The On-Call Rotation Rule of Six

Six engineers minimum. Below six, the rotation is unsustainable. The math, the symptoms of an under-staffed rotation, and the path back.

The math

The math is concrete. A 5-person rotation puts each engineer on call 1 week in 5: 10 weekends and 70 weeknights per year. A 6-person rotation drops to 1 in 6: 8 weekends and 56 weeknights. The difference between heavy-but-possible and sustainable is exactly that one extra body. Below five, the rotation collapses fast.

• 5-person: 1 in 5 weeks. 10 weekends and 70 weeknights on call per year. Heavy but possible for a finite stretch.
• 6-person: 1 in 6 weeks. 8 weekends and 56 weeknights per year. Materially different from 5-person, and sustainable.
• Below 5: collapse. Rotation falls apart fast. People quit; the survivors carry more; the rotation gets smaller.
• Published per-person count. Visible weekend-count per engineer per year. Catches drift toward unsustainable before it becomes a departure.

Under-staffed symptoms

Three visible symptoms signal a rotation under stress: burnout (tired-even-off-shift pattern), departures (on-call rotation as proximate cause), quality drops (tired on-calls miss things during incidents). Each is a leading indicator of the rotation falling apart further. Quarterly symptom audit catches the slow slide.

• Burnout. Tired-even-off-shift pattern across the rotation. Sleep debt accumulates faster than recovery.
• Departures. On-call rotation as proximate cause for most SRE departures. Exit interviews surface it consistently.
• Quality drops. Tired on-calls miss details during incidents. Incidents take longer; postmortems get worse.
• Quarterly symptom audit. Explicit "are we seeing these?" check per quarter. Catches the slow slide before it becomes irreversible.

Path back to six

The path back is clear but slow: hire, cross-train, outsource non-critical hours, document the plan with a target and timeline. Each option has different speed and quality trade-offs. The documented plan is what prevents indefinite under-staffing from becoming the steady state.

• Hire. Most direct path. Takes 3-6 months to source, hire, and onboard a new SRE.
• Cross-train product engineers. Faster capacity than hiring. Quality is mixed but better than running 4-person rotation.
• Outsource non-critical hours. Vendor night-shift coverage for tier-2 alerts. Imperfect but maintainable.
• Documented plan. Named "back to six" target and timeline per team. Catches indefinite under-staffing before it normalises.