At 7:40 on a Wednesday morning a planner in Rotterdam opens the exception queue and finds 312 open items. Four people are working them. By lunchtime there will be more, because the vessels that left Asia three weeks ago are arriving inside the same ten-day window, and because the factories those orders came from are about to close for a fortnight. She starts at the top and works down, oldest first, because that is the only rule anyone ever agreed on. Somewhere around item 190 sits the order that stops a production line in Ohio on Friday.
Why the queue is about to get longer
China's Mid-Autumn Festival runs September 25 to 27 and National Day Golden Week runs October 1 to 7, with three working days between them. Weeks of export manufacturing are being compressed into a narrow shipping window, and then the origin goes quiet for most of a fortnight. The dates are published a year ahead, which is why it pays to plan around a factory shutdown at origin rather than meet the backlog when it arrives.
Carriers have responded the way they always do. Maersk and others have pulled transpacific capacity around the holiday, which is the same blank sailing pattern that rewrites an arrival plan without telling anyone. At the same time port congestion in China is absorbing more than 4 million TEU of containership capacity, with berth waits at Shanghai and Ningbo running as long as 12 days. Vessels held during the weather closures earlier in the month are now arriving in clusters rather than on a schedule.
The far end of the journey is no calmer. Rail-destined container dwell at Los Angeles and Long Beach reached 6.75 days in August, up from 6.34 in July and the highest in twelve months, after the Port of Los Angeles closed out its busiest three-month stretch on record.
None of that is new information to anyone running an operation. What is worth saying plainly is what it does to a team: it does not create one large problem, it creates several hundred small ones, all at once, for the same four people.
Working the list in order is a ranking, just not one you chose
Most teams believe they do not have a prioritization method. They do. It is just implicit, and it is usually one of three.
Oldest first, which ranks by how long an item has been ignored rather than by what it costs. Loudest first, which ranks by which customer emailed twice and which account manager sits nearest the ops desk. Biggest delay first, which ranks by the size of the number in the delay column.
All three feel fair. None of them is connected to money, and the third is actively misleading. A 12-day berth wait at Ningbo is a much bigger number than a two-hour missed drayage cutoff in Long Beach. The drayage miss is the more urgent item, because the decision window on it closes this afternoon and the decision window on the vessel closed a week ago. Size is not urgency.
This is a different problem from the one most visibility work solves. Detecting an exception early is necessary, and catching it before the customer feels it is the right goal. But detection that works produces volume, and volume is where the second failure lives. A team that sees everything and still works the list top to bottom has moved the bottleneck, not removed it.
Rank by the cost of waiting
A priority rule that survives a bad week ranks each exception by what it costs to leave it for another hour. Four inputs do most of the work.
Consequence. What actually happens downstream if nothing is done. A line stoppage, a retailer chargeback, an expired lot, a missed vessel, or nothing at all. This is the only input most teams try to capture, and they usually capture it as a severity flag set at detection, which is not the same thing.
Recoverability. Whether a decision still exists. An exception with no remaining option is reporting, not work. It needs a customer conversation, not a place near the top of the queue.
Decision window. How long the option stays open. Rerouting a container before it is discharged is cheap; after it is discharged and in a rail stack it is a different conversation, which is why tracking a container past the terminal gate matters more in a congested month.
Coupling. How many other commitments hang off this one shipment. One late pallet that holds an entire customer order, or a component that gates a build, outranks five independent late shipments with slack in front of them.
Score those four and the ranking stops being a matter of opinion.
The scoring needs data the exception list does not hold
Here is the practical obstacle. An exception raised by a carrier feed knows about the transport event and almost nothing else. It does not know the order value, the customer, the contractual delivery window, what else is on the same order, or whether there is inventory at the destination that makes the whole thing moot.
That information exists. It is in the ERP, the TMS, the WMS and the carrier portals, in four different shapes, which is why scoring by hand takes a person twenty minutes per exception and nobody does it when there are 312 of them.
Closing that gap is the specific job of a command center layer sitting above the systems of record. Orkestra normalizes order, shipment and inventory data across ERP, TMS and WMS, so an exception arrives already attached to the order behind it and already scored for impact rather than waiting for someone to look it up. Teams running exception management this way report around an 80% reduction in manual triage, which is mostly the twenty minutes per item disappearing.
Agree the tiers before the week you need them
Ranking under pressure does not work if the ranking is invented under pressure. Write the tiers down while the queue is short, in language an operator can apply in ten seconds. The tiers are one component of a wider disruption playbook: the pre-made decisions a team falls back on when a strike or closure fills the queue overnight.
Three tiers is usually enough. Tier one is anything that stops production, breaches a contractual window, or risks product integrity. Tier two is anything that costs money but not a commitment: demurrage accruing, a rebooking fee, an expedite premium. Tier three is everything that will resolve itself or cost only an apology.
Then settle the part teams skip: who can override the ranking, and what happens to tier three in a week like this one. The honest answer is often that tier three does not get worked at all, and saying so out loud is better than letting those items rot silently until a customer finds one. Set a rule, close them with a status, and let the team stop feeling behind.
Re-rank as the clock moves
Priority is not a property an exception is assigned once. It is a function of time, and it changes while the item sits in the queue.
A tier three delay with eight days of buffer becomes a tier one when the buffer is gone. A rerouting option that was cheap on Monday is gone by Thursday. If the queue only carries the score it was given at detection, it is showing last week's answer. The same is true across shifts: a ranking agreed by one regional desk only holds if the next one knows what it means, which is why handing exceptions off across time zones needs the tiers written down before the night starts. The ranking has to re-run as vessel positions, gate times and inventory positions change, which again is not something a person does manually across several hundred items.
The payoff for getting it right is measurable in time rather than tonnage. OIA Global saw a 60% reduction in exception resolution time in Q1 after moving to this model, and resolution time is the number to watch, not the number of alerts raised. If you are choosing which metrics to report end to end, resolution time and the share of exceptions closed before customer impact tell you more than queue depth ever will.
What to do before the vessels arrive
Pull last month's exception log and sort it by what each item actually cost, not by how it was flagged. Most teams find the expensive items were not the ones that got worked first. Write the three tiers down this week, agree who overrides them, and decide now what happens to the bottom of the list when volume arrives faster than the team can absorb it. Tiering works on exceptions that have already fired; the same consequence logic applied earlier, to the lanes themselves, gives you a supply chain risk score that says where to look before anything is late.
The same tiers do the work in the other direction too, when a dated trade rule lands and you have to scope which shipments it touches before the cutoff rather than work the queue in the order it arrived.
Congestion is also bunching vessel arrivals, so the same tiers decide which boxes move first when a week of containers lands on one day.
The wave from this holiday window is already loaded. It will land in a fortnight either way. The only variable you still control is the order in which your team works it, and whether that order came from a rule you chose or from whoever shouted first.
