| name | diagnose-attrition |
| description | Diagnose an attrition spike with cost and cause — split regretted from non-regretted, localize to team/manager, price the loss, name the driver. Reach for this on a retention question. |
Skill: Diagnose attrition
A turnover number with no split, no cost, and no segment is noise (§3 #1).
Step 1 — Split regretted vs non-regretted
Non-regretted (managed-out) is often intended; only the regretted share is a loss to recover (§3 #1).
Step 2 — Localize it
Segment by team / manager / level / tenure cohort. A concentrated spike is a manager/span problem (§3 #7); broad-based is structural.
Step 3 — Name the driver
Comp, manager, growth, or workload — usually two co-occur. Route a comp driver to total-rewards-comp-analyst.
Step 4 — Cost it
Regretted exits × replacement cost (recruiting + ramp + vacancy). Use ../../scripts/people_calc.py attrition.
Output
An annualized, regretted-split, segmented attrition read with a dollar cost and a named driver — and a ranked action list with owners/dates. Traverse Tree 1 in ../../knowledge/people-ops-decision-trees.md.