A carrier posts a schedule change at 02:40 on a Tuesday in Chicago. The vessel will omit Ningbo, and your containers roll to the next sailing. In Ningbo it is 15:40, the origin team is at their desks, and the rebooking window is still open. Nobody acts, because the booking sits with the North American desk and the North American desk is asleep. By 08:00 Chicago time the window has closed and the boxes have rolled. The exception did not get worse overnight. Your coverage did.
Why the dead hours cost more this quarter
Asian port congestion has deepened through the autumn, with roughly 4.3 million TEU tied up in congestion worldwide and schedule reliability falling with it. Golden Week ran October 1 to 7, factories and export desks paused, and carriers announced blank sailings around the holiday. Those schedule changes are posting now, and they will keep landing as revised ETAs at destination for weeks. Most of them originate in Asian business hours, which means they arrive on North American and European desks in the middle of the night.
The volume is not the only problem. A rolled container is recoverable at 03:00 and expensive at 09:00, because the rebooking, the drayage slot, and the customs filing all have their own cutoffs. The same is true of a blank sailing or port omission caught early enough to move cargo to a different string. Exceptions have a clock, and the clock does not pause for your working day.
Being awake is not the same as being accountable
Most global teams believe they already have round-the-clock coverage, because somebody in the network is always online. Then you map it properly and find a window, usually six to ten hours wide, where an exception can be raised, seen by two or three people, and acted on by nobody. It is not that the hours are unstaffed. It is that they are unowned.
Detection is the easier half to close. A command center that sits above your ERP, TMS, and carrier feeds raises the exception the moment the carrier message lands, rather than when someone opens a portal at the start of their shift. That is the gap behind the industry figure that 62% of exceptions are found only after the customer feels them, and it is the gap Orkestra's exception management is built to close: the event is detected, scored for impact, and queued against an owner while the originating desk is still awake to do something about it.
Detection without a named owner just means the alert waits in better lighting. The harder half is the handoff.
A handover note is not a handoff
Shift handover notes usually describe what happened. A handoff transfers accountability. Every exception crossing a regional boundary should carry four things, and if any one is missing the receiving desk will stall:
Current state. Not the original problem, the live one. What changed since it was raised, and what the shipment is doing right now.
The clock it is racing. The cutoff that actually governs the decision: the next sailing, the gate close, the filing deadline, the production slot. "Urgent" is not a deadline.
What has already been tried. Which carrier was called, what they said, which alternative was priced and rejected. Without this, the next desk repeats the first two hours of work.
What the receiving desk may do without asking. This is the one that gets left out, and it is the one that decides whether the night is productive or simply observed.
Decide the authority before you need it
If the overnight desk has no standing authority, its only safe action is to escalate, and escalation at 03:00 means wait. Write the limits down while nobody is under pressure: rebook on an alternative string up to an agreed cost delta, approve a premium move below a set threshold, split a shipment, accept a later ETA on replenishment stock but never on a production-critical part.
Those limits should mirror the tiers you already use to rank exceptions when several break at once. A tier that means "work this first" during the day has to mean "you may act on this alone" at night, or the ranking stops at the border of the shift.
Then make the limits run by themselves. Workflow automation is where this becomes real: the rule that a roll on a production-critical purchase order pages the on-call planner at any hour, while the same roll on safety stock waits for the morning, is written once and applied every night without anyone remembering to apply it. The point is not that software decides. The point is that the decision was made in advance, and the overnight desk inherits it instead of guessing.
Route to a person, not a distribution list
Broadcasting an exception to a regional mailbox feels like coverage and behaves like diffusion. Groups do not own things. Name an on-call person per region per shift, put the name on the exception, and give it a response clock the way you would an incident. Teams that do this consistently see the result quickly: OIA Global reported a 60% reduction in exception resolution time in Q1 after consolidating this kind of work into one operational view.
The handoff also needs somewhere to live. When the conversation happens in email and WhatsApp while the shipment lives in the TMS, the next desk starts by reconstructing the story from three places. Tie every message, document, and decision to the shipment or case itself, so a planner in Rotterdam picking up a container from a Singapore colleague reads the whole history in thirty seconds and starts from the real current state. That is the difference between catching an exception before customer impact and documenting it afterwards.
Measure the hours, not the headcount
You cannot fix a window you have not measured. Track time to first touch on exceptions, bucketed by the hour they were raised in origin local time. Most teams who run this for a month find a clean cliff: a median of twenty minutes during the covered day and five or six hours for anything raised between 22:00 and 06:00.
Be careful comparing regions on that number. A desk that raises exceptions manually looks faster than one where the system raises them automatically, for the same reason one region can look cheaper simply because it reports less. The rules for making regional numbers comparable are the same ones that apply when you benchmark performance across regions: agree the definition and the clock before anyone is ranked.
Where to start this week
Three things, none of which require a project. Map your actual dead window by plotting when exceptions are raised against when each desk is staffed. Write the standing authority limits for the overnight desk and get them approved by whoever currently gets woken up. Instrument time to first touch by hour so you can prove the window closed.
The backlog from Golden Week will arrive as a stream of revised ETAs, rolled bookings, and gate changes over the coming weeks, most of them posted while one of your desks is dark. Teams that handle it well will not be the ones with more people. They will be the ones where every exception had an owner, a clock, and a pre-agreed limit on what could be done without a phone call. That is what Orkestra's execution layer is for: one view of every order and shipment across your systems and carriers, exceptions routed to a named owner with the context attached, and the response running on rules your team wrote in daylight.
Book a demo to see how your exceptions would be routed overnight.
