SF Holding September Revenue Up 6.34% to RMB28.72bn as Parcel Volume Falls 7.25%

The Analysis

Every week I take two or three calls that open the same way. The shipper did nothing wrong. The goods were packed on time, the label was printed, the collection was booked. Then the tracking sat still for nine hours and the delivery promise printed on the product page quietly stopped being true. Nobody on the service desk will say the word breach. They say delay. They say weather. They say volume.

S.F. Holding's September operating bulletin, filed to the Hong Kong exchange on 19 October, explains more of those calls than any weather report does. Revenue from express logistics combined with supply chain and international came to RMB28.719bn, up 6.34% against a year earlier. As a headline that reads like a healthy month. Take it apart and it is a month in which the domestic network carried 7.25% fewer parcels while the supply chain and international arm grew 28.13%.

I cover the last mile, which means I read a revenue bulletin the way a dispatcher reads a vehicle schedule: what is moving, what has stopped moving, and at what hour of the day the promise stops being true. This one says the domestic cut-off is about to move.

Begin with the four numbers and the arithmetic sitting behind them, because the arithmetic is where the story lives. Express logistics revenue was RMB20.835bn, down 0.09%. Volume was 1.395bn parcels, down 7.25%. Revenue per parcel was RMB14.94, up 7.71%, the seventh consecutive month of year-on-year gain. Supply chain and international revenue was RMB7.884bn, up 28.13%, covering international express, forwarding and contract logistics.

Now run those backwards. If volume fell 7.25% to 1.395bn, the year-ago month handled about 1.504bn parcels, which means this network processed roughly 109 million fewer pieces in a single month. Divide that across thirty days and twenty-four hours and the network is carrying about 151,000 fewer parcels an hour through sorting halls, trunk trucks and delivery rounds. Hold on to that figure. It is the one that decides whether your five o'clock cut-off still holds.

If revenue per parcel rose 7.71% to RMB14.94, then last September it was about RMB13.87, so the gap is RMB1.07 a parcel. If supply chain and international rose 28.13% to RMB7.884bn, the year-ago figure was about RMB6.153bn, an increase of RMB1.731bn. Run the same exercise on the total and RMB28.719bn at plus 6.34% implies RMB27.006bn last year, an increase of RMB1.713bn.

Put those last two side by side and the month arranges itself. The group earned RMB1.713bn more than it did in September 2025. The international and supply chain arm alone earned RMB1.731bn more. The domestic express arm contributed minus RMB0.19bn. Every yuan of September's growth came from outside the domestic parcel network, and the domestic network was a small drag on it. The international share of group revenue moved from about 22.8% to about 27.5%, a shift of roughly 4.7 percentage points in twelve months. No company reallocates like that without moving aircraft, chartered belly space, line-haul drivers and sorting headcount inside the same buildings.

Why now is less of a mystery than people make it, but the explanation doing the rounds has it backwards. The story is that the price war ended and S.F. repriced. The timing supports half of that: if September was the seventh consecutive month of year-on-year gain in revenue per parcel, the turn happened around March 2026. Seven months of per-parcel gains on falling volume is not a repricing story. It is a mix story. The cheapest parcels walked out. What is left in the bag is heavier, more urgent, and more likely to be a return, a spare part, a cold-chain box or a cross-border leg.

Three pushes arrived from upstream at the same time. Domestic e-commerce platforms kept steering volume into consolidated fulfilment and self-collection, which strips the lowest-yield pieces out of the independent express network first, because that is where the margin was thinnest to begin with. Export platforms pushed more sellers onto semi-managed and fully managed models, which consolidates parcels into bigger, earlier, better-planned shipments and takes pricing authority away from the individual merchant. And cross-border air capacity tightened into the fourth quarter, which is exactly where the 28.13% sits.

September matters because of what sits immediately behind it. The October-to-December window carries Singles Day, Black Friday and Christmas. Capacity for that window is not bought in November; it is committed in the second half of October, and air cargo allotments for the peak are normally locked two to three weeks ahead of the flight, a month ahead for charter blocks. A bulletin released on 19 October therefore lands inside the precise week when next quarter's slot decisions are being signed. That is why this number deserves an hour of your attention rather than a scroll past.

Who eats it, and how much, splits five ways. Sellers on standard domestic parcels, the small and mid-sized merchants shipping direct to consumers, are the most exposed group and the least able to see it coming. Their per-parcel cost rises by the RMB1.07 the bulletin reports, and their same-day window shrinks in the same month. They absorb a price increase and a service reduction together, and only one of the two ever appears on an invoice.

Cross-border exporters get a different problem. Growth of 28.13% in supply chain and international means more of the network's lift is committed to international legs, and international legs carry the earliest handover times in the building: export documentation, consolidation and build-up typically pull the effective cut-off forward by six to eight hours against a domestic parcel, more when screening tightens. When the international book grows by nearly a third and the domestic book shrinks by 7%, the sorting hall does not run two schedules. It runs one schedule, and the earlier one wins the labour.

Contract logistics customers, the ones moving pallets rather than parcels, are on the winning side this quarter. Capacity follows revenue, and revenue is plainly moving their way. Their risk is not the rate; it is complacency, because a carrier growing 28% internationally is a carrier whose domestic contingency options are thinning out underneath them.

Time-definite shippers are the fragile group: pharma, cold chain, critical spares. Their economics depend on the last departure of the day, which is precisely the departure that gets cancelled first when volume falls. A network that loses 151,000 parcels an hour does not cancel the eight o'clock trunk. It cancels the two o'clock one that only half filled.

Last-mile pickup points and locker banks absorb the residual. Fewer parcels an hour erodes the drop density that made a locker wall or a village collection point viable at all, and the first thing to go in that situation is the evening collection, which pushes the effective cut-off earlier once more. It is a loop rather than a list.

Time it out, because timing is the entire point. Take a standard domestic parcel on a next-day promise. A merchant finishes packing between 2pm and 5pm. The station cut-off sits between 5pm and 6pm. Local collection runs 6pm to 8pm. The sort hub works 8pm to 10pm. Trunk departures leave between 10pm and midnight, run overnight, and land at the destination hub between 2am and 5am. A second sort runs 5am to 7am. Delivery stations load between 7am and 9am, vans roll from 9am, and the signature lands somewhere before 6pm. Cut-off to doorbell is roughly twenty hours, and every one of those blocks is sized to volume.

Now remove 7.25% of the volume and ask which block breaks. It is not the sort, because sorting lines get quicker when they have less to sort. It is not the trunk run, because the trucks that remain run to plan and arrive early. It is the last departure of the night and the earliest collection of the evening, the two blocks that only fill when volume is high. When a trunk departure is cancelled or merged, the seller's five o'clock cut-off does not change on paper. What changes is that the 6pm collection now feeds a 1am departure instead of an 11pm one, and the parcel loses two hours before it has moved a kilometre. When a station drops its evening collection, the effective cut-off moves from 6pm to around 4.30pm, and no letter goes out about it.

That is the transmission path, and it runs on a specific clock. The bulletin covers September, which means the network decisions were taken in August and executed through September. The observable symptom, a cut-off that has drifted from 6pm to somewhere between 4.30pm and 5.30pm, a tracking scan that stalls eight to twelve hours at the origin hub rather than three to four, shows up in the merchant's own data within one to two weeks of the change, and it surfaces on a Friday or a Monday, because those are the days a merged schedule hurts most. By the time it reaches a service-level report the quarter is half gone.

Cross-border runs on a longer and harder clock. If a domestic parcel needs twenty hours from cut-off to door, a cross-border parcel on an economy service needs three to five days, of which export clearance and build-up consume four to eight hours that cannot be compressed by paying more. Air allotments for the November-December peak are committed in the second half of October. A merchant who has not lodged volume forecasts and booking requests by roughly the end of this month is bidding for whatever is left, and in a peak season when one carrier's international book is up 28%, what is left is both expensive and late.

Here is the part the headlines skip, and it is the part that costs shippers real money.

Every write-up I have read frames the 7.71% gain in revenue per parcel as good news: the price war is over, yields are recovering, the sector is rational again. For an investor that may be right. For a shipper it is very close to irrelevant, and the reason is mechanical rather than commercial. Networks are scheduled on pieces, not on revenue. RMB1.07 more per parcel does not add a van, does not add a driver, does not add a departure. But 109 million fewer parcels in a month takes about 151,000 an hour out of the halls, and departures are filled on pieces. Revenue up 6.34% while pieces fall 7.25% is not a stronger network. It is a network carrying less work at a higher price each, which is precisely the condition under which frequencies get merged.

So the thing nobody is saying out loud is this. The breach is not going to happen in transit, and it is not going to be reported as a breach. It is going to happen at the cut-off, in the ninety minutes before the collection van leaves the station, and it will be described to your customer as a delay. Your published delivery promise will not change. Your carrier's terms will not change. What changes is that your five o'clock pack-and-ship window becomes a half-past-three window, your same-day dispatch rate drops by some number of percentage points you are not currently measuring, and your product page goes on promising a delivery date the operational chain can no longer reliably hit. The promise was written in days. The failure arrives in minutes. That gap is where the money drains away.

There is a second-order version of the same problem and it sits on the cross-border side. The 28.13% line is the fastest-growing piece of the business, international parcels carry earlier cut-offs than domestic ones, and international documentation is far less forgiving of an error. When a sorting hall's best labour and its best departure slots get pulled toward the growing book, the domestic parcel is the one that waits. Merchants running both will notice their export lanes improving and their domestic replenishment lanes quietly degrading inside the same month, and they will diagnose it as two unrelated problems.

Run the money, with the assumptions stated out in the open. Assume you are an exporter sending a 2kg parcel to Europe or North America on a standard economy line, and assume your all-in freight before fuel and handling is RMB100 per parcel. Assume further that the 7.71% move in per-parcel revenue is passed through to you in full, which it will not be, but take the worst case first. That is RMB7.71 more per parcel, and at 10,000 parcels a year the annual bill rises by RMB77,100. Now assume competition forces carriers to absorb half of it, so you see 3.86%: RMB3.86 a parcel, RMB38,550 a year. The honest answer sits between the two, call it RMB40,000 to RMB77,000 a year on a 10,000-parcel account, and it is a number you can argue against because it is visible on an invoice.

Now cost the failure that is not visible. Same account, 10,000 parcels a year, average declared value RMB800. Assume a cut-off that has quietly moved ninety minutes earlier costs you a two-percentage-point drop in same-day dispatch. That is 200 parcels a year leaving a day late. Assume that on a five-day service, one late day raises the complaint rate on affected parcels by one percentage point and the return rate by half a point: two complaints and one return. At RMB20 to handle a complaint and RMB60 of round-trip freight and markdown on a return, that is about RMB100 a year. Negligible, and this is where most people stop counting.

Run it again with the assumption no carrier will make for you, which is that a moving cut-off hurts hardest on the days your volume is concentrated, before and after a promotion. Same arithmetic on a Singles Day week carrying 1,500 of your 10,000 parcels: same-day dispatch drops five points, that is 75 parcels, complaint rate up five points on those, return rate up two points. Roughly four complaints and one and a half returns per hundred affected parcels inside a single week, plus the marketplace ranking penalty that follows a late-dispatch flag. Call it RMB6,000 to RMB9,000 in direct cost and an unknown but non-zero amount of lost listing traffic. Still smaller than the freight increase, and that is precisely the trap. You will fight hard over RMB77,100 you can see. You will not notice the dispatch leak for two quarters, by which point it has moved your service tier and your marketplace ranking at the same time.

One more, on the capacity side, because it is the only one with a date on it. Assume you move 3 tonnes a month on an export lane into Europe, and assume peak air allotment is committed two to three weeks before the flight date, which puts the window for November and December capacity in the second half of October. Assume spot rates in a peak week run 30% to 60% above contracted allotment. At a contracted RMB28 per kilogram, the difference between a booked month and a spot month is roughly RMB25,000 to RMB50,000, and the difference between booked and unbooked in the last ten days of November is sometimes a departure that does not exist at any price. That one has a deadline: the end of this month.

Two things would turn this conclusion around, and one of them is a measurement argument rather than a market argument. The measurement argument first, because it is the weaker link in what I have written. The supply chain and international line bundles international express, freight forwarding and contract logistics, and forwarding revenue in this industry is generally recognised gross, meaning the full amount billed to the customer, including what is passed on to the airline or the ocean carrier, is booked as revenue. If a large share of the 28.13% came from forwarding rather than from owned lift, that line can grow by nearly a third without a single additional aircraft or chartered block, and my claim that capacity has physically migrated to international weakens a long way.

The bulletin does not split the three. Nobody should be drawing capacity conclusions from a revenue line that may be gross. I think the shift is real, because contract logistics and international express are the two pieces that move metal, and a 4.7 point move in revenue mix in twelve months is too large to be pure accounting. But it belongs in the next quarterly filing rather than the next monthly bulletin, and it is the number I would check before betting a budget on it.

The market argument is simpler. If domestic parcel volume turns positive again in the fourth quarter, which a strong Singles Day can do on its own, the merged departures come back, the cut-offs drift back toward six o'clock, and the ninety-minute problem dissolves without anybody having done anything at all. The September bulletin is a snapshot of a network in the two months before peak, and the two months before peak are the trough. Read it as a permanent restructuring and you will over-hedge and pay for capacity you do not need. Read it as noise and you will be the account that missed the October booking window.

There is a third possibility I would not rule out either, which is that part of the 7.25% volume decline is deliberate, a continuation of shedding the lowest-yield accounts that began around March. If so, the price rise and the volume decline are one decision rather than two events, and the merchants who were dropped are not reading this because they are no longer shipping on this network. For everyone still on it, the operational consequence is identical.

So stop reading this as a rates story. It is a clock story. The money moved, the pieces did not follow, and the gap between them lands in the ninety minutes before your van leaves.

↑ Back to the key points

— By Nadia Pryce