J&T Express Hits 104.2m Parcels a Day in Q3: Volume Up 24.8%, New Markets Up 159.5%

The Analysis

A client of mine selling phone cases and cheap home goods into Indonesia, Thailand and Mexico had me on a call yesterday morning, and he started by reading me his own calendar out loud. Eleven eleven, twenty seven November, cyber Monday, twelve twelve. Four dates inside five weeks, each one forcing every parcel he owns into the same network on the same night, and he wanted to know whether the network is bigger than it was last year. It is. I told him that is precisely the problem, and he did not like that answer, so let me show him the arithmetic.

On 15 October J&T Global Express put out its third quarter operating update. Nine point five eight billion parcels in the quarter, up twenty four point eight per cent year on year, one hundred and four point two million a day, the second quarter in a row above one hundred million. Nine months in, twenty seven point zero nine billion parcels, up twenty five per cent, ninety nine point two million a day. Southeast Asia did two point eight nine billion, up forty four point five, thirty one point four million a day. China six point four three billion, up fifteen point three, sixty nine point nine million a day. The segment the headlines grabbed, other markets, did two hundred and seventy point eight million, up one hundred and fifty nine point five per cent, two point nine million a day, driven by Latin America and the Middle East on TikTok, SHEIN, Temu, Kwai, AliExpress and Mercado Libre volume.

Every headline written about that release stops at the growth rates, and every one of them is asking the wrong question. The question that decides whether your November parcels land on time is not how much bigger this network got. It is whether it got bigger faster than its own load, and how many hours of slack are left in it when four million extra pieces land on the same Tuesday night in the same sorting centre. I work in hours, so let me take this apart in hours.

Start with the facts that actually carry weight, which are not the percentages. The network ended September with twenty thousand three hundred outlets, two hundred and ninety two sorting centres and four hundred and fifty seven automated sorting lines. Against the end of 2025 that is roughly a thousand more outlets, forty six more sorting centres and forty four more automated lines. Nine months of building for nine months of growth, so put the two columns next to each other and stop looking at them separately.

Volume grew twenty four point eight per cent in the quarter. Outlets grew about five point two per cent, automated lines about ten point seven, sorting centres about eighteen point seven. Nothing in that network grew as fast as the freight moving through it, and the two lines that actually decide throughput, automated lines and sorting centres, grew at roughly half and three quarters of the pace of volume. That gap is the whole story and it did not appear in anybody's write up. A network can be bigger this year than last year and still enter the fourth quarter with less room in it, because what matters in November is not how much capacity exists but how much sits empty on a normal Tuesday waiting to absorb a Tuesday that is not normal.

Then do the per unit arithmetic, which is where it gets physical. One hundred and four point two million parcels a day across four hundred and fifty seven automated sorting lines is about two hundred and twenty eight thousand pieces per line per day, and I am going to hand you my assumption rather than hide it: assume each parcel meets roughly one automated line and assume those lines run something like twenty effective hours. That puts each line at around eleven thousand four hundred parcels an hour, every hour, of every day. It is a rough number and the direction is what matters. There is no manual army waiting behind that line. The moment anything interrupts it, a belt, power, a storm, a road, that volume does not quietly route around the problem, it stacks up in front of it, and everything downstream inherits the delay in units of hours.

Now the same sum at the outlet, because that is where you actually touch this network. Twenty thousand three hundred outlets handling one hundred and four point two million pieces a day is about five thousand one hundred pieces per outlet per day. Five thousand pieces a day per counter is a queueing machine, not a shop. Your parcel's first six hours are decided by whatever that queue looks like at the moment it arrives, and every other seller on those same platforms is feeding the same queue at the same time. This is why the lever is never the cheap freight rate. The lever is whether your box gets into that queue before the wave or after it.

Before I get to your clock, deal with the number everyone shared and nobody translated. Other markets grew one hundred and fifty nine point five per cent and produced two hundred and seventy point eight million parcels in the quarter. Reverse the growth rate and last year's quarter was about one hundred and four million pieces, so the increment is about one hundred and sixty six million parcels across ninety two days, call it one point eight million extra pieces a day arriving in networks in Latin America and the Middle East that had almost nothing built for them. Real, explosive, and small. Now run the same trick on Southeast Asia. Two point eight nine billion against forty four point five per cent growth means last year's quarter was about two billion, an increment of eight hundred and ninety million pieces, about nine point seven million extra pieces a day. China adds another nine point three million a day on the same arithmetic.

Put those three increments side by side and the headline falls apart. The fastest growing segment contributed one point eight million extra pieces a day. The boring one contributed nine point seven. If you ship into Southeast Asia you are standing under the real load, and the Southeast Asian network absorbed it with ten extra sorting centres, taking it to one hundred and thirty one, seventeen extra automated lines, taking it to eighty one, and roughly a hundred extra line haul vehicles, taking it to five thousand nine hundred. Sortation capacity in the region grew about twenty six point six per cent and trucks grew under two per cent, against forty four point five per cent more parcels. Whatever thin padding existed in that regional network last November is thinner now, and it is thinner in exactly the places your customers live.

China looks different and I want to give you that because it changes where you push. Line haul vehicles went up by about five hundred to seven thousand seven hundred, and about three hundred of that increase was self owned capacity, self owned now at five thousand six hundred. So two thirds of the extra trunk capacity came in house rather than being brokered, which tells you something about how the fourth quarter gets planned. Centralised capacity is reliable when it is planned well and rigid when plans change, and campaigns always change. There is less negotiation available to you in November inside a company owned trunk fleet than there was in the version of that network where a bigger share of trucks belonged to contractors who wanted your freight.

Why is any of this happening now, up this steeply, in these particular corridors? Follow the money upstream and it is not demand in the classic sense. It is the campaign calendars of about six platforms, all of them pushing sellers into livestream and flash formats that concentrate order generation into windows measured in hours rather than spreading it across a week. That has a different physical shape from ordinary marketplace demand. Ordinary ecommerce demand arrives across twenty four hours and sorts across twenty four hours. Livestream demand arrives in ninety minutes and then has to sort across a day. Your parcel is not competing with the average parcel. It is competing with everyone else's parcel generated inside the same ninety minutes, landing in the same intake queue, in a network whose automated line count grew at half the speed of its volume.

That reframes the whole thing from a capacity problem to a timing problem, and timing problems have cheap answers. So take your own operation hour by hour and look for where you actually control something.

Hour zero is the order. The platform is counting your promise from this moment, not from the moment your warehouse notices. Whatever dispatch window your seller terms specify, and you should go and read yours today rather than trusting my description of them, the clock is already running while your team is still asleep. Every hour you lose between order and pick comes straight off the end of the delivery promise, and nobody outside your building can win it back.

Hours zero to six are yours and they are the cheapest hours you own. Pick, pack, label, manifest. Your own cut off time here is a decision you made years ago and have probably never revisited against a campaign calendar. If your cut off is six in the evening for a driver arriving at eight, you have built your day around a normal Tuesday. On the night of the eleventh there is no normal anything. Pulling that cut off forward by four hours during campaign windows does not cost you a rate increase, it costs you a shift pattern, and buying back four hours at the front of the chain is worth more than any concession anybody will ever win at the back of it.

Hours six to eighteen are the first mile and they are where most sellers lose their nerve because they have no visibility. Your parcels sit waiting for collection. On a normal day the van comes when it says it comes. On a campaign day the driver is covering twenty stops of pickup requests that all went in at once, and pickup slides by hours that nobody reports to anybody. Measure it. Time from your manifest event to their first sort scan, per destination, per day. It takes one person half a day to build that report and it is the single best early warning instrument you can put in front of yourself, because it moves before the delivery failures do, sometimes two or three days before.

Hours eighteen to thirty six are the intake queue at the sorting centre, and everything about this piece converges here. This is where your parcel meets everybody else's. It is also where that thin padding gets spent. A network handling eleven thousand four hundred pieces an hour per line does not have a quiet corner for the surprise. If a shift's worth of volume arrives in one window, the delay does not stay in that window, it carries into the next sorting cycle, then the next line haul departure, then arrives at the destination sort a half day late, and by the time it reaches the van it has cost you thirty six hours you will never get back and cannot bill anybody for.

Day two to four is the line haul and this is honestly the part that worries me least. Trucks move, capacity was added, the network has practice. Then comes destination sort and the last mile, and if you have read any of my other pieces you know what I think about last mile during peak. First attempt delivery drops for reasons that have nothing to do with the network carrying your parcel halfway across a sea, and each failed attempt costs twenty four hours in a normal week and more in a campaign week. Add thirty six hours of inherited upstream delay to a first attempt failure and you are now comfortably past whatever the platform promised, and past it in the tail of your distribution, which is exactly where platforms measure you.

Here is the part nobody wrote about this release, and it is not the growth. It is that the operating update itself has quietly become a leading indicator you should be reading every quarter, whether or not you ship with this company. Compare three lines against one line: parcel volume growth against sortation capacity growth, against line growth, against network density. When volume growth outruns them all, as it did this quarter, twenty four point eight against ten point seven and eighteen point seven, the slack in that network is going down, and slack is the only thing that absorbs your peak. Nothing else absorbs it. A network with no slack does not fail on average days, it fails on the days you care about, and then it recovers slowly, because a backlog leaves in the same units it arrived in, hours at a time.

The second thing nobody wrote is about the calendar rather than the network. Eleven eleven, the twenty seventh of November, cyber Monday on the thirtieth, twelve twelve. Four waves inside roughly five weeks, and what matters between them is not the peak, it is the recovery. Campaign waves are not independent events. Each one leaves a backlog, and the next one arrives before the last backlog has finished clearing, so the second wave starts from a worse place than the first and the third starts worse than the second. Nobody plans for that. Everybody plans each date as though the network starts empty. If you are deciding what to promise customers, you should be adding recovery time to waves two, three and four, not to wave one.

Now count it, with every assumption written down so you can replace mine with your own numbers.

Assume you are doing four thousand pieces a day into Southeast Asia across the platforms, at an average order value of twenty two dollars, with contribution per order after product, freight and platform fees of eighteen per cent, three dollars ninety six. Assume your normal week P90 transit order to door is six point one days. Assume the campaign window pushes that P90 out by one point four days to seven point five. Assume the platform's own late threshold triggers cancellation or non collection, and assume every day of slippage past the promise costs you zero point six of a point of lost orders. Assume twenty of the forty three selling days between eight November and twenty December sit inside those campaign or recovery windows. Assume customer acquisition cost of one dollar eighty an order, already spent and unrecoverable when the order dies.

Do it. One point four days of slippage times zero point six points is zero point eight four of a point, which on four thousand pieces is about thirty four lost orders a day, one hundred and thirty three dollars of contribution. Add thirty four times one eighty of sunk acquisition, another sixty one dollars a day, call it one hundred and ninety four a day. Across twenty affected days that is three thousand eight hundred and eighty dollars, and if you think that number is too small to bother with, multiply it by the size of the operation you actually run rather than the one I invented, and then remember there is a refund handling cost and a review cost sitting behind every one of those orders as well.

Now test the expensive answer against it. Take pre positioning thirty per cent of your campaign inventory in country. Assume it costs you sixty two cents a piece extra in inbound freight, handling and storage, twelve hundred pieces a day, seven hundred and forty four dollars a day. Against that you remove about one point two days of cross border transit on those pieces, which is zero point seven two of a point of losses on twelve hundred units, roughly nine orders a day, thirty four dollars of contribution plus sixteen of acquisition. Fifty dollars saved against seven hundred and forty four spent. Pre positioning does not pay for itself inside one fourth quarter at these unit economics. It pays across a full year of repeat volume on your top twenty SKUs, and anyone pitching it to you as a November fix is doing arithmetic on someone else's business.

So what does pay? Cheap things, precisely because the constraint is timing rather than capacity. Moving your own cut off forward four hours on the six heaviest days of the year costs you a shift pattern and nothing else, and it buys you hours at the front of a queue where hours compound. Launching your own flash promotion thirty six to forty eight hours ahead of the platform's big day costs you nothing at all, only coordination between your marketing and operations people who probably do not talk to each other often enough, and it puts your volume into the intake queue before everyone else's arrives. Converting your public delivery promise from a fixed number of days to a window costs you a copy change and saves you the refunds and the angry tickets that come from promising a number you now know you cannot hold. None of these require the network to change anything.

Where does this turn out to be nothing much? Two places, and one of them is testable next quarter. It is noise if January's fourth quarter update shows sortation capacity and line count growing faster than parcel volume, because that would mean slack is being rebuilt and this network enters next November looser than it enters this one. Watch that update the way you watch your own lane data. It is also noise for any exporter whose campaigns genuinely do not sit on the platform calendar, which is rarer than people admit, because even sellers with their own storefronts get pulled into those dates by their own customers' expectations.

The way it flips hard is the platform dependence itself. Those relationships are named in the release as the source of the growth: TikTok, SHEIN, Temu, Kwai, AliExpress, Mercado Libre. That is concentration. A single platform changing its logistics policy, insourcing more fulfillment, altering which carriers it presents at checkout, or shifting responsibility for failed deliveries, moves twenty four point eight per cent growth in one direction or another inside a quarter, and it moves your service levels with it while you have no seat at that table. Hold your corridor plan lightly. Keep a second last mile option warm in each of your top three markets, not because this network will fail you, but because your access to it is not entirely your own decision.

I know how this reads, so let me be clear about the bottom line. This is not a warning about a company in trouble. Twenty four point eight per cent, twenty five per cent over nine months, two straight quarters above a hundred million a day, forty four point five in Southeast Asia, one hundred and fifty nine point five in new markets: that is a machine that is working. My point is narrower and it is about your clock, not their health. Growth like that eats the padding first. Padding is what absorbs your four worst weeks of the year. It is the twentieth of October today. You have about two and a half weeks before the first wave, and every one of the cheap moves above is worth more this week than it will be worth on the ninth of November, because the only thing left to buy later is expensive.

↑ Back to the key points

— By Nadia Pryce