On-Call Cost Calculator: $61K Per Engineer Per Year (2026)
Updated June 2026 | Sources: Google SRE Book (Ch. 6), SHRM replacement-cost formula, Gloria Mark 2023, Levels.fyi
An engineer handling 42 pages per week at $180K fully-loaded costs $61,000/year in alert-handling time alone, before burnout, replacement, or incident amplification.
The Four Cost Components
Minutes per page multiplied by pages per week multiplied by hourly rate. Includes: reading the alert, opening the runbook (or searching without one), investigating, attempting resolution, documenting the outcome. At MTTA + 23 minutes per page and 42 pages per week, this is 24.5 hours per week of engineering time per on-call engineer. That is more than half a working week consumed by alerts.
Gloria Mark (UC Irvine, 2023) found that after an interruption, engineers require an average of 23 minutes to return to full cognitive engagement with the prior task. A 5-minute alert investigation actually costs 28 minutes of productive time. For engineers doing complex distributed-systems work (incident investigation, capacity planning, architecture), this penalty is severe.
Night and weekend pages carry a cost premium beyond the direct time. Sleep disruption degrades next-day cognitive performance (Stanford research: one night of disrupted sleep reduces complex problem-solving ability by 30-40%). On rotations with frequent night pages, on-call engineers can spend extended stretches operating below cognitive capacity.
The SHRM replacement-cost formula puts senior technical talent replacement at 1x-1.5x annual salary, covering: recruiter fees (20-30% of base), interviewing time (40-80 engineering hours), offer overhead, signing bonus, new hire ramp (6-9 months to full productivity). At a $200K senior SRE base, replacement cost is $200K-$300K. The calculator lets you estimate how many of your own engineers sit in the at-risk cohort.
Burnout Signals Table
Signals that indicate on-call load has crossed into burnout territory. Track these against your own pager and HR data; the thresholds below are practical triggers, not survey-derived prevalences.
| Signal | Prevalence | Threshold to act |
|---|---|---|
| Weekly sleep disruption from pages | Track via 1:1s | Any frequency above monthly |
| Considered leaving due to on-call load | Track via eNPS | Track quarterly; >20% = urgent |
| Acknowledging alerts without investigation | Not surveyed | >10% of ack'd alerts closed without action |
| Skipping post-mortems | Not surveyed | Any skip on a P1/P2 incident |
| Declining to participate in game days | Not surveyed | Two consecutive skips |
| Requesting rotation removal | Not surveyed | Immediate: single request is a data point |
| Performance degradation in code review | Not surveyed | Noticeable quality drop after on-call weeks |
| Sick days spiking after on-call weeks | Not surveyed | Correlation over 3+ months |
| Late alert acknowledgement (MTTA rising) | Industry: 8-15min median | MTTA > 20min on stable systems |
| Expressed resentment in retros / 1:1s | Qualitative | Any: this is a direct signal |