rulebase-co After-hours coverage
Customers do not stop needing help at 17:00. The common failure is choosing a coverage model for cost and treating handoffs as free. Follow-the-sun, overnight shifts and deferred async each trade money, quality and speed differently — but all three fail the same way when the receiving shift inherits opaque tickets, missing context and wrong expectations. Design coverage and handoff together.
Choose the model deliberately
Three patterns, often combined by channel:
Follow-the-sun
Regional or global teams pass work across time zones — APAC → EMEA → Americas. Works when:
- You have qualified staff in each region on the languages and products you serve.
- Handoff windows are short and the ticket state is machine-readable.
- Product and policy are stable enough that regional interpretation does not diverge.
Fails when: only one region has depth, handoffs land in local morning peaks, or macros and policy differ by site without documentation.
Dedicated night / overnight shift
Same geography, off-hours hours. Works when:
- Real-time channels (voice, chat) need sub-hour response overnight.
- Volume justifies a roster; part-time overnight blocks are hard to recruit but cheaper than 24/7 follow-the-sun complexity.
- You can staff to measured overnight volume, not guess from daytime ratios.
Fails when: overnight is staffed for "presence" with no demand, or night team lacks escalation paths and ships everything to the day shift anyway.
Deferred async
No live coverage; queue until next business day with explicit promise. Works when:
- Channel SLA allows it (email, many ticket types).
- Expectations are set in product — status page, auto-reply, portal messaging — not implied by daytime responsiveness.
- Backlog at open is staffed as a surge, not absorbed into normal capacity invisibly.
Fails when: customers use async like chat, or marketing runs sends into a closed queue without surge planning.
| Factor | Follow-the-sun | Night shift | Deferred async |
|---|---|---|---|
| Real-time voice/chat | Possible with regional pools | Natural fit | Not suitable alone |
| Cost at low volume | High coordination cost | Fixed roster cost | Lowest |
| Handoff count | High — daily by design | Lower — one AM handoff | One open-the-queue spike |
| Language / product depth | Risk of shallow regional pools | Single-site depth | Day team only |
| Customer expectation | "Always someone" | "Someone overnight" | "We respond next day" |
Hybrid is normal: defer L1 async, staff overnight for P1 and enterprise, follow-the-sun for a global product line. Document which channel uses which model.
Handoff design
Handoffs are where coverage models lose quality. A handoff is a manufacturing step, not a status change.
Handoff packet — minimum contents
Every ticket crossing a shift or region boundary should carry:
| Field | Purpose |
|---|---|
| Customer goal in one sentence | Stops re-discovery |
| Actions taken and systems checked | Prevents duplicate work and customer re-telling |
| Open questions and blockers | Names what the receiver must resolve |
| Promises made to the customer | Prevents contradiction |
| Priority and SLA clock state | Receiver knows urgency |
| Sensitive flags — billing, legal, VIP | Routes attention before volume |
Free-text alone is insufficient; structured fields survive language and skill variance better than narrative essays.
Process rules
- Handoff window — last 30 minutes of shift is stabilise-and-document, not new complex work unless P1.
- No "pending customer" without documented ask — vague pending states die overnight.
- Escalation ownership follows the clock — name who is on call for product, billing and engineering overnight; "escalate to day" is not an overnight escalation path.
- Language match — do not hand Spanish tickets to a region with no Spanish coverage without explicit translation workflow.
Measuring handoff loss
Handoff quality is measurable without guessing:
| Signal | Likely handoff failure |
|---|---|
| Re-open or new message within one interval of shift start | Receiver restarted work |
| Repeat customer narrative in next message | Context not captured |
| Handle time spike on inherited tickets vs self-started | Re-work cost |
| First-response SLA miss on queue-open only | Opening surge understaffed |
| Escalation rate jump on handoff tags | Wrong skill or missing blocker note |
| CSAT drop on tickets touched by 3+ agents across shifts | Coordination tax |
Compare inherited vs self-started tickets within the same queue and skill — difference is handoff tax. Track by sending site and receiving site if follow-the-sun.
Opening the queue
The first hour of the day is a coverage event, not normal operations:
- Staff opening separately from steady-state requirement if backlog aged overnight.
- Triage rules — what gets auto-reply, auto-close, merge — published before peak season.
- Do not count overnight deferral as "within SLA" if the clock never stopped; be explicit which SLAs pause and which do not.
Traps
- Follow-the-sun without regional parity — three shallow teams instead of one deep one.
- Night shift handling only easy tickets — complexity lands at 09:00 anyway.
- Handoff tags without packet fields — "handed to APAC" is not information.
- Measuring only response time, not re-work — fast wrong answers cost more than slow right ones.
- Changing coverage model by channel without telling customers — expectation mismatch drives repeat contact.
- Ignoring daylight-saving and holiday calendars — handoff gaps appear twice a year on predictable dates.
Present results to the user
- Coverage model by channel — follow-the-sun, night, defer — with rationale and customer promise stated plainly.
- Roster implication — where volume was measured; gaps if overnight is assumed not staffed.
- Handoff packet spec — required fields and who completes them.
- Escalation and on-call map for off-hours — named paths, not "day team."
- Handoff loss metrics — which signals you will track and baseline if available.
- Opening-queue plan — surge staffing or triage rules for backlog at day start.
- Risks and unverified assumptions — regional skill depth, SLA clock rules, language coverage.