PeopleAnalyticsAI
Enterprise People Data & AI

Case 3 · Workforce Intelligence + AI

Why is Engineering voluntary attrition increasing?

Engineering trailing-12m annualized voluntary attrition is 16.0% (as-of 2026-08-31). The month view is 17.9%, up 1.9 pp from 2026-07. Concentrated in APAC-SIN among cells visible at this access level.

based on cells visible at this access level

16.0%

Scope Engineering · Window trailing-12m (annualized) · as_of 2026-08-31

Company trailing-12m (parity) 14.0%.

Scope Company · Window trailing-12m (annualized) · as_of 2026-08-31

Month view (secondary) 17.9%.

Scope Engineering · Window month (annualized) · as_of 2026-08-31

2026-07: 16.0% (as_of 2026-07-31)

Identity demo-external-viewer · min_cell 50 · 44 of 104 location × tenure × grade cells suppressed. Rates use trailing-12m; suppression uses as-of month headcount n, not trailing-12m average n.

Cell suppression is not cosmetic. Across the four demo identities the hidden cells are 44 / 42 / 34 / 30. The headline location is computed from cells still visible after min-cell suppression at this access level. At People analyst min_cell 5, two six-person slices (AMER-NYC · 3–5y · G10 and EMEA-LON · <1y · G3) enter the ranked list. Neither is visible to the site visitor. Suppression changes the conclusion — that is the governance demonstration.

What changed

Engineering voluntary attrition · trailing-12m annualized · 24 months

13.7%15.1%16.5%13.7,15.1,16.52026-03 scenario start2024-09-30: 14.9%2024-10-31: 14.7%2024-11-30: 14.6%2024-12-31: 14.8%2025-01-31: 14.8%2025-02-28: 16.0%2025-03-31: 14.7%2025-04-30: 14.6%2025-05-31: 14.6%2025-06-30: 14.4%2025-07-31: 14.3%2025-08-31: 14.3%2025-09-30: 14.2%2025-10-31: 14.5%2025-11-30: 14.6%2025-12-31: 14.4%2026-01-31: 14.4%2026-02-28: 14.5%2026-03-31: 14.7%2026-04-30: 14.9%2026-05-31: 15.3%2026-06-30: 15.6%2026-07-31: 15.7%2026-08-31: 16.0%2024-092026-08

Y-axis starts at 13.7%, not 0. Unit: trailing-12m annualized rate.

Vertical marker: 2026-03 scenario start

Where × tenure × grade (suppression on)

Window: trailing-12m (annualized). Rate window is trailing-12m. Suppression uses as-of month n, not trailing-12m average headcount.

  • APAC-SIN · <1y · G7 · 34.3% · n=248Window trailing-12m (annualized)
  • APAC-SIN · <1y · G4 · 34.1% · n=233Window trailing-12m (annualized)
  • APAC-SIN · <1y · G5 · 33.8% · n=241Window trailing-12m (annualized)
  • APAC-SIN · <1y · G6 · 29.1% · n=226Window trailing-12m (annualized)
  • APAC-SIN · 1–3y · G5 · 27.4% · n=298Window trailing-12m (annualized)
  • APAC-SIN · 1–3y · G4 · 25.5% · n=241Window trailing-12m (annualized)
  • AMER-NYC · <1y · G7 · 25.3% · n=354Window trailing-12m (annualized)
  • APAC-SIN · 1–3y · G6 · 25.1% · n=296Window trailing-12m (annualized)

Top 8 visible cells · 44 cells hidden under min-cell.

Related signals

  • Compa-ratio comparison is available to internal People identities.
  • Reorg-class manager change control: 1,913 events / 47,674 workers
  • Reorg-class manager change slice: 213 events / 2,149 workers
  • BLS JOLTS quits rate: BLS series calibrate GlobalTech attrition. They are not this company's employees.

What we know

  • The voluntary attrition definition, period, and grain are certified.
  • Rates are not uniform across Engineering locations and tenure bands.
  • Reorg-class manager change is a related signal, not a cause.

What we do not know

Hypotheses, not proven causes.

  • Why a specific employee left — the mart does not contain exit-interview text.
  • Whether pay, mobility, or skill gaps caused the pattern. They can only be compared.
  • Whether a local labor-market shock is operating outside this dataset.

Could we build critical skills internally?

Internal workforce: Synthetic · O*NET: Public · Microsoft Learn: Public

  • Engineering · Engineering - Platformcoverage 62.9%

Relevant learning (Microsoft Learn)

What I would investigate next

  1. Inspect the top two Engineering locations with the HRBP before a company-wide program.
  2. Compare promotion and lateral volume on the same tenure and level slices.
  3. Use the current trusted snapshot only; keep the APAC HRIS replay out of the board pack.

Reasonable leadership actions

  • Ask HRBPs in the highest-rate locations for stay-risk context before changing policy.
  • Review promotion velocity on the same tenure slices as a mobility signal, not a cause.
  • Offer targeted learning against the largest critical-skill gaps where coverage is thin.