Follow-The-Sun On-Call: 3-Region Cost Model 2026
Updated May 2026. Sources: Google SRE Book Chapter 11 (Being On-Call), incident.io 2024 State of On-Call Report, Catchpoint 2024 SRE Report, AASM shift-work research.
How Follow-The-Sun Actually Works
The classical three-region model splits the 24-hour day into three 8-hour primary shifts: Asia-Pacific covers roughly 00:00 to 08:00 UTC, Europe covers roughly 08:00 to 16:00 UTC, and Americas covers roughly 16:00 to 24:00 UTC. Each region's primary on-call engineer is working during their local daytime, so a pager call lands during the work day rather than at 3 AM. Each region also runs a secondary back-up to cover absence (sickness, vacation, escalation latency). A structured handoff at each shift boundary transfers context from the outgoing primary to the incoming primary.
The model is described in detail in Google SRE Book Chapter 11. Google operates a more complex version of this model internally across multiple regions, with overlap windows for handoff and explicit response-time SLOs per shift. Most organisations adopting follow-the-sun do not need that level of formalism; the basic three-region model with a written handoff template and a 15-minute handoff meeting captures most of the value.
The most important property of follow-the-sun is what it eliminates rather than what it provides. It eliminates the night-pager component of on-call burden entirely from the primary rotation. The incident.io 2024 State of On-Call Report found that 62 percent of on-call engineers report regular sleep disruption from night pages; the AASM and broader sleep-medicine literature consistently links sleep disruption to cognitive impairment that lasts well beyond the night of the disruption. Removing the night-pager load is the single most-leveraged retention move available to an SRE team, when it is financially feasible.
Minimum Staffing Per Region
A useful staffing rule of thumb: each region needs roughly 5 engineers on the rotation, which gives each engineer a primary rotation slot of about 33 percent of working hours plus a secondary backup at about the same frequency. This is the lowest you can run without coverage fragility. With 4 engineers per region, any single absence (a week of vacation, a sickness, a sabbatical) cascades into either a coverage gap or a rotation hot-spot that produces measurable burnout in the affected region.
| Per-region size | Total team | Burden per engineer | Coverage risk | Verdict |
|---|---|---|---|---|
| 3 | 9 | ~ 56% primary slots | Critical | Below threshold; one absence creates pages-or-burnout choice |
| 4 | 12 | ~ 42% primary slots | High | Marginally workable, no slack for absences |
| 5 | 15 | ~ 33% primary slots | Acceptable | Realistic minimum; absorbs single absence |
| 7 | 21 | ~ 24% primary slots | Low | Comfortable, absorbs multiple absences |
| 10 | 30 | ~ 17% primary slots | Very low | Generous; team values it visibly |
The Hidden Costs
Naive financial comparisons of follow-the-sun overstate the saving because they ignore three operational overheads that scale with rotation complexity. None are large individually; they compound into roughly 10 to 20 percent of the visible engineering headcount cost.
Handoff overhead is the largest of the three. A structured handoff at each shift boundary, three times per day, is realistically 15 to 30 minutes per shift across at least two engineers per handoff (outgoing primary plus incoming primary). At 20 minutes per handoff and three handoffs per day, this is 60 minutes of engineering time per day per region, across two engineers, every working day. Annualised that is approximately 130 engineer-hours per region per year of pure handoff, or roughly 1.5 to 3 FTE-weeks across the full rotation. At a fully-loaded $180,000 annual cost, that is $50,000 to $100,000 of annualised handoff overhead for a three-region team. Real, not invented.
Runbook and documentation duplication is the second hidden cost. Without conscious investment, regional teams develop slightly different practices: different escalation tolerances, different first-debug-step preferences, different criteria for declaring an incident. This drift is not catastrophic but it eats coordination time across handoffs and creates inconsistency in customer experience. A named runbook maintainer with allocated time (roughly 10 percent of one engineer's capacity, ongoing) is the realistic cost of preventing this drift. That is another roughly $18,000 per year in fully-loaded cost.
Multi-region compensation strategy is the third hidden cost, and the hardest to quantify. If your three regions are all major tech markets (San Francisco, London, Sydney), compensation parity is expensive but conceptually clean. If one region is a lower-cost market (typically Eastern Europe or India), the parity question becomes operationally fraught: pay-to-market and risk perceived inequity, or pay-to-headquarters and absorb the cost. Either choice has trade-offs, and the absence of a clear policy creates retention risk in the cheaper region. Budget for explicit handling.
Comparison to Single-Region With Night Pages
The financial alternative to follow-the-sun is a single-region rotation that takes night pages. The cost of night pages is typically modelled as a multiplier on the daytime page cost: the engineer doing the off-hours work is owed compensation (overtime, time off in lieu, or simply attrition risk), the cognitive cost of an interrupted night is higher than the cognitive cost of an interrupted afternoon, and the MTTA at night is consistently 20 to 50 percent slower than during the day across the public PagerDuty and incident.io datasets.
A useful comparison framing: take your current annual night-page count and multiply by a fully-loaded hourly cost at a 2x night premium (covering the wage premium plus the cognitive disruption plus the retention probability). For a 12-engineer team taking 8 night pages per engineer per month, that is 1,150 night pages per year. At a $180,000 fully-loaded cost (~$87 per hour) and an average 35 minutes of total disruption per night page (12 minute MTTA plus 23 minute Gloria Mark refocus), the direct cost is 1,150 * 0.58 hour * $87 * 2 (night premium) = $116,000 per year. Plus retention probability impact (incident.io 2024: 41 percent of on-call engineers report considering leaving because of alert load); at typical SHRM replacement costs that contingent liability easily exceeds $200,000 per year for a 12-engineer team.
The follow-the-sun overhead at the same 12-to-15-engineer scale is roughly $70,000 per year in handoff plus $18,000 in documentation maintenance plus the multi-region compensation premium (variable). If the multi-region premium is moderate (say, $50,000 per year across the team), the total follow-the-sun overhead is approximately $140,000 per year. Compared to the $116,000 in direct night-page cost plus $200,000-plus in contingent retention liability, the follow-the-sun path is financially favourable, materially so when retention risk is real.
When It Pays Off, When It Does Not
Follow-the-sun pays off when three conditions are met together. First, you have meaningful night-page volume: more than roughly 5 to 10 night pages per engineer per month, sustained. Second, you have at least 12 engineers in the on-call rotation: below that, regional staffing minimums break down. Third, you have the organisational capacity to actually staff in multiple regions: either existing offices, willingness to hire fully remote, or a clear plan to seed a new region.
Follow-the-sun does not pay off when night-page volume is low (under 5 per engineer per month is usually better fixed with alert tuning), when the team is small (under 12 engineers makes the regional minimums unachievable), or when the organisation cannot credibly staff in three regions. In any of these cases, the right move is to invest in alert hygiene first to reduce night-page volume below the threshold where the burden becomes unsustainable. The fix sequence is documented in /alert-tuning, /correlation-dedup, and /slo-vs-threshold.
For teams in the middle of these ranges, a useful intermediate step is a two-region split (US Eastern plus Europe, or US Pacific plus India) which covers roughly 16 of every 24 hours and leaves a shared evening that both regions handle as best-effort. This delivers most of the night-page reduction at a fraction of the multi-region overhead, and is more achievable at the 8-to-12 engineer scale.