Those Who Bring You Things

III remember waiting for a parcel in Maputo once. Not waiting at the door. Waiting in the sense of having already adjusted to the reality that it probably wasn’t coming. You don’t stand by the window. You reroute. Couriers for anything that matters. Suitcases carried by friends visiting from abroad, then acquaintances, then eventually anyone going in the right direction. The postal system exists the way bad weather exists: you work around it, not with it. SAPO, as people in South Africa say, stands for Sorry, Awaiting Processing Outcome. I always thought that was too generous. Outcome implies there was a process.

I didn’t know that was the wrong baseline. Then I moved to Japan.

Yamato disagreed. They launched TA-Q-BIN [from taku (home), kyuu (speedy), bin (transport)] as a direct bet against the consensus. Not a premium service. A mass service. Door-to-door, nationwide, next-day. The bet was to make residential delivery reliable enough and people would use it at the volume required to cover the cost. They were right. The service became so synonymous with what it represents that Yamato has spent decades legally protecting the trademark from becoming generic, the way Xerox defends Xerox.

Sagawa Express followed. Japan Post upgraded. The result is a triad that competes on service quality because each is watching the others: Yamato at roughly 41% market share, Sagawa strong on business accounts, Japan Post running the national letter-and-parcel backbone through Yu-Pack. on service quality because each is watching the others.

Here is the process: You select a delivery window at checkout: a two-hour slot within a day. The driver arrives in that window. If you miss it, the slip in your mailbox carries the driver’s direct number. If you call before end of business and they may turn the truck around. Fail to reach them, scan the QR code, reschedule for that same evening, or redirect to the nearest konbini: Lawson, FamilyMart, 7-Eleven, open at 11pm or whenever you’re finally free.

Sagawa’s English support is thinner than Yamato’s. The redelivery tools default to Japanese, and calling the driver requires at least conversational ability in the language. These are all very understandable, but what isn’t is the hanko requirement on some registered items.

And the model has a ceiling: by 2017, Yamato was running 1.8 billion parcels on workforce capacity built for less. Drivers were clocking eleven-hour shifts. Yamato cut the morning delivery slot because drivers were collapsing under the schedule. They announced this publicly, explained the constraint honestly, and took the hit to recipient convenience. That decision is as important as any of the successful ones. It shows what an accountability architecture looks like when it’s working: acknowledging limits rather than papering over them.

The core mechanic is accountability running through the driver’s professional identity. When a Yamato driver fails to deliver, they are failing a specific person. The slip carries their number. The customer can call them. They may come back. Note that is not policy. It is culture the system produced and then relied upon. It makes the incentive structure different from a system where the driver logs an attempt and the package becomes someone else’s problem.

Let’s pause on Japan and go back further. Before Yamato, before DHL, before FedEx. Every version of the postal system has been built around the same question: infrastructure for whom?

The Thurn und Taxis family ran a postal monopoly across most of Europe from 1512 to 1867, three and a half centuries, 20,000 couriers on routes from Spain to Hungary, from the Low Countries to Italy. They got rich on state information: dispatches between courts, diplomatic correspondence, the connective tissue of empire. The public could use the network, but the network was not built for the public.

Rowland Hill changed this in 1840 with a pamphlet. His Post Office Reform: Its Importance and Practicability demonstrated that the cost of a letter bore almost no relationship to the distance it traveled. Infrastructure cost was fixed; routing cost was negligible. British postage rates at the time were graduated by distance and destination, which meant a letter from Dublin to London could cost more than a laborer’s daily wage. Hill argued, correctly, that dropping the price would increase volume enough to more than compensate for the reduction per unit.

Uniform penny postage launched January 10, 1840 and letter volume in the UK doubled in the first year. Within a decade it had doubled again. Within three years, Switzerland and Brazil had adopted stamps. By 1860, ninety countries had. The system had been pricing itself out of the population it was supposed to serve, and nobody had stopped to ask why it was structured that way.

The operational innovation came from FedEx, 133 years later. Fred Smith proposed the hub-and-spoke model in a Yale economics paper in 1965. His professor gave him a C (roughly a 70%). The idea: instead of routing packages on direct paths between every city pair, consolidate everything through a single central hub in Memphis, where it can be sorted overnight and redistributed by morning. Smith launched Federal Express in 1973 with fourteen Dassault Falcons and 186 packages on the first night. The company was nearly bankrupt in 1974; Smith flew to Las Vegas with the company’s last $5,000 and won $27,000 at blackjack, which kept the lights on another week. The hub-and-spoke model survived and became the structural backbone of modern logistics. Adopted by UPS, copied by every major carrier, extended internationally by DHL’s global network. The hub is where the efficiency lives. Everything up to and including the regional depot is solved.

Then there is the last mile. The final distance from the depot to your door. It costs between 50% and 70% of total delivery expense (the estimate varies by density and geography) because it is inherently point-to-point. You cannot consolidate delivery to 47 different apartment doors the way you consolidate 5,000 packages in a sorting hub. Every stop is different. Traffic, parking, building access, recipient availability. The efficiency logic of hub-and-spoke runs out of territory at the door of your building.

Which brings us to Germany.

Six major parcel carriers: DHL, Hermes, DPD, GLS, UPS, and Amazon Logistics. Together they handled about 98% of the 4.6 billion parcels delivered in 2024. DHL dominates at roughly 40-48% market share.

In 2025, Deutsche Post received over 55,000 official complaints via the Federal Network Agency, 90% about DHL, mostly for damaged parcels and wrong-address deliveries. Those are only the people who filed with the regulator. Deutsche Post separately disclosed receiving around 420,000 direct complaints in 2024, the first year they were legally required to report them. In Q3 2024, GLS’s first-attempt delivery success rate was 83.96%. DPD’s was 90.53%. DHL’s own Q2 2024 figure was 98.24%, which across 6.7 million daily parcels is still tens of thousands of failed deliveries per day, before asking whether ghost deliveries register in the first-attempt figures at all.

They probably don’t. The ghost delivery: tracking shows “attempted,” no bell was rung, no slip left. It is not anomalous. It is structurally produced. Drivers in Germany are routed by software calibrated to a fixed number of stops per hour. E-commerce volume has grown, particularly because of the pandemic and the rise of remote work. Headcount has not kept pace. DHL announced 8,000 job cuts in 2024-2025 while distributing 2.1 billion euros in dividends to shareholders, citing a 7.2% decline in operating profit even as the worker wage agreement cost 360 million euros. When the schedule becomes physically impossible to complete honestly, it gets completed dishonestly and the package is rerouted to a Filiale three kilometers away, open Tuesdays through Fridays from ten to six.

And here I need to step out of the analytical register for a moment because like many: I work. I have hours. I cannot simply sacrifice my lunch break every time DHL decides my Tuesday is less important than their quarterly targets. I have had the specific experience of planning to work from home to receive a package, then getting an email two days before saying it’s coming to my address (great, I live there) and then watching the missed delivery notification arrive at 11:43 on a Wednesday like a gnat I couldn’t swat. You cannot do anything. You have bills. The notification is there. The package is not.

The driver is not a bad person. They are a person inside a system that resolved the tension between volume and labor cost in the direction of volume, and then attributed the consequences to individual error.

Hermes, rebranded in the UK as Evri partly to escape its own reputation, and then ran a model that had couriers sorting packages before their paid shift began because the company hadn’t hired enough sorters. The Trustpilot reviews from early 2026 are identical to 2020: 3 out of 4 times the courier didn’t ring the bell. The driver said no one was home. No slip in the mailbox. While the recipient was probably at home. On the other hand, DPD has genuine innovations (real-time location tracking, precise delivery windows, app-based redirection) and a 2.2% issue rate in Q3 2024 alongside a first-attempt success rate of 90.53%. Better than nothing, not good enough. GLS is the quiet outlier: lowest issue rate among the main German carriers at 0.47%, fastest transit time at 1.32 days in Q2 2024, rated in the German Institute for Service Quality top three. Nobody talks about GLS.

DHL’s structural answer to last-mile failure is the Packstation: 15,500 automated lockers nationwide, free to use, 24-hour access, planned to reach 30,000 by 2030. It is a good idea. It is also an honest acknowledgment that home delivery at current volume cannot be completed reliably, so the system routes around the problem. The Packstation works for DHL packages. If your parcel comes via Hermes, DPD, or GLS, and given their combined market share a large fraction of German e-commerce does, the infrastructure does nothing for you. You take the Filiale trip, as I often do. Unless the item is large. I don’t drive. My international license doesn’t work here because I live here now rather than visiting. Getting a German license means a Fahrschule, upwards of 2,500 euros, often closer to 3,000 these days. So on the days when the parcel is large and the driver definitely rang the bell: I don’t know what to tell you. I figure it out.

Explaining Japan’s delivery standard as a product of omotenashi, a national cultural disposition toward service, lets the structural explanation off the hook. Yamato’s standard was produced by a commercial decision in 1976 to serve individual recipients when the industry consensus said they weren’t worth serving. The competition that followed locked in a quality floor through market pressure. The driver’s professionalism is downstream of that structural choice. The accountability architecture, the slip with the number and the turn-around-the-truck norm, is culture the system produced, not culture the system inherited.

Germany has excellent infrastructure at every layer except accountability at the final step. The hub works. The regional sorting works. The Packstation works for the people it serves. The driver-recipient relationship has not been redesigned around what the recipient actually needs. And the labor model at the bottom, DHL’s route density targets and Hermes’s contractor arrangements, removes accountability from the driver by making honest completion impossible rather than merely optional.

What a better model looks like is not hard to describe. Getting there is.

First, let’s take Yamato’s accountability architecture: time windows that mean something, same-day redelivery as a default right, a slip that names the driver and gives a number. The recipient should be able to reach the person responsible for the package. Not a bot. Not a queue. The driver.

Next up, Germany’s Packstation infrastructure, made carrier-neutral. The plan to double the network to 30,000 by 2030 is DHL-led and DHL-exclusive. A regulatory standard requiring locker infrastructure in public space to be accessible to all carriers would change the economics for every competitor. The locker solves the recipient-not-home problem cleanly. The current version solves it for about 40% of the market. Estonia routes over 90% of trackable items through carrier-neutral automated lockers. The first-attempt completion rate makes the German numbers look like a rough draft.

Third, make the fallback accessible without a smartphone. A physical card that opens the locker, a magnetic stripe or something similarly passive and durable, should be available to anyone. The system should work better with a phone, not only with one. The delivery slip, when one must be left, should be printed: clear type, unambiguous options, a checkbox for preference on the next attempt. The thermal printers exist. The drivers carry phones that could trigger them. What is missing is the requirement.

This is the piece that requires political will rather than product design. The locker infrastructure, the accountability slip, the accessible fallback: all of that is solvable with money and standards. The labor piece requires someone to say that delivery workers are not an input cost to be minimized, and mean it when volume grows.

Rowland Hill’s central insight was that the rates were structured to seem inevitable, and the structure was keeping ordinary people from using the post at all. He demonstrated it was a choice, not a constraint. The system had spent decades pricing itself out of the population it was supposed to serve, and nobody had stopped to ask why.

The ghost delivery is the same logic at the point of contact. The unreliability looks like an inevitable consequence of scale. It is not. It is a choice embedded in route density targets, headcount decisions, and dividend payments, made by people who also decided the person waiting at home would absorb it. 420,000 direct complaints in a single year is not a 0.003% complaint rate. It is 420,000 people whose Tuesday got restructured around someone else’s quarterly target.

In Mozambique, nobody pretended otherwise. In Germany, the Packstation is elegant, the app is sophisticated, and Deutsche Post will tell you with a straight face that the complaint rate is within normal parameters. The infrastructure is impressive. The last step is not.

When I first moved to Japan and the Yamato driver handed me a lamp from Nitori inside the two-hour window I had selected, I stood at my door a moment longer than the transaction required. It felt faintly embarrassing, that much relief over a parcel. But that’s what happens when the baseline has been wrong long enough.