A 24-person platform group in one country pages its engineers at night. Leadership offers to fund a second site in another time zone so nobody is woken. What does follow-the-sun buy, what does it cost, and what would you check before agreeing?

What is gained, quantified

Nobody is woken. That is not a comfort argument: an engineer paged at 03:00 makes worse decisions, and the following day is largely lost. A team taking two night pages a week loses far more capacity than the two hours the pages consumed, and attrition among people who cannot plan an evening is the largest hidden cost in most platform groups.

A second site in a complementary time zone converts every page into a page during someone's working hours, which improves both the decision quality and the recovery time.

What is paid

Handover is now a system. Twice a day, context has to move between groups: what is degraded, what was changed, what is being watched. Handover quality decays quickly without structure, and an incident that spans a handover is the one that goes badly because the responders who understood it went home.

Ownership dilutes. The strongest incentive to fix a flaky service is being woken by it. When the person woken is on another continent, that loop is broken and the alert stays noisy for years. This is the failure mode of follow-the-sun that nobody predicts and everybody eventually has.

Knowledge splits. A service built in one site and operated at night by another is operated by people who cannot debug it beyond the runbook, so the second site escalates, and the first site is woken anyway - now with an extra hop and 20 minutes of delay.

It costs a team. Not headcount but a functioning group of at least five or six engineers with real ownership, because a rota of two people in the second site is worse than no second site.

What I would check first

1. The page volume and its causes. If the group takes 40 pages a week, the problem is the alerts, not the rota. Follow-the-sun applied to a noisy system exports the noise and removes the incentive to fix it. Get to a small number of actionable pages per shift first; the usual target is no more than one or two per 12-hour shift.
2. What fraction of night pages actually require action within the hour. Most do not. Reclassifying those to a ticket queue removes the night problem without an organisational change.
3. Whether the second site would own services, not just shifts. A site that owns products has the context to respond; a site that only takes pages is a call centre for an on-call rota and will not retain engineers.
4. The handover cost in engineer-hours. Two structured handovers a day across a 24-person group is a real recurring tax, and it must be smaller than the night-page cost it replaces.

When this is the wrong answer

When the page volume is the problem, which it usually is. The cheaper sequence is: delete alerts that do not correspond to user harm, make the remaining ones actionable, add auto-remediation for the recurring ones, and only then consider a second site. Most groups find night pages drop by most of their volume from alert hygiene alone, at a fraction of the cost of a second site.

Follow-the-sun earns its place when the service genuinely requires an engaged human at all hours - regulated availability commitments, real-time trading, live events with contractual uptime - and when both sites can own real work. Short of that, it buys sleep at the price of ownership, and ownership is what keeps the alert count low in the first place.