Systems and records
How to build a support handoff continuity dataset across time zones
By SourceX Editorial · Updated
10 sections
Short answer
A support handoff continuity dataset should store every shift-to-shift transition as its own event, with the context packet, an owner timeline, open commitments and the clock conditions, then grade it against the resulting state. In a 2026 replay of an illustrative Manila, Dublin and Chicago desk, 33.3% of 783 handoffs left a gap with nobody on shift, and the live-handoff share doubled from January to July.
System export
Support handoff continuity
Key takeaways
- Record each handoff as an event with its mode: live, edge to edge or gapped, plus minutes of overlap or gap and the tz release used.
- In the illustrative three-hub replay, 411 of 783 handoffs in 2026 were live, 111 met edge to edge and 261 left a gap.
- The live share is 33.3% in winter months and 66.7% from April to September, so a July sample and a January sample describe different conditions.
- One ticket's four-business-hour deadline moved 14 hours depending on which calendar the SLA used.
- Five failure codes, from dropped request to stale state, catch problems a closed status or a thank-you never shows.
- Hold out by hub pair and calendar regime, and use the 20 weekdays of US-only summer time as a stress set.
What should a support handoff continuity dataset prove?
It should prove that every time a case crossed from one shift to the next, three things survived: what the receiving agent needed to know, a named owner, and each commitment already made to the customer, and that the case then reached a verified resulting state. A support handoff continuity dataset stores those crossings as events in their own right, together with the clock conditions under which each one happened, and grades them against what the records show afterward.
The clock conditions are what set desks spread across time zones apart from a single-site queue. When sender and receiver share an hour, a missing detail can be asked for out loud. When they never overlap, the written note is the only channel, and a gap of two or three hours, or a whole weekend, opens between one owner leaving and the next arriving. The calendar sets those conditions, not the agents, so a dataset that keeps only ticket text and a final status cannot tell a careless handoff from one that never had a chance.
This guide measures the calendar effect with a reproducible replay of one illustrative three-hub desk across 2026, then sets out the record each transition needs, an annotated decision packet, five failure codes and a held-out test that stops calendar effects from passing as skill. General transfer completeness and cross-border timestamp normalization are covered in other SourceX guides and are not repeated here.
A year of handoffs at a three-hub desk, replayed from the tz rules
In the replay, 411 of 783 handoffs in 2026 (52.5%) had live overlap, 111 (14.2%) met edge to edge with neither overlap nor gap, and 261 (33.3%) left a gap in which nobody was on shift. Nobody chose those proportions. They follow from three hubs keeping local office hours under clock rules set by three different governments.
The footprint is illustrative rather than a real company: Manila, Dublin and Chicago, each staffed Monday to Friday from 09:00 to 17:00 local time, with no holidays and no weekend cover, passing open cases west in that order. Release 2026e of the IANA time zone database supplies the clock rules. The United States starts summer time on the second Sunday in March and ends it on the first Sunday in November; the European Union changes on the last Sundays of March and October at 01:00 UTC; the Philippine rules show no clock change since 1990. In 2026 that puts the American changes on March 8 and November 1 and the European ones on March 29 and October 25.
Each hub pair behaves differently. Manila to Dublin meets edge to edge in winter, when 17:00 in Manila is 09:00 in Dublin, and gains a one-hour overlap only while Ireland is on summer time, from March 30 to October 23. Dublin to Chicago always overlaps, by two hours for most of the year and by three on the 20 weekdays when only the United States has changed its clocks, March 9 to 27 and October 26 to 30. Chicago to Manila never overlaps: on weeknights the case waits two hours in winter and three while Chicago is on daylight time, and from Friday evening it waits 50 or 51 hours until Monday morning in Manila.
Method, as the script applies it: Python 3.13's zoneinfo module reads the 2026e rules from the tzdata 2026.5 package rather than the system's zone files. Each local weekday of each hub is one shift. A handoff runs from a shift to the next hub's first shift that starts at or after the sending shift began; overlap is the time both shifts share, gap is the time between the sender's end and the receiver's start, and a handoff is live when the overlap is at least one minute. Counts cover shifts whose local date falls in 2026. Nothing random is involved, and the script (handoff_windows.py) is kept with the review files.
The table gives the window for each hub pair in each part of the year.
| Handoff | Winter weeks (to Mar 6, from Nov 2) | US-only summer time (Mar 9-27, Oct 26-30) | Both on summer time (Mar 30-Oct 23) | Friday evening |
|---|---|---|---|---|
| Manila to Dublin | Edge to edge, no overlap or gap | Edge to edge, no overlap or gap | 1 hour of overlap | Same day, no weekend crossing |
| Dublin to Chicago | 2 hours of overlap | 3 hours of overlap | 2 hours of overlap | Same day, no weekend crossing |
| Chicago to Manila | 2-hour gap | 3-hour gap | 3-hour gap | 50-hour gap on 18 Fridays, 51 hours on 34 |
| Live handoffs per week | 5 of 15 | 5 of 15 | 10 of 15 | Fridays follow the same split |
Why the live-handoff share doubles between January and July
Because Manila and Dublin overlap only during European summer time, the share of handoffs with live overlap is 33.3% in January, February, November and December and 66.7% in every month from April to September, with March at 36.4% and October at 59.1%. A sample drawn from July contains twice the live handoffs of one drawn from January, with no change in staffing, tooling or training.
That shift reaches any label built on handoff quality. In a live handoff, part of the context can pass in conversation and never reach the ticket; in an edge-to-edge or gapped one, the note is all the receiver gets. A model trained or graded on a summer-heavy sample therefore meets a different mix of handoff conditions than one built from winter months, and a note-quality score that ignores the condition will drift with the season while the team stays the same.
The fix is cheap at collection time and expensive later. Store the handoff mode (live, edge to edge or gapped) and the overlap or gap in minutes on every transition, computed from both agents' schedules and the zone rules in force that day, and split every reported metric by mode. Rebuilding the mode afterwards needs historical rosters; if those sit in a workforce management tool rather than the helpdesk, ask for them in the same request.
Same ticket, four deadlines: which calendar does the SLA label use?
An SLA breach label is only as reliable as the calendar behind it. In the replay, a ticket opened at 16:00 Chicago time on Friday, November 6, 2026 with a four-business-hour first-reply target falls due at 22:00 on Sunday, Chicago time, on a follow-the-sun calendar that counts any staffed hub; at 23:00 on Sunday on Manila's calendar; at 07:00 on Monday on Dublin's; and at 12:00 on Monday on Chicago's own. The four deadlines land 54 to 68 hours after the ticket opened, 14 hours apart.
Helpdesks make that choice in configuration, and an export can easily drop it. Zendesk pauses business-hour SLA targets outside business hours and restarts them when the schedule opens; on accounts with several schedules, each schedule can carry its own time zone, triggers decide which schedule a ticket receives, and the first schedule in the list is the default for every ticket. Jira Service Management asks for a time zone on each SLA calendar and lets each SLA goal pick a calendar, while Atlassian notes that its Resolution Time gadget ignores calendars and counts elapsed time around the clock. ServiceNow's support knowledge base describes SLAs set to the caller's time zone that ran on the system default instead, because the SLA record was created before the caller field was filled in.
For a continuity dataset, keep the calendar identifier and its time zone with every SLA event, the zone rules release used to compute it, and both elapsed and business-hour durations. Without them, an exported breach cannot be told apart from a configuration artifact, and a case that crossed a weekend looks like a failure by whoever held it last.
The four due times for the same ticket:
| Calendar behind the 4-hour target | Due, in Chicago time | Hours after the ticket opened |
|---|---|---|
| Any staffed hub (follow-the-sun) | Sunday 22:00 | 54 |
| Manila 09:00-17:00 | Sunday 23:00 | 55 |
| Dublin 09:00-17:00 | Monday 07:00 | 63 |
| Chicago 09:00-17:00 | Monday 12:00 | 68 |
Who owns the case while its next owner is offline?
Someone must, and the record should say who for every minute. In the replay, each Chicago-to-Manila handoff opens a two- or three-hour window on weeknights, and a 50- or 51-hour window over the weekend, in which the assignee of record has gone home and the next one has not yet started. If the ticket still names the departed agent, a customer reply in that window lands with someone who cannot act on it; if it names a queue, it lands with no one in particular.
Research on follow-the-sun work has treated the handoff as the hard part for a long time. Writing in the Journal of Management Information Systems in 2010 about software development, Carmel, Espinosa and Dubinsky note that follow-the-sun has few documented industry successes because it is extremely difficult to implement, and conclude that handoff efficiency is paramount. A support desk inherits the same problem with a customer waiting on the other side.
So ownership belongs in the dataset as a timeline, not a single field: assignee changes with timestamps, the receiving agent's first acknowledgment, and every interval in which the owner of record was off shift. This guide calls those intervals orphan windows. They are where commitments slip, and they can only be measured when the roster is joined to the assignment history.
The handoff record: what each transition carries
Each transition becomes one row that points back to its case, carrying enough to grade the handoff without rereading the whole ticket. Free text stays in the ticket and is referenced by pointer rather than copied.
The fields, and why the grade depends on each:
| Field | What it holds | Why the grade needs it |
|---|---|---|
| Transition and case ids | One id per shift change, linked to the case | A case with three handoffs is graded handoff by handoff |
| Sender, receiver and hub pair | Pseudonymous agent keys and their hubs | Separates hub-pair effects from individual skill |
| Handoff mode and window | Live, edge to edge or gapped, with minutes of overlap or gap and the tz release used | Takes the calendar out of note-quality scores |
| Context packet | The note plus pointers to each fact it asserts: steps tried, evidence, policy version | Lets a grader check every claim against the record |
| Open commitments | Each promise with its deadline in UTC and in the customer's local time | Catches deadlines missed in a zone conversion |
| Owner timeline | Assignee intervals and the receiver's first acknowledgment | Measures orphan windows |
| SLA clock state | Calendar id, its time zone, elapsed and business-hour durations | Separates real breaches from calendar artifacts |
| Resulting state | What the downstream system shows after the handoff, and any return contact within the window | Grades the outcome rather than the closing message |
Annotated decision packet: a payroll export crosses three hubs (illustrative)
The packet is invented to show the annotations a grader adds; it does not describe a real company or customer. It runs on Thursday, October 15, 2026, when both regions are on summer time: Chicago is UTC-5, Dublin UTC+1 and Manila UTC+8, so the Chicago-to-Manila gap is three hours and Manila and Dublin share one.
Each entry gives the event, then the grader's annotation:
- Thursday 15:40 Chicago (20:40 UTC): the customer reports that a scheduled payroll export failed and must run before their payroll cutoff at 12:00 Chicago time on Friday. They also ask that the partial file from the failed run be deleted. Annotation: two open requests, with a hard deadline at 17:00 UTC Friday.
- Thursday 16:50 Chicago (21:50 UTC): the Chicago agent finds an expired authorization, sends a re-authorization link and writes: "We will rerun as soon as you re-authorize and update you by 10:00 your time tomorrow." The handoff note lists the cause, the rerun step and the deletion request. Annotation: commitment recorded as 15:00 UTC Friday; both requests are in the packet.
- 22:00 to 01:00 UTC: nobody is on shift and the case still names the Chicago agent. The customer re-authorizes at 23:30 UTC. Annotation: 180-minute orphan window; the customer's action waits 105 minutes for an owner.
- Friday 09:15 Manila (01:15 UTC): the Manila agent takes the case, reruns the export, sees the job succeed at 01:40 UTC and updates the customer at 01:45 UTC without asking again for anything already captured. Annotation: context retained, no re-ask, commitment met more than 13 hours early.
- Friday 16:00 Manila (08:00 UTC): the note to Dublin reads: "Export rerun OK, waiting for the customer to confirm." Annotation: the deletion request is missing from the packet, and Manila and Dublin are inside their one live hour, so the gap could have been closed in conversation.
- Friday 09:00 Chicago (14:00 UTC): the customer confirms payroll ran and thanks the team; Dublin closes the case at 14:10 UTC. On Monday the customer writes back asking whether the partial file was deleted. Annotation: dropped request at the Manila-to-Dublin handoff; outcome partial despite a closed status and a thank-you.
Five failure codes for grading a handoff
Grade each transition against the record with one pass condition and five failure codes; a single transition can carry more than one code. Because the codes name what went wrong at the seam, they can be counted by hub pair and by handoff mode, which is where the replay shows the conditions differ.
Four of the five leave no trace in a status field or a thank-you, which is the case for building the dataset at all. The last column says whether a closed status or a thank-you would reveal each one.
| Code | What happened at the handoff | Evidence that settles it | Visible from closure or thanks? |
|---|---|---|---|
| Dropped request | A customer request open in the sender's record is absent from the receiver's packet and never actioned | Requests in the conversation matched against the packet and later actions | No |
| Orphan window | The owner of record was off shift when the customer acted or a deadline fell | Assignee intervals joined to the roster and message times | No |
| Clock-shifted commitment | A deadline was stated or tracked in the wrong zone, or a zone rule changed in between | Commitment text, customer zone, UTC deadline and tz release | No |
| Re-ask | The receiver asked the customer for information already in the record | Outbound questions matched against facts already captured | Rarely |
| Stale state | The packet described a state that had already changed, such as a pending step already complete | Packet claims compared with system events before the handoff time | No |
| Pass | None of the above, the resulting state is verified and no return contact arrives within the window | Downstream system after the handoff plus a return-contact check | Only partly |
A solved status after a gap: resolution or reset?
Treat it as unproven until the system of record agrees. On a follow-the-sun desk the agent who closes a case can be two or three shifts removed from the one who opened it, and a customer's thanks usually arrives before anyone can know whether the fix will hold. In the illustrative packet, the case closed with a thank-you and still produced a return contact on Monday about a request that fell out of a note.
Time zones add an artifact of their own. If an inactivity rule closes tickets after a quiet period measured in elapsed time, a customer who replies only during their own working day can be closed out by a desk staffed on the opposite hours, and the export will show a resolution that never happened.
The durable test is one the records can run: the verified state in the downstream system, such as the export job, the refund or the account change, plus no return contact on the same issue within a fixed window counted in calendar days from the last handoff rather than from closure.
Held-out tests by hub pair, and the request that makes them possible
Hold out by hub pair and by calendar regime, not by random tickets. A rubric or model tuned on Dublin-to-Chicago handoffs, which always overlap in the replay, has not been tested on Chicago-to-Manila handoffs, which never do. The 20 weekdays of US-only summer time make a natural stress set, because overlaps and gaps move by an hour while staffing stays fixed. Freeze the rubric before scoring the held-out cases and report results separately for live, edge-to-edge and gapped transitions.
When requesting records, name the handoff fields above, the rosters needed to compute modes and orphan windows, the SLA calendar definitions in force over the period and the time zone database release the supplier used. Rosters and assignment histories describe employees as well as customers, so both sides need pseudonymous keys before anything leaves the company.
SourceX does not keep datasets in stock: it sources them on request from U.S. businesses, and its request form collects scope, volume, language, format, timing and licensing needs. A handoff requirement fits there with the acceptance test written out, and whether a particular company's records can meet it is decided case by case after rights and privacy review.
Frequently asked questions
What is a follow-the-sun support handoff?
It is the point where an open case passes from a hub ending its working day to a hub starting its day in another time zone, so work continues around the clock. Depending on office hours and clock rules, the two shifts may overlap, meet edge to edge or leave a gap in which nobody is on duty.
Why does daylight saving time matter for a support dataset?
The United States and the European Union change clocks on different dates, so overlaps and gaps between hubs move by an hour for several weeks a year. In the illustrative replay, Dublin and Chicago overlap for three hours instead of two on 20 weekdays in 2026, and Manila and Dublin overlap only during European summer time. Labels that depend on timing need the zone rules release recorded.
Should a handoff dataset store timestamps in UTC or local time?
Both. Store the UTC instant for ordering events, the IANA zone name for each agent and the customer, and the time zone database release used for conversions. Commitments such as an update by 10:00 your time can only be graded in the customer's local time.
Can ticket closure serve as the outcome label for handoffs?
Not on its own. The agent who closes a case may be several shifts removed from the one who opened it, and inactivity rules can close tickets while a customer in another zone is asleep. Use the verified state in the downstream system plus a return-contact window counted from the last handoff.
Which records outside the helpdesk does a handoff dataset need?
Agent rosters or schedules to compute handoff modes and orphan windows, the SLA calendar definitions in force at the time, and events from the downstream system that confirm the resulting state, such as job logs, refunds or account changes.
Sources
- IANA time zone database release 2026e: US, EU and Philippine clock rules used in the replay; the replay counts come from handoff_windows.py run on these rules. Source
- Zendesk SLA targets run in calendar or business hours, and business-hour targets pause outside business hours. Source
- Zendesk schedules: per-schedule time zones on multi-schedule accounts, triggers that apply schedules and a default first schedule. Source
- Jira Service Management SLA calendars carry a time zone and each SLA goal selects a calendar. Source
- Atlassian: the Resolution Time gadget ignores SLA calendars and counts total elapsed time. Source
- ServiceNow known issue: a caller time zone SLA can fall back to the system time zone when the SLA starts before the caller is set. Source
- Carmel, Espinosa and Dubinsky (2010) on follow-the-sun workflow: few documented successes and handoff efficiency as the key factor. Source
Tell us what data you need
Share scope, volume, format, timing and licensing needs. Datasets are sourced on request.