WiseTech: Asia-North America on-time fell to 26.9% in Sept, Asia-Europe 9.9%

The Analysis

WiseTech's October outlook opens with a number that should stop anyone who books containers in the middle of reading it: Asia-North America vessel on-time performance came in at 26.9% in September. August was 33.9%. The most recent week printed 26.8%, which means September was not a bad month with a good week tacked on the end. It was a bad month that is still bad.

Asia-Europe is the ugly one. On-time fell from 22.9% to 9.9%, and in the same measured week 45.5% of volume on that trade arrived more than eight days behind schedule. Nine point nine percent means nine ships in a hundred showed up when the schedule said they would. Forty-five point five percent means close to half the cargo moved a week late or worse. Those are two different statements, and only one of them tells you what to do.

Published transit times are 26.2 to 27.6 days on the transpacific and 37.8 to 39.9 days on Asia-Europe, and both read better than what actually happens at the berth. Europe to North America, the lane nobody writes headlines about, sits at 39.2%, while the spread between individual carriers on that lane runs from roughly 14% to 61%. WiseTech tells forwarders to book earlier and build in buffer. The advice is fine. We take the numbers apart first.

Start with the caliber, because it decides everything downstream. On-time performance in datasets like this one is scored against the carrier's own proforma schedule, not against the arrival date you promised your customer. Most providers count a vessel as on time if it arrives the same day, sometimes within a day either side. That makes 26.9% a statement about how many ships crossed a line drawn at zero, not a statement about how late the late ones were. A ship six hours behind and a ship eleven days behind both count as late, and they count exactly the same. The measurement is sound. As a planning input on its own it is nearly useless, because what your warehouse needs is the shape of the distribution, not the share of it sitting on the right side of zero.

Now the thing the headline does not say, and the thing that should change what you do on Monday. Ask what is in the denominator. A sailing that was cancelled is not a late sailing; it leaves the sample. Blank sailings are the standard carrier response when a network is slipping, and they do two things at once: they take away the ships most likely to arrive late, and they take them out of the calculation before anyone measures. So 26.9% means that of the ships which actually sailed, 26.9% arrived on time. It does not count the ship you got rolled off. If capacity was pulled into September, the same decision that cost you space also improved the number you are being shown.

The figure that cannot be cleaned that way is 45.5%. That one is measured on cargo that did move. Nearly half of Asia-Europe volume in the measured week was more than eight days late, and no amount of blanking makes it disappear. The release led with 26.9% and 9.9% because those are the numbers that travel. The number that should set your buffer is 45.5%, and it is the one buried halfway down.

We can do more with it than the release did. If 45.5% of volume sits beyond eight days late, the median delay lands just under eight days, call it seven to eight. Half your Asia-Europe cargo is arriving roughly a week behind plan, and the tail goes further. Most importers I have sat across from built their safety stock in years when a three-to-five-day wobble was the honest range. If that is your position, you are not marginally under-covered. You are short by most of a week, and you will discover it one purchase order at a time, which is the most expensive way to discover anything.

The proforma numbers make it worse rather than better. Twenty-six point two to twenty-seven point six days transpacific, thirty-seven point eight to thirty-nine point nine days Asia-Europe. These were published for the sale, and they are prettier than actual arrivals. That gap is not noise you can average away; it is structural. Here is the part that should bother you more than the on-time rate: the proforma is what your ERP anchors on. Your reorder point, your production slot, your customer's committed date were all almost certainly calculated from a number the carrier never intended to hit. Put a median seven-to-eight-day slip on top of a 39-day Asia-Europe proforma and your real planning figure is 46 to 47 days before a single other thing goes wrong.

One caution before you compare this number with anything else. Reliability data is not standardised across providers. Some score arrival on the proforma day, some allow a day of grace, some measure at pilot station rather than at berth, and a couple measure at container availability rather than vessel arrival. Twenty-six point nine percent from WiseTech is therefore not directly comparable with a reliability figure from another source, and anyone building a carrier scorecard by pasting two providers together is measuring the definitions, not the carriers. What a single provider's series is good for is direction and speed of change: August 33.9% to September 26.9% to a 26.8% week is a real move measured on a consistent caliber, and that is the part you should act on.

The transpacific deserves its own look, because the range is doing some talking. Twenty-six point two to twenty-seven point six days is a spread of a day and a half across the loop, and the latest week at 26.8% tells you there was no late-month recovery. When a trade prints bad in August, worse in September and no better in the first week you can measure, you are not looking at a shock. You are looking at a level. Treat 26.9% as the current state of the lane, not as an outlier month that reverts on its own, and plan November on that basis rather than on hope.

There is a second-order consequence worth pricing while you are at it: when four ships in three are late, the ones that are late are late by different amounts, so your arrivals stop coming in the sequence you shipped them. A purchase order released on the first of the month can land after one released a week later. If your warehouse or your production line consumes in shipping order rather than in arrival order, that inversion costs you more than the average delay does, and no buffer fixes it.

Why now? WiseTech publishes measurement, not causation, and I am not going to invent a typhoon to make the paragraph more satisfying. What the data does show is the shape. Both trades fell month over month at the same time, and one week in Asia-Europe put nearly half the volume a week or more behind. Two trades moving together in the same direction points at something shared, whether that is capacity coming out, a hub under stress, or rotation changes rippling through a network, rather than one port's weather. Asia-Europe dropping from 22.9% to 9.9% in a single month is a violent move for a lane with a 39-day transit, which means whatever caused it was already operating on the network weeks before September arrivals were scored. That is the useful part of the timing: the deterioration you are reading about today is already inside your October arrivals.

Layering matters because these are three different problems wearing one headline. Asia-Europe importers are in the worst position: 9.9% on time, 45.5% of volume beyond eight days, and a 38-to-40-day transit to absorb all of it. Transpacific at 26.9% is genuinely bad and roughly half as bad, three ships in four late but on a shorter run with a shallower tail. Europe to North America at 39.2% is the best of the three and gets no coverage at all, because it is not the lane people write about. If you have any discretion over origin or routing, the reliability ranking in this dataset is Europe-origin first, Asia-to-North-America second, Asia-to-Europe last, and the distance between first and last is about thirty percentage points. That is not a rounding difference. That is a sourcing argument.

The most actionable number in the release is not any of those. It is the carrier spread: roughly 14% to 61%. Same lane, same month, a fourfold difference in service between the worst and best operator measured. If you are buying ocean freight on rate alone you are not buying a cheaper service, you are buying an unknown one, and you find out which end you got somewhere around day thirty. Assume you ship forty FEU a month on the lane; that is the assumption, put your own number in. At the 14% end, about thirty-four of those forty boxes arrive late. At the 61% end, about sixteen do. The difference is eighteen boxes a month, and it does not come from your forwarder's tracking, your broker, or your trucker.

It comes from which carrier's paper you signed. Nothing else in this story moves eighteen boxes a month. Put a price on the premium instead: ask what the reliable carrier wants per FEU over the cheap one, multiply by forty, and set it against the eighteen boxes. If the premium is two hundred dollars a box, you are paying eight thousand a month to avoid eighteen late boxes, and the answer depends entirely on what a late box costs you, which is the number almost nobody has written down anywhere.

Timing is where the arithmetic gets uncomfortable. Take Asia-Europe: 37.8 to 39.9 days proforma, plus a median seven-to-eight-day slip, plus inland and customs. If your planning number is forty days door to door, reality at the median is forty-seven or forty-eight, and the share of volume beyond eight days pushes a meaningful slice past fifty-five. Your reorder point is therefore wrong by ten to fifteen days, not by two or three. When a carrier or a forwarder tells you to book earlier, you should, but booking earlier against a broken planning number just means being wrong earlier. Fix the number, then move the order date, in that order.

Let us price the delay honestly, because most people overpay for the wrong half of it. Assume forty FEU a month to North Europe, assume sixty thousand dollars of goods per FEU, assume eight percent annual carrying cost. One day of carrying works out to roughly thirteen dollars a box. Eight days is about a hundred and five. Across the roughly eighteen boxes a month running eight-plus days late, that is about nineteen hundred dollars a month. Nineteen hundred dollars. I have watched companies pay a freight premium larger than that to avoid a problem that cost them less than two thousand.

So the money is not in carrying cost. It is in the stockout, the line stoppage, the missed retail window, the demurrage and detention clock that starts when the box lands, and the air freight somebody expends to rescue one purchase order. That gives you an arithmetic rule instead of an instinct: pay a premium for reliability only when the premium is smaller than your expected cost of being late. Most of the time it is not, which means the fix is buffer and carrier selection, not a higher rate.

Sizing the buffer is the obvious next question, and the honest answer is that two data points do not give you a full distribution. You know where the median is, roughly seven to eight days, and you know one point on the tail, 45.5% beyond eight days. You do not know where the ninetieth percentile sits. So make the assumption explicit and write it down: if the tail beyond the median runs about as long again, which is the common shape when a network loses a hub window rather than losing speed, covering nine shipments in ten takes roughly fifteen days of buffer rather than seven. That is the number I would carry, and I would revisit it the moment a new weekly figure moves 45.5%. If it falls below thirty percent, the buffer comes back down with it, because carrying twice the stock forever is its own kind of loss.

One more angle, and it belongs to somebody else. Reading a corridor node by node is Ömer Kaya's habit; he wants to know which specific point ate the week before he will discuss an average, and that is exactly the right question here. Ask your carrier or your forwarder where the eight days happened: load port, transshipment hub, or waiting on a berth at destination. The answer changes the fix completely. Congestion at origin means book earlier. A hub problem means change the rotation rather than the buffer. A berth problem at destination is the one you cannot fix at all, and it is the one that arrives on your invoice as demurrage.

What you ask for this week follows from that. Get the proforma arrival for each individual sailing in writing, not a lane average, because about thirty-eight days is not a term anybody can claim against. Get each carrier's own on-time figure for your lane rather than the trade figure; you are buying one carrier, not a market. And ask what happens when you are rolled: is there a written undertaking on the next available sailing, or is it goodwill? While you are at it, renegotiate the two clocks that turn a late vessel into an invoice. Free time at destination is quoted for a network that arrives on schedule; at 45.5% of volume running eight-plus days late, the boxes land in clusters and the free time you were granted on a normal week does not cover the week your whole month arrives together.

Ask for extended free time now, in writing, per shipment rather than per contract year, and ask what the per diem is after it. The same argument applies to your inland appointment windows: if the vessel slips seven days, the trucking slot you booked is gone, and rebooking it in a tight market costs real money. Rolling is where schedule failure actually turns into cost, and it is the clause almost nobody negotiates while rates are cheap.

Then ask what would make me wrong, because there is a version of this that reverses. If carriers respond by blanking more sailings, on-time performance will improve, partly because the network gets slack again and partly because cancelled sailings leave the denominator while late ones stay in it. In that world the headline number goes up, your space gets tighter, and the rate goes with it. You would be reading better reliability data while having a worse quarter. So do not use on-time alone as a trigger. Watch three things together: on-time, blank sailing count, and rate. When on-time improves while blanks and rate rise, you are not getting a better service, you are getting a smaller one.

The other condition that flips the conclusion is what is inside the box. Low value, high volume, no fixed delivery date: take the cheap unreliable carrier and carry more stock, and stop paying for punctuality you do not need. Promotion-dated cargo, project cargo, anything with a contractual delivery window: pay for the reliable carrier and stop pretending the buffer will save you. Same data, two opposite decisions, and the only variable is what being late costs you.

My own threshold is thirty percent. Below thirty percent on-time I stop treating the schedule as a plan and start treating it as a guess with a distribution around it. Asia-North America at 26.9% is below the line. Asia-Europe at 9.9% is not a lane at the moment, it is a sailing schedule with optional arrival dates. Crossing that line changes four things for an importer: buffer goes from seven days to fifteen; booking lead time goes out by about three weeks; the forty boxes get split across two carriers with an on-time floor written into the contract rather than discussed over email; and the per-sailing proforma gets put in writing.

I would also set the review date now rather than waiting to feel like it: the first week of every month, pull the new series, and move the buffer and the carrier split only when the number crosses back above thirty percent or falls below twenty. Numbers between twenty and thirty mean hold what you have. Reacting to every weekly print is how companies end up carrying stock they bought for a crisis that ended.

WiseTech's advice, book earlier and add buffer, is correct and incomplete. Booking earlier without fixing the reorder point is motion without progress. Adding buffer without knowing whether your carrier runs at 14% or 61% is buying insurance against the wrong risk. Denominator, distribution, spread, then the money, in that order. One last check on yourself: if you cannot say, in one sentence, what one late FEU costs your business, then every reliability decision you make this quarter is a guess with a spreadsheet attached. Work that number out before you argue about freight, because it is the only input that turns this release into a decision.

A proforma schedule is a letter a carrier writes to itself. File it under marketing and plan against the arrivals. And when on-time sits below thirty percent with your buffer already stretched to fifteen days, that is the moment to lock space and rate, not the moment the rate looks cheap.

↑ Back to the key points

— By Vivian Zhao