Analyze this employee attrition data and tell me what's signal vs. noise.
Inputs:
- Departures over {{period, e.g. last 12 months}}: {{paste — name, dept, level, tenure, stated reason, manager}}
- Total headcount by department/level at the start of period: {{paste — for rate calculations}}
- Industry benchmark for attrition rate (optional): {{paste}}
Output:
## Rate vs. benchmark
| Segment | Departures | Avg headcount | Rate | Industry benchmark | Delta |
|---|---|---|---|---|---|
(By dept, by level, by tenure band, by manager if data allows.)
## Concentration patterns
Where is attrition clustering:
- Department(s) with rate >1.5x company average
- Manager(s) with disproportionate losses
- Level(s) over-represented
- Tenure cliff (e.g. 1–2 year mark)
For each cluster: is it a pattern (≥3 departures with shared reason) or a coincidence?
## Stated vs. likely reasons
Stated reasons (from exit surveys) vs. underlying reasons (from patterns + manager + tenure clustering). They often diverge.
## Single-point-of-failure risk
Names not on this list yet that are at elevated risk: the only person in a critical area, a manager whose direct reports just left, anyone reporting to a manager-cluster identified above.
## What's missing from the data
- Departures we don't have exit-interview data on
- Segments too small to draw conclusions from
- Reasons probably under-reported (comp, manager problems are systematically under-disclosed in exit interviews)
## Actions, ranked by reversibility
- This week: who to talk to (skip-level 1:1s)
- This month: structural changes worth considering
- This quarter: bigger bets (comp, structure, scope)
Hard rule: if a manager is in a cluster, surface it. Don't soft-pedal because they're senior.attritionpeople-analyticsretention