Real attrition cohort math: what 42 hires over 36 months told us

60-Second Summary

Trailing 36-month annualized attrition: 19.4%. The headline hides the structure: 8% in months 0–12, 24% in months 14–22, 11% thereafter.
The second-year wall (months 14–22) accounts for 58% of all exits across all cohorts. Conventional 'first-year cliff' models predicted 12%.
Exit interview coding shows comp is the cited reason in 67% of second-year exits but the actual root cause (per skip-level interviews 90 days post-exit) is stalled growth in 71% of the same cohort.

Every HR leader quotes an attrition number. Almost no one shows the cohort math. Here is ours — 42 hires across 36 monthly cohorts, with the exit timing, the exit reasons, and a 90-day post-exit follow-up that surfaced the difference between 'what they told HR' and 'what was actually true.'

The dataset

Window: May 2023 – April 2026 (36 months).
Hires in window: 42. Exits in window: 19. Survivors at end: 23.
Roles: 32 engineering, 6 product, 4 ops.
Follow-up: skip-level interview 90 days post-exit with 14 of 19 leavers (74% response rate).

Cohort survival curves

Benchmark: LinkedIn Tech Industry Attrition Report, South Asia subset, 2024.

Months since hire	% still employed	Industry benchmark	Delta
0–6	100%	94%	+6 pts
7–12	92%	87%	+5 pts
13–18	78%	82%	−4 pts
19–24	60%	78%	−18 pts
25–36	55%	70%	−15 pts

The second-year wall

58%

of all exits occurred in months 14–22

n=11 of 19

20 months

Median tenure at exit, all cohorts

vs. 14-month industry median

Exits in months 0–6

Probationary period worked

The conventional 12-month cliff model would have predicted 5 exits in months 11–13. We had 1. The same model predicted 3 exits in months 19–22. We had 8. The shape of attrition in our org is fundamentally different from the model HR textbooks teach.

Cited vs. actual reasons

Stated reason (exit interview) vs. actual reason (90-day skip-level)

Cited at exit

Compensation: 67%
Better opportunity: 23%
Personal/family: 10%

Identified 90 days later

Stalled growth / no visible promo path: 71%
Manager mismatch: 14%
Compensation as root cause: 15%

Why the gap matters

Exit interviews are nearly worthless as a diagnostic tool. Comp is the polite, safe, externally-defensible reason. The honest reason almost always involves the manager or the absence of a credible growth path — and is only surfaced once the employee has nothing to lose by saying it.

What we changed

Introduced a written career ladder for engineering at month-of-hire +6, with explicit criteria for L3→L4 and L4→L5 promotions.
Quarterly promo calibration meetings — previously ad hoc, now on a schedule.
Manager 1:1 effectiveness scored anonymously by reports each quarter; lowest-scoring managers receive coaching, not removal.
Compensation refresh moved from annual to bi-annual to close the second-year gap before it became an exit conversation.
Result so far (8 months in): exits in months 14–22 dropped from 4 (prior 8 months) to 1 (current 8 months). Sample size still small — early signal only.