Alert Volume → Burnout Correlation

Studies show alert volume correlates with burnout. The data.

What the data shows

PagerDuty's State of Digital Operations and Catchpoint's SRE survey both show direct correlation between page volume and on-call attrition. Above 5 pages per night, retention drops sharply.

Internal Google SRE data published in the SRE Workbook shows the same pattern: teams above 2 incidents per shift report higher burnout scores.

Burnout is lagging. By the time the engineer quits, the team has tolerated noisy alerting for 6 to 12 months.

The mechanism

Sleep interruption is the dominant factor. A single overnight page can reduce next-day cognitive performance by 20%; three pages destroy the day.

Cognitive load compounds. Engineers who page-shift cannot deep-focus the following day, which delays platform work, which produces more pages.

Predictability matters more than volume. 10 expected pages on Tuesday is less damaging than 3 random pages over a week.

What to track

Pages per shift, p95 across the rotation. Target under 2 per night.

After-hours pages as a ratio of business-hours pages. Above 30% suggests poor batching or genuine underinvestment in resilience.

On-call survey every quarter. Two questions: "Was your sleep affected?" and "Could you have prevented the page?". Bin and trend.

Interventions that work

Quarterly noise audits with a fixed budget for tuning. Make tuning a tracked metric, not background work.

Pay for on-call. Even modest stipends signal that the time is valued; the cultural effect is larger than the dollar amount.

Spread the load. Two-week rotations are easier to absorb than week-long ones; primary plus secondary plus manager-on-call halves the load on each.

Apply this quarter

Pull last quarter of pages from PagerDuty. Compute median, p95, and after-hours ratio per rotation.

If any rotation exceeds 5 pages per shift p95, freeze feature work on the owning service until the noise budget is back under target.

Run the burnout survey once. The first run sets a baseline; subsequent runs measure intervention effectiveness.