DHL Express Expands European Network Capacity Ahead of Peak Season

The Analysis

It is 05:20 at an express gateway and the night sort is running half an hour behind. A line-haul from the continent touched down at 04:40 instead of 04:05, so the belt is still full when the first wave of delivery vans starts loading, so three routes pull out fifteen minutes late, and fifteen minutes at 05:20 is the whole difference between hitting a consignee goods-in window and missing it.

That is the picture in my head when I read what DHL Express put out on 23 June: more European network capacity, additional aircraft leases, upgraded sorting hubs at Leipzig and East Midlands, all pointed at the peak volumes everybody expects. Behind it, the parent group has put 160 million euros into French infrastructure. Good news, and I mean that. It is also news about somebody else's building, and my job is to work out how many minutes of it actually reach my dock.

Here is my reading in one line. This is a capacity and resilience story, and the only parts of it that touch you are a cutoff time and a recovery option. Everything else is the carrier's problem, and carriers fix their own problems first. So by Friday I want two answers in writing: does my latest tender time move later and by how many minutes, and if the hub my freight clears through chokes in November, what is my second path and how many hours does it take to switch.

Let me unpack the announcement the way a dispatcher would, not the way a press office would. Additional aircraft leases means more metal in the night, and more metal in the night usually means a rotation that did not exist before. A new rotation matters to you in exactly one way: it either moves your latest tender later, or it creates a recovery option that did not exist when a flight was missed. If it does neither on your lane, it is invisible, and you should keep planning as if it does not exist.

Upgraded sorting hubs at two named locations, Leipzig and East Midlands, is the more interesting half. Sorting capacity is measured in pieces per hour, and pieces per hour is what decides whether a belt clears at 04:40 or at 05:20. When a hub is being upgraded, though, there is a phase where throughput is worse before it is better: lanes closed, equipment commissioned, staff retrained on a layout they do not know yet. If your freight runs through either of those two hubs, ask your account contact one specific question: what is the contingency plan during the works, and which gateway takes the overflow. Ask it in June, while there is still time to route around it.

The stated goal, in the carrier's own framing, is better service resilience and less exposure to single-hub bottlenecks. Read that carefully, because resilience belongs to whoever owns the decision. A network that can move a bottleneck is a network that can move your freight somewhere else without asking you, and it will do that when it is under pressure, not when it is convenient for you. So the useful move on your side is to decide in advance which alternative you are willing to accept, and to say so on the account before peak starts. If you let the carrier choose for you in the middle of November, you will get whatever path protects the network, not the one that protects your delivery window.

Now the number everybody will quote at you: 160 million euros into French infrastructure from the parent group. I am not going to pretend to know how it is split or when each piece lands, because the announcement does not say. What I will say is this. That is a group betting on ground infrastructure in one country, at a scale that tells you where it expects volume to sit for years. If a meaningful share of your European freight touches France, that is worth a conversation with your account team about road network timing and cutoffs over the next few quarters. If none of your freight touches France, it is context, not a lever, and you should not build a plan on it.

Let me put hours on the hub question, because that is the part that turns into money. Assume you move 800 pieces a week through a gateway that sorts at one of the two named hubs. Assume peak multiplies that by 1.6, which is a normal fourth-quarter shape for a lot of importers, so 1280 pieces a week. Spread over a six-night operating week that is roughly 213 pieces a night, and across a night sort of about eight hours, call it 27 pieces an hour. Now assume a hub event that stops the sort for six hours. Six times 27 is about 160 pieces that sit. Of those, assume twenty percent need a re-delivery the next day rather than a late same-day run. That is 32 re-deliveries, and at fifteen minutes of handling each, including the re-book and the phone call to the consignee, that is eight hours of your team's time plus roughly 20 hours of delay on the customer side.

That is the number I care about. Not the 160 million, not the aircraft count. About 160 pieces sitting and roughly 28 hours of combined internal and customer-facing time, from a single six-hour event. Now price the insurance. Assume you pre-qualify a second gateway and move fifteen percent of your weekly volume through it as a matter of routine, so 192 pieces a week on the alternative path. Assume that path costs you two extra hours of road feeder time at each end on average. That is four hours per load, and if a load is 60 pieces, that is roughly three loads a week and about twelve hours of added transit per week, spread over 192 pieces.

Twelve hours of added transit on 192 pieces, against 28 hours of recovery time on 160 pieces when something breaks. On paper the insurance looks expensive. It stops looking expensive the first night you need it, because the 28 hours land all at once and the twelve hours are spread out and planned.

There is a cheaper version of the same insurance and I would start there. You do not have to move volume to hold a second path open. Run one test load a week through the alternative gateway, sixty pieces, and log four timestamps per piece: tender, depart scan, hub scan, delivery signature with a time. After eight weeks you know the real transit difference in hours, you know whether the customs clearance at that gateway behaves the same way, and you have a relationship and a booking reference that already exist. When you need to shift three hundred pieces in a hurry in November, you are activating something you have used eight times, not something you have read about once.

The same arithmetic applies to failure and retry at the far end, which is where most of the November pain actually shows up. Assume a first-attempt failure rate of seven percent during peak, higher than your steady-state rate, because consignees are congested, because their goods-in windows are oversubscribed, and because temporary staff are taking signatures. On 1280 pieces a week that is about 90 failed attempts a week. Assume each one costs fifteen minutes of internal handling and adds twelve to twenty hours of delay. That is 22 hours a week of your team's time spent on re-delivery admin alone. Cut the failure rate from seven percent to four percent with booked delivery windows and a named contact at every high-volume consignee, and you get back roughly ten hours a week, which is a quarter of a person.

That is the kind of number this announcement should push you to go and measure. Capacity at the top of the network does not fix a failed doorstep attempt. Nothing in Leipzig or East Midlands will knock on a consignee's door at 09:00 on a Tuesday. What fixes it is a booking system on your side and a rule that every consignee above a threshold volume has a delivery window and a human name attached to it.

So what do you actually do this week. Get the cutoff in minutes for every lane you run, in writing, from the carrier, and then check it against your own dock reality. If the published latest tender is 18:00 and your pick-pack operation finishes at 17:30 on a good day, your real buffer is thirty minutes, and a single late order destroys it. Write that buffer down for every lane. Any lane with less than sixty minutes of buffer between your last realistic pack time and the tender cutoff is a lane that will fail you in November, and you should either move the cutoff conversation now or move the pick-pack start earlier.

Then build the exception ledger. Every exception gets detected-at, human-acted-at, action taken, minutes lost, and cost, the same five fields. It takes about four minutes per exception to fill in, and after a month it tells you which lane, which gateway, and which consignee is eating your hours. I have never once seen a team guess this correctly from memory. They always blame the lane that shouts loudest rather than the one that leaks most hours.

Then get your peak allocation in early. Capacity commitments are made against forecasts, and forecasts are built months ahead. If your peak starts in the first week of November, the conversation with your account team happens in September, with weekly piece numbers by lane, not with a hopeful phone call in October. Ask two questions in that meeting. What is my committed weekly capacity during the peak weeks, in pieces. And what happens, in hours, when I exceed it. The second answer is the one that matters, because exceeding your allocation during peak usually does not mean a polite rejection, it means your freight rolls to the next available rotation and you find out after the fact.

Now the alternatives, and the traps in each one. Splitting across two gateways is the obvious hedge and it has a real cost: two sets of cutoffs to track, two clearance behaviours to learn, two relationships to maintain, and usually a slightly worse rate on each because you have split the volume. My rule of thumb is the one I gave earlier, hold a tested second path at ten to fifteen percent rather than a full split, unless a single gateway already carries more than about seventy percent of your peak volume. Above that concentration, the hedge is worth the overhead.

Switching to road for part of the leg is the other option, and for European freight it is more viable than people assume. A road feeder of six to ten hours can beat an air leg plus two hub touches when the network is tight, especially on lanes under about 1200 kilometres. The trap is the cutoff arithmetic: a road leg has a departure time and it is usually early evening, and if your operation cannot be loaded by then, the road option is theoretical. Check the departure time in hours against your own pack time before you put road into a plan.

Moving your inventory forward beats both of them for the right product. Two weeks of forward stock at a European warehouse turns a November capacity problem into a September inbound problem, and September inbound is a much easier problem. It costs you carrying cost and it costs you forecasting discipline, and for anything with a short shelf life or a fast-changing spec it is the wrong call. For everything else, it is the cheapest hour insurance available, and it is entirely under your control, which no carrier capacity promise ever is.

The traps around the announcement itself are worth naming. Trap one: reading additional capacity as a promise about your lane. Capacity gets added where the network needs it, which is usually where the volume density is. If your lane is thin, more capacity in the system does not mean more capacity for you, and peak surcharges will still apply to whatever you ship. Trap two: assuming hub upgrades reduce your transit immediately. During the works, transit often gets worse on the affected flows. Ask about the schedule for the works, in weeks, and about where the overflow goes.

Trap three is the billing basis on peak surcharges, and it catches people every year. Surcharges get quoted per piece or per kilogram or per shipment, and the three produce very different bills on the same freight. Get the basis in writing, model it against your actual piece count and weight distribution, and compare quotes only after you have normalised the basis. A quote that looks cheaper per shipment can be more expensive per piece once you run your real mix through it.

Trap four is the commitment shape. If you commit weekly volumes to get capacity, ask what the penalty is for undershooting, and in what units. A take-or-pay clause measured in pieces per week across a twelve-week peak is a very different exposure from a soft target with no penalty. Sign the one you can actually hit if your demand forecast is off by twenty percent, because in most years it is off by at least that much in one direction.

Trap five is the quiet one: cutoffs move. A network rebalancing ahead of peak shifts tender times, hub injection times, and last-belt times, sometimes by thirty minutes, sometimes by an hour. If you are not checking your cutoff board weekly from September onwards, you will discover the change the hard way, on a night when a van leaves without your freight on it. Put a recurring fifteen-minute check in somebody's calendar every Monday morning and make them read the numbers out loud against last week's.

Write the switch trigger in hours, not in feelings. Mine would look like this: if the hub scan for more than twenty percent of my pieces runs more than two hours late against the same weekday two weeks earlier, I move the next three nights of volume to the second gateway. That is a rule a duty manager can apply at 03:00 without calling anybody. Without a rule, the decision gets made by whoever shouts loudest on the phone, and it gets made about two hours later than it should have been.

And a word on the leased aircraft, because people over-read this bit. The announcement says additional aircraft leases and does not put a number on it, so plan on zero until your own lane changes on paper. Leased metal enters the network on the densest lanes first, and if yours is not one of them, your cutoff does not move at all. Ask your account team which lanes gain rotations and in which week, and if they cannot tell you, assume the answer is not your lane yet.

Two named hubs is the part of this I like most, and here is why. If your freight can be sorted at either Leipzig or East Midlands, you have a natural A and B inside one carrier network, which is a cheaper hedge than building a second relationship from scratch. Ask which of the two your freight is planned into today, whether a lane can be reassigned at short notice, and what notice period that reassignment needs. Get the answer in writing before October, because an A and B you cannot switch between is just one hub with a longer name.

The other thing capacity announcements never cover is your own labour. Peak adds hours to your shift before it adds hours to the carrier's. Assume your pack operation has to start ninety minutes earlier in November to hit the same tender cutoff, because pieces per hour go up and your pack rate does not. Ninety minutes a night, six nights a week, across ten peak weeks, is ninety hours of added shift time, which is more than two full people at forty hours each. That is the number to take to finance in September. If you raise it in the last week of October, the answer will be no, and you will spend November explaining misses instead of preventing them.

Decide now what gets sacrificed, because when capacity is short, something does not fly. Write three tiers: contractual or time-critical freight, then high-value or customer-committed freight, then everything else. When your allocation is exceeded, the third tier rolls first, and the customer hears about it the same night rather than seventy-two hours later. A late honest message costs one phone call. A silent roll costs a chargeback and sometimes an account.

One more measurement, and it is the cheapest leading indicator I know: hub dwell in hours, per piece. Take the gap between the arrival scan and the departure scan at the sort hub, per piece, per night, and chart the median. It moves before your on-time numbers move. If your median dwell drifts from three hours to five hours across two weeks in October, the network is tightening under you and you still have time to shift volume. By the time it shows up in your delivery performance, you are reacting instead of moving.

Where does that leave me on 23 June. Cautiously pleased. A carrier adding aircraft, upgrading two hubs, and talking openly about reducing single-hub exposure is doing the right work before peak rather than during it, and the group putting 160 million euros into French infrastructure says it expects European volume to keep growing. None of that is a substitute for knowing your own buffer in minutes. The window shuts at the same time whether or not the belt has been upgraded.

If I were sitting in your planning meeting, I would leave with three things decided. A cutoff and buffer figure in minutes for every lane, written down and owned by a name. A second gateway tested with one load a week starting this month, so that November has somewhere to put freight. And a peak allocation conversation booked for September with weekly piece numbers and a written answer on what happens when you exceed it. Those three are boring and they are the whole game. Get the parcels moving first, then argue about the rate.

↑ Back to the key points

— By Nadia Pryce