with employment counsel.
### Universal Process Elements
Regardless of jurisdiction, effective underperformance management follows a consistent methodology:
1. Identification trigger : Specific, documented gap between expected performance (from job level expectations) and actual performance (from evidence gathered during the performance cycle)2. Root cause assessment : Performance gaps can stem from capability (skill gap), motivation, clarity (unclear expectations), resources (tools, support, time), or personal circumstances. The intervention must match the root cause.3. Formal documentation : Every step documented with dates, specific behaviors, expected standards, and agreed actions. This documentation is critical regardless of jurisdiction.4. Improvement plan : Specific, measurable improvement objectives with a realistic timeline, defined support (training, coaching, adjusted workload), and clear success criteria5. Regular checkpoints : Scheduled reviews during the improvement period with documented progress assessment6. Outcome decision : At the end of the improvement period, a documented assessment of whether improvement criteria were met, partially met, or not met, with the resulting next steps
> **CRITICAL**: Verify the specific procedural requirements, notice periods, employee representative involvement, and legal protections with employment counsel in the applicable jurisdiction BEFORE initiating any formal underperformance process. Getting the process wrong can expose the organization to significant legal risk.
## Bias Mitigation in Performance Assessment
Performance calibration shares bias types with structured hiring (see structured-hiring skill). The mitigations below are specific to the performance context:
- Recency : Require evidence from the ENTIRE review period, not just the final quarter. Managers should maintain running documentation throughout the cycle.- Leniency/Strictness : The primary purpose of calibration -- cross-manager distribution comparison surfaces systematic over/under-rating.- Central tendency : Require evidence-based justification for middle ratings too, not just outliers. "Meets Expectations" is a positive assertion, not a default.- Halo/Horns : Rate each competency or goal independently before forming an overall rating.- Attribution : Distinguish individual contribution from environmental factors (team resources, market conditions, organizational changes).- Contrast : Anchor each discussion to job level expectations, not to the person discussed immediately before.
Facilitate explicit bias checks during calibration: after each segment, ask "Are we seeing any patterns in who receives the highest/lowest ratings? Are there demographic, tenure, or team-based patterns?"
## Guardrails
- Never generate employee performance assessments, ratings, or talent classifications. All performance data must come from the organisation.- Never present any single performance model as inherently superior or claim "industry standard" for rating scales, distributions, or calibration approaches. Distribution is an organisational design choice.- Flag when recommendations touch employment law : "Verify procedural requirements with employment counsel in [ jurisdiction ] ."- Source-label all output : [Methodology framework] , [From organizational data] , or [AI assessment — verify with HR] .
> **Tip:** Ask for DOCX or Markdown output via companyFILES to get a formatted, ready-to-share document.