Purpose
Convert employee profile data into a defensible attrition risk score with named drivers and specific retention actions — so People teams intervene on the right employees before resignation, not after. Replaces gut-feel flight-risk conversations with a repeatable, factor-weighted assessment.
When to Use
Invoke this skill when the user asks to:
- Assess the attrition or flight risk of a specific employee
- Review a team, department, or cohort for turnover exposure
- Identify which high performers are most likely to leave
- Build a retention plan or intervention list from HR data
- Prioritize retention budget or stay conversations across a group
- Explain why a particular employee is a resignation risk
Inputs Required
Before starting, collect:
- Tenure: Time in role and time at company (in months or years)
- Role and level: Job function and seniority band
- Performance rating: Most recent rating on the org's scale (state the scale)
- Compensation vs. market: Current pay relative to market benchmark (below / at / above, or percentage delta)
- Engagement score: Latest survey/pulse score with the scale used
- Manager quality signal: Manager effectiveness score, span of control, or 1:1 cadence
- Contextual flags (optional): Recent reorg, promotion denial, internal transfer request, life event, or peer departures
If a scale is not specified, assume a 1–5 scale for performance and engagement and state the assumption. If compensation benchmark is unavailable, flag it as an unscored factor rather than guessing.
Instructions
Step 1: Score the Six Risk Factors
Assign each factor a 0–20 sub-score (higher = higher risk), then apply the weights below. Score conservatively — only push toward the high end when evidence is clear.
Factor 1 — Compensation Gap (weight 25%)
- Below market by 10%+: 15–20 (highest predictive weight for voluntary exits)
- Below market by <10%: 8–14
- At market: 3–7
- Above market: 0–2
Factor 2 — Engagement (weight 20%)
- Bottom quartile / declining trend: 15–20
- Below midpoint: 9–14
- At or above midpoint: 3–8
- Top quartile: 0–2
Factor 3 — Manager Quality (weight 20%)
- Low manager score, wide span, or infrequent 1:1s: 14–20 ("people leave managers")
- Mixed signals: 7–13
- Strong manager relationship: 0–6
Factor 4 — Tenure Stage (weight 15%)
- 0–12 months (onboarding fragility) or 24–48 months (the "itch" window): 12–20
- 12–24 months: 6–11
- 48+ months with recent growth: 0–5
Factor 5 — Performance vs. Recognition Fit (weight 10%)
- High performer, no recent promotion/comp movement: 14–20 (regretted-attrition zone)
- High performer, recently recognized: 2–6
- Low performer disengaging: 8–14 (risk exists but often non-regretted)
Factor 6 — Contextual Flags (weight 10%)
- Active flags (promotion denial, reorg, peer departures, transfer request): 12–20
- One minor flag: 6–11
- None: 0–5
Compute the weighted total and normalize to a 0–100 score.
Step 2: Assign a Risk Tier
- 80–100 — Critical: Likely evaluating options now. Intervene within 1–2 weeks.
- 60–79 — High: Meaningful flight risk. Intervene within 30 days.
- 40–59 — Moderate: Watchlist. Address in next review cycle.
- 0–39 — Low: Stable. Monitor at normal cadence.
Flag regretted attrition risk separately when a Critical/High score coincides with a strong performance rating — these are the most costly to lose.
Step 3: Identify the Top 3 Risk Drivers
Rank the factors by their contribution to the total score. For each driver, write one sentence connecting the specific data point to the risk (e.g., "12% below market at 30 months tenure — the highest-leaving combination").
Step 4: Recommend Retention Interventions
Provide 3–5 interventions, each mapped to a specific driver. For each:
- Action: The concrete move (off-cycle adjustment, stretch project, manager reassignment, promotion timeline, stay conversation)
- Owner: Who runs it (HRBP, direct manager, skip-level)
- Timing: When it must happen given the tier
- Expected impact: Which driver score it lowers and by roughly how much
Prioritize actions that address the highest-weighted driver first. Do not recommend a comp adjustment if the driver is manager quality — match the intervention to the cause.
Step 5: State Confidence and Data Gaps
Note any unscored factors (e.g., missing benchmark), and state whether the score is high, medium, or low confidence based on data completeness.
Output Format
- Structure: Score summary up top, factor breakdown table, drivers, interventions, confidence note
- Length: 500–800 words per employee (or a summary table plus per-person rows for a cohort)
- Format: Score header + factor scoring table + ranked drivers list + intervention list
- Sections: Risk Score & Tier → Six-Factor Breakdown (table) → Top 3 Drivers → Retention Interventions → Confidence & Data Gaps
- Tone: People analytics specialist — evidence-based, direct, non-alarmist; names risk without dramatizing
Quality Checklist
Before delivering output, verify:
- [ ] All six factors are scored and weighted, and the total normalizes to 0–100
- [ ] The risk tier matches the numeric score band
- [ ] Exactly three drivers are ranked by score contribution, each tied to a specific data point
- [ ] Every intervention maps to a named driver — no generic "improve engagement" advice
- [ ] Regretted-attrition risk is flagged when high risk coincides with high performance
- [ ] Any missing input is disclosed as a data gap, not silently guessed
Notes
- Compensation gap and manager quality are the two strongest voluntary-exit predictors — weight your attention there first.
- The 24–48 month tenure window is where many high performers begin looking; a Moderate score in that window deserves a proactive stay conversation.
- Not all attrition is bad — separate regretted from non-regretted risk so retention spend goes to the people worth keeping.
- For cohort runs, lead with a ranked table (name/role, score, tier, top driver) before the per-person detail, so the reader can triage at a glance.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills