Cambodia Makes PASNet Port Platform Mandatory at Sihanoukville from January

The Analysis

Ninety people sat in a room at Sihanoukville and heard one word that should have every customs desk in the region paying attention. Mandatory. From early next year every operator at the port has to work on the Port Autonomous System Network, PASNet, an electronic platform built for paperless processing, live container tracking and automated clearance. Deputy Prime Minister Sun Chanthol made the announcement in front of roughly ninety delegates, with chambers representing Chinese and European business in the audience. The numbers are designed to be repeated. Around two hundred dollars off the handling cost of every container in the first phase, another twenty to thirty percent further down the road, and un-invoiced charges cleared out. Nearly every write-up since has gone straight at the two hundred dollars.

I went somewhere else, because my first question is never the interesting one. When this goes live, which piece of paper stops counting as evidence? Clearance becomes automated, so the release instruction stops being a stamp somebody presses and becomes a row in somebody's database. A human being types that row in. Whoever types it owns the first mistake, and the exam order, the detention days and the amended declaration line up behind that mistake in that order. That is why I filed this under paperwork and not under software, and why anyone who took the government's wording at face value and filed it under technology has got the story the wrong way round.

I did a bit over ten years in a brokerage before I went in-house, and the one thing that never changes is that the manual is always cleaner than the yard. The PASNet manual will be very clean. Paperless clearance, real-time tracking, no off-book fees, all beautiful on paper. What happens in January depends on which terminal operator trained which data entry clerk, on whether your forwarder's booking office and its Phnom Penh office are typing into the same template, and on whether anybody has written down who pays when those two do not agree. Let me walk one shipment through, booking to gate-out, and point at the five places this jams.

Start with what mandatory actually attaches to, because the word is doing a lot of work. It attaches to all operators at Sihanoukville Autonomous Port, not to the port authority alone. Operators means whoever touches the box: the terminal company, the shipping line's local office, the forwarder, the customs broker, the trucking company, and anyone else the authorities have decided has to hold an account. If you ship through this port you are downstream of every one of those logins. Your cargo's freedom to move depends on data typed by somebody you have never met and cannot instruct directly, sitting in a screen you will never see.

Here is the route map I keep in my head for any shipment. Classification comes first, then the declared fields, then inspection, then tax, then release. Look at the HS code before you look at anything else, then look at the duty treatment that follows from it. In a manual world each of those five gates had its own folder and its own person, and your agent could walk down the corridor and argue with that person. Under a single platform all five gates read from the same data set. That is the whole revolution, and it is also the whole risk. It is no longer five separate arguments you can win one at a time; it is one argument repeated five times, and you either got it right at the beginning or you did not.

Gate one, classification. The tariff line for this commodity has to sit in the system before the container moves, not after it lands. In the old flow the code lived on the declaration form and you had a day or two of slack to convince someone it was right. Now the code is an input to whatever does the risk scoring. Wrong code, wrong score, and a wrong score is invisible to you until a container gets pulled. Classification is also what drives the duty rate and the preferential treatment, which is why I always say look at the tariff line first and argue about relief afterwards. The order matters. A code that is right for the duty rate can still be wrong for a preferential claim, and the platform will happily accept it either way.

Gate two is the declared fields, and this is where you will either be fine or bleed. The fields that have to be frozen before the truck leaves your supplier are these: the tariff line to whatever digit depth the national tariff uses, country of origin, the customs value and the basis you are declaring it on, quantity in the legal unit as well as your own commercial unit, any licence or permit number that attaches to the goods, the consignee's tax identifier, and container and seal numbers.

The basis you declare the value on deserves its own line, because it is the field people get wrong most often and the one they least notice getting wrong. If your commercial invoices are raised in one place on an FOB basis and your logistics desk books freight prepaid, the value for customs purposes changes with the freight amount. Do that inconsistently and the same SKU walks in every month with a different duty base. A human officer might not notice. A pattern-matching system notices patterns; that is the entire point of it.

There is a legal edge to the origin and preference part that I am going to stop short of, because it is not mine. Dr. Ingrid Voss is the person who can tell you which originating-status statement survives a paperless flow and who carries liability for a declared field under Cambodian law; I would only be guessing, and my guess from other jurisdictions is that it does not automatically follow the model most of us grew up with, where the declarant carries the consequence. That guess is worth nothing to you. Put it to local counsel before you rely on it. What I am giving you is the operational side, and it is not a legal conclusion.

Gate three, inspection. The instruction now comes out of a machine, and it comes with an audit trail. That part is genuinely good news; real-time tracking means you stop calling three people to find out where the box is. The bad news is quieter. If your container gets pulled because two fields disagree with each other, then what you need first is not a forklift, it is an amended invoice. The bottleneck moves from the yard into your own office and your supplier's office. I have stood in a yard and watched supervisors shrug while two companies argued for three days about who reissues a corrected commercial document, with the clock running on the box the entire time. That is what this looks like when it goes wrong, and it is never visible on the port's cost saving.

Gate four, tax. Assessment gets generated automatically off the declaration, which means the money leaves earlier than it used to. Nothing dishonest about it, but if your working capital is tight and you have been effectively financing yourself on the gap between declaration and payment, that gap is about to close. Tell whoever runs your cash in December, before it turns into a February problem.

Gate five, release. This is the sentence I would like you to read twice. When release becomes a record rather than a stamp, whoever controls that record controls your cargo. That is normally the terminal, and normally it is fine. But go and ask your agent one specific question this month: in the new flow, what actually flips a box to released — payment confirmation, a status flag, or a person clicking? And whose hand is on the mouse? If the answer comes back vague, that vagueness is your January.

Now the two hundred dollars, which is three different things wearing one number. Part of it is the handling saving, the operational cost of moving less paper between more desks, and that saving mostly belongs to whoever employs the crane driver. Part of it is gate time, truck turnaround, which genuinely reaches the trucker and sometimes reaches you through trucking rates. And part of it is removing money that never appeared on an invoice in the first place. That third part deserves a raised eyebrow. If a charge was un-invoiced, it was not on your books. You cannot save money you were never billed for. What you may get is a quieter, more predictable gate and a driver who stops asking for cash at the barrier. Converting any of that into a smaller number on your freight quote is a negotiation, not a system upgrade. Nobody has ever handed a beneficial cargo owner two hundred dollars because the port bought software.

Let me put the arithmetic on the table with every assumption showing, because that is the only honest way to do it. Assume you are moving thirty forty-foot high cubes a month through Sihanoukville, so three hundred and sixty boxes a year. Assume the two hundred dollars is real and lands on the handling bill. Thirty times two hundred is six thousand dollars a month and seventy-two thousand over twelve months. Now assume, because you have to assume something and I would rather be wrong in the open, that only forty percent of it reaches you as a lower bill and the other sixty percent stops somewhere between the terminal's margin and your forwarder's. That leaves twenty-eight thousand eight hundred dollars a year. Still a real number, still nowhere near the one in the headline.

Now the other column, and this is the part none of the coverage is running. Assume that during the first six months, while everybody is learning which screen does what, three percent of your declarations carry a data error serious enough to stop a container rather than simply get it amended later. On three hundred and sixty boxes that is roughly eleven incidents. Assume an incident costs four days — two waiting for the corrected document to be issued and accepted, two more for the exam that usually follows — and assume a hundred and twenty dollars a day in combined detention and demurrage on a forty-footer, plus something like two hundred dollars in amendment and local handling once it does get pulled. That is about six hundred and eighty dollars per incident. Eleven incidents is a little under seven thousand five hundred dollars in the year. Set that against the twenty-eight thousand eight hundred you were promised and about a quarter of it has already gone.

Push those same assumptions a little and see what happens. Take the error rate to five percent and give each incident six days instead of four, because in a first year nobody gets anything fixed in two days. That is eighteen incidents at roughly nine hundred and twenty dollars each, sixteen and a half thousand dollars, and more than half of your saving is gone. These are my numbers from an assumed box count and an assumed error rate, not measurements, and I have nothing from the pilot to tell me the true figure. But every hour that error rate sits above three percent is an hour the headline becomes fiction.

So here is the thing everyone else missed, and it is the reason I sat down to write this. Every story leads with two hundred dollars a container and nobody leads with the liability transfer that arrives on the same day. When a paper form walked between five desks, an error had five chances of getting caught by five people whose job it was to catch it. When one data set feeds five gates, an error has exactly one place to be created and then it propagates perfectly through the rest. The efficiency and the fragility are the same feature. In year one — and January will be year one — the fragility is the story, and the two hundred dollars is the decoration.

Who absorbs that cost when the input was wrong is the question to settle on paper before go-live, and I mean literally on paper, in the service agreement, before the first January booking. My ulcer tells me this will otherwise get settled on the fifteenth of January, standing in a yard, with a shift supervisor pointing at a screen showing six containers that cannot move.

Who gets hurt, in the order I would bet on. Exporters with a stable, repetitive SKU list and their own clean master data will barely notice a change; their codes and values were always internally consistent, they simply never had to prove it. Trading companies whose descriptions get written fresh every month by whoever happens to be typing that afternoon are the exposed ones, because their data has never had to survive contact with a matching engine. Anyone running LCL through Sihanoukville has an extra layer of trouble, since consolidation means one box carries the data of many declarants and one bad line holds up everybody else's cargo, which is not a conversation anybody enjoys having with a customer. And small shippers who let the agent originate everything are exposed in the most uncomfortable way: they never see the data, so they never see the error, they only see the detention invoice arrive with somebody else's handwriting on it.

When does it bite. Early next year sounds comfortably far away and it is not, because the preparation does not start in January, it starts as far back from January as your own lead time reaches. If your declaration data is currently final two days before sailing, and you now need clean data at booking rather than at departure, then your effective documentation cut-off has moved backwards by your entire booking lead time. For most shippers that is somewhere between three and five working days. Stack that across every shipment of a peak season and you can see that the actual window for doing anything about it is this quarter.

There is also a version of this where it does not launch everywhere at once. These things tend to phase in by terminal or by operator class, and it genuinely depends on which berth your carrier is booked into. If there is a phased start, there is a short-lived hedge in pointing bookings toward whoever goes live last. I do not love that hedge for long, because within two or three months everyone has converged and nobody remembers who was first. For the first six weeks it might buy you somebody else paying for the learning curve, which is not a bad trade.

And the honest counterfactual, because I am not certain about any of it. If it turns out that corrections inside PASNet are cheap and fast — an amended field accepted in a couple of hours without automatically triggering an exam — then every line of arithmetic above collapses and this is a good news story after all, and I will come back and say so. Everything here hangs on one unknown: how forgiving this system is to being corrected. Nobody outside the pilot knows that yet. It is not a number the government is going to publish. It is the number I want in February, and it is the number that decides whether you spent December for nothing.

The worse version is also easy to imagine. If liability for declared data defaults to whoever holds the account rather than whoever owns the cargo, then your forwarder or broker is carrying costs that you thought you had priced away. Nothing gets absorbed permanently in this industry; it comes back in the next rate round with an explanation attached. The cost does not disappear, it just changes its name and arrives on a different invoice.

One last thing that is actually useful and costs nothing. The tracking is real, so keep it. Pull the system's own timestamps — gate-in, exam instruction, exam start, exam finish, release — into a file of your own every month. Every detention invoice you have argued about in the last five years becomes arguable with evidence once those timestamps exist, and most shippers will not bother. Keep three months of them before January so you have a baseline, and bring them out when a yard tells you it was only two days.

↑ Back to the key points

— By Derek Xu