HR Engineering Performance Diagnose

Overview

Systematically diagnose team and individual performance issues using quantitative and qualitative signals, separate systemic problems from individual gaps, and generate concrete, fair interventions.

When to Use

Use this skill when:

• You’re seeing missed commitments, low velocity, or quality issues and want to understand why.
• You need a structured performance diagnosis for a specific person or team.
• You want to differentiate systemic constraints (scope, process, leadership) from individual performance.
• You’re preparing for a performance conversation, PIP, or team health review.

Do not use this skill when:

• The primary concern is burnout risk → start with hreng-burnout (then come back here if needed).
• You’re designing career ladders or promotion criteria → use hreng-ladder.
• You’re doing a skills inventory vs. roadmap → use hreng-skills.

Inputs Required

• Time window to analyze (e.g., last 1–3 quarters).
• Performance signals (both quantitative and qualitative).
• Role expectations and level definitions (ideally from your ladder).
• Any recent org or product changes that could affect performance.

Outputs Produced

• A structured diagnostic JSON capturing patterns, root causes, and recommended actions.
• A manager/HR-friendly narrative report summarizing findings and risks.

Tooling Rule

• If MCP tools are available, prefer them for performance data sources (issue tracker, CI, incidents, HRIS).
• Use the templates in this repo for all structured outputs instead of ad hoc formats.

Core Pattern

1. Collect multi-source data (velocity, quality, reviews, feedback, attendance).
2. Cluster patterns into individual vs. systemic findings.
3. Identify root causes across skills, expectations, context, and leadership.
4. Map causes to interventions with owners, timelines, and success metrics.
5. Document clearly for managers/HR, including risks of inaction and fairness checks.

Data Collection

Quantitative Signals

Look across:

• Velocity/throughput: stories per sprint, PRs/month, throughput trends.
• Quality: bug rates, incident frequency, regressions, escaped defects.
• Code review: time to review, review depth, number of rework cycles.
• On-time delivery: % of work delivered on time vs. re-scoped or delayed.
• Availability: attendance patterns, responsiveness to critical comms.

Important:

• Compare individuals to team baselines, not arbitrary expectations.
• Use trends over time, not single bad sprints.

Qualitative Signals

Gather:

• Peer feedback (360, ad hoc comments).
• Stakeholder satisfaction (PM, design, support, GTM).
• 1:1 notes from the manager (with themes, not private details).
• Code review comments (patterns: “needs more tests”, “unclear design”, etc.).
• Meeting participation and cross-functional collaboration patterns.

Pattern Analysis

Individual Underperformance

Indicators:

• Performance concerns are isolated to one person vs. the whole team.
• Peer/stakeholder feedback repeatedly mentions specific behaviors.
• Clear gaps vs. documented level expectations (from career ladder).
• Issues persist even when constraints are removed (e.g., others succeed under same conditions).

Sanity checks:

• Has the person had clear expectations and feedback?
• Are they working on problems appropriate for their level?
• Are there signals of burnout or wellbeing issues that should be addressed first?

Systemic Issues

Indicators:

• Multiple team members showing similar performance symptoms.
• External factors: org reshuffle, unstable roadmap, shifting priorities.
• Resource constraints: missing tooling, unclear ownership, under-staffing.
• Process problems: constant interruptions, context switching, broken workflows.

Rule of thumb:

• If 3+ people show the same pattern, treat it as systemic until proven otherwise.

Root Cause Identification

Common categories:

1. Skill gaps
   • Missing technical skills, design skills, or communication skills.
   • Often fixable via training, pairing, and targeted work.
2. Unclear expectations
   • No concrete level expectations, vague goals, shifting targets.
   • People “underperform” against expectations they never actually received.
3. Personal/wellbeing issues
   • Health, family, mental health, burnout (see hreng-burnout).
   • Requires supportive, confidential handling, not just performance framing.
4. Management gaps
   • Irregular feedback, poor prioritization, unclear strategy, weak sponsorship.
   • Team-level issues are often management performance problems.
5. Role misfit
   • Wrong level, wrong role (IC vs. manager), or misaligned strengths.
   • May call for role redesign, level correction, or transfers.
6. External blockers
   • Dependencies, cross-team bottlenecks, unowned work, flaky infra.

For each finding, explicitly identify:

• Primary root cause category (from list above).
• Contributing factors (other categories that matter).

Diagnostic Output Structure

Use templates/metrics-template.json and templates/performance-diagnosis-report.md as baselines. A typical JSON summary:

{
  "team": "Platform",
  "period": "Q4 2025",
  "findings": [
    {
      "type": "individual",
      "employee": "Name",
      "pattern": "Consistently missing sprint commitments",
      "evidence": [
        "Missed 4 of last 6 sprint goals while peers hit 80%+",
        "Peer feedback: unclear communication about blockers"
      ],
      "root_cause": "Skill gap in task estimation and reluctance to surface blockers early",
      "recommendation": "Mentorship on estimation, explicit expectations about raising risks, smaller scoped tickets for next 2 sprints",
      "urgency": "medium"
    },
    {
      "type": "systemic",
      "pattern": "Team velocity down 30% this quarter",
      "evidence": [
        "Average velocity: 45 → 32 story points",
        "Incident load doubled (2.1 → 4.3 incidents/week)"
      ],
      "root_cause": "Increased on-call burden and unclear Q4 priorities",
      "recommendation": "Reduce scope, clarify top 3 priorities, run incident postmortems to reduce repeat pages",
      "urgency": "high"
    }
  ]
}

Interventions & Fairness Checks

For each individual finding, specify:

• Concrete interventions: coaching, mentorship, training, project rotation, or PIP (as last resort).
• Timeline: when you’ll re-evaluate (e.g., 4–8 weeks).
• Success metrics: what “improved” concretely looks like.

For each systemic finding, specify:

• Process or structure changes: clarify priorities, fix ownership, improve tooling.
• Leadership actions: manager training, clearer strategy, reducing WIP.
• Risk of inaction: impact on attrition, quality, or roadmap.

Fairness checklist before labeling someone “underperforming”:

• Are expectations documented and shared?
• Has the person received specific, recent feedback?
• Are peers with similar conditions performing better (evidence)?
• Have systemic constraints and burnout been evaluated?
• Is there any bias risk (location, identity, communication style, etc.)?

Using Supporting Resources

Templates

• templates/metrics-template.json – Structured performance metrics to track consistently.
• templates/performance-diagnosis-report.md – Narrative diagnosis report for managers/HR.
• templates/diagnostic-framework.md – Checklist-style root cause analysis guide.

References

• references/overview.md, references/checklist.md, references/pitfalls.md – Process and pitfalls.
• references/intervention-library.md – Intervention options and fairness checks.
• references/output-schema.md – Diagnostic output structure.

Scripts

• scripts/check-hreng-perf-diagnose.sh – Pre-run checks for the skill.
• scripts/validate-hreng-perf-diagnose.py – Validate diagnostic JSON and report structure.

Common Mistakes

• Jumping straight to PIP or “low performer” labels without checking systemic factors.
• Treating a few bad sprints as a verdict instead of a signal.
• Overweighting communication style over actual outcomes (especially across cultures).
• Ignoring burnout indicators and reading them purely as poor performance.
• Producing diagnoses with no explicit owner or follow-up date.

Your goal is to produce a diagnosis that a reasonable third party would consider fair, evidence-based, and actionable if they read it six months later.