The honest ETA: why "before lunch" beats "10:23"

A dispatcher puts 10:23 on a confirmation because the routing engine produced 10:23. The receiver reads it as a commitment, staffs the goods-in door from 10:15, and rings at 10:40 to ask where the truck is. It arrives at 11:05, an entirely ordinary morning, and everyone involved has spent it mildly angry.

Run the same day with before lunch on the confirmation. Same truck, same 11:05. Nobody rang anyone.

The difference was never the arrival. It was what had been promised.

Precision is a claim about your own uncertainty

Every arrival time you publish carries a second, unwritten number: how sure you are. Before lunch says four hours of uncertainty out loud. 10:23 claims one minute of it. If you do not have that minute — and at 06:00, with four stops ahead of that one in the sequence, you do not — the precise number is not more information than the vague one. It is the same information with a false confidence bolted onto it.

That matters because receivers plan against the confidence, not the number. A one-minute promise tells a warehouse manager it is safe to book the dock either side of it and have two people standing there. A four-hour window tells him to keep the morning loose. Only one of the two is true, and the untrue one costs somebody wages.

What an ETA can honestly say, and when

An arrival time is three different claims at three points in the job. The useful discipline is to stop pretending it is one.

At booking. Nothing has been planned. There is no vehicle, no driver, no sequence — often the day is a week away. What you know is whether you have capacity that date and roughly how the lane usually runs. The honest unit is a day or a half-day: Tuesday, Tuesday morning. A booking-time commitment is a promise about capacity wearing the costume of a promise about a clock.

At dispatch. Now the stop has a position in a sequence, a vehicle and a driver, and everything upstream of it is at least decided. This is where two to four hours becomes defensible for a mid-tour stop. The first stop of the day is the exception, and the only one you can nearly promise, because nothing has yet had the chance to go wrong ahead of it. Treat that asymmetry as a feature: sell tight morning slots to the customers who need them, and stop pretending stop seven is the same kind of commitment as stop one.

En route. Position, remaining stops, dwell so far. This is the only point at which an hour — sometimes half an hour for the next stop — is a statement rather than a hope. It is also the most perishable: true until the next gate queue, and not a minute longer.

Notice the shape: the window should narrow as the day commits itself. Putting 10:23 on the booking confirmation inverts that — it is at its most precise when you know the least.

Where the minutes actually go

A routing engine answers a driving question well, then gets blamed for answering a different question badly. The working day is mostly not driving:

• Dwell at earlier stops. Ten minutes over at each of four stops is forty minutes, and overruns are not symmetrical — nobody unloads faster than planned to make up for it.
• Gate queues and booking slots. A distribution centre with a timed slot will hold a truck that arrives outside it, so arriving early can cost more than arriving late.
• The statutory break. Under Regulation (EC) 561/2006 a driver must take 45 minutes after four and a half hours of driving, splittable as 15 then 30. On a long tour that break lands somewhere, and it lands in the middle of somebody’s window.
• A failed delivery. Nobody at the address, the wrong pallet, a refused consignment — the phone call and paperwork happen before the wheels turn again.
• Weather and road conditions, a season in the Nordics rather than an incident.

Each is small alone, and they compound in one direction: a delay is almost never cancelled out by a later leg running fast, because the plan already assumed it would. That is the real argument for windows over points — the same gap between plan and day we wrote about in route planning versus reality.

The phone chain fails on exactly the day you need it

Here is how a slip travels without a system. The driver knows at 08:50, but he is driving, so realistically he calls from the next stop at 09:40. The dispatcher is on another line. The receiver, who has had someone at a dock since 10:15, finds out at 11:00 by ringing the customer, who rings you, who rings the driver.

That chain is slowest precisely on the busy days when things slip most, because everyone in it is busy for the same reason. And the person best placed to notice the slip is the one you least want making phone calls, since he is at the wheel.

The alternative is not that the driver communicates more. It is that he communicates once, as a by-product of work he is already doing — a one-tap status at the stop, a signature, a reason for a failed delivery — and the system tells everyone else. That is also what quietly removes most inbound status calls, which we covered in the customer portal and fewer status calls; position data from a tracker sharpens it further, with the trade-offs set out in GPS tracking options for fleets.

Be honest about the cost: automation broadcasts your bad news faster too. A slip that used to surface at 11:40, once it was over, now surfaces at 09:05. That feels worse and is much better. A customer told at 09:05 can move two people onto another task; one told at 11:40 can only be annoyed.

The discipline of widening

Widening a window is an act, not an admission. The rules that make it work are unexciting and mostly about timing.

1. Widen when the cause is known, not when the miss is certain. The moment the second stop overruns, the fourth stop’s window is already wrong. Say so then, while the receiver can still use the information.
2. Widen once, generously. Three successive ten-minute slips read as incompetence; one honest move to 14:00–16:00 reads as control.
3. Give the category, not the essay. Held at an earlier delivery is enough. Detail invites negotiation about a fact.
4. Never re-narrow to look better. If the day recovers, arrive early inside the wide window — a pleasant surprise. Re-narrowing spends the trust you have just bought.
5. Never let a published window expire in silence. The complaint is rarely the delay itself. It is the deadline passing with nobody saying anything — the moment a customer concludes you do not know where your own truck is.

Measure width and hit rate together, or you get fiction

A thirty-minute window hit 55% of the time is a worse product than a three-hour window hit 96% of the time, and only one of those looks impressive on a slide. A team measured on narrowness alone learns to publish optimism; a team measured on hit rate alone learns to publish sometime this week. Track both, per customer.

Which leads to the trade-off worth stating plainly to anyone pushing for tighter slots: narrow windows are entirely purchasable, but you buy them with vehicles and slack, not with software. A dedicated run with two stops can promise a 60-minute slot honestly; a fourteen-stop distribution tour cannot. What a system can do is stop the fiction being generated, and get the truth to the receiver quickly when the day moves.

navichain is built around that second job. Delivery windows are one of the real limits routes are judged against; stops are planned in sequence, so a driver’s one-tap status update at one stop is what updates everyone else’s view of the next; and when something changes, the right person is told in the app, by push or by email, while the customer follows their own shipments in a portal scoped to their organization. Every truck and trailer with a known position sits on one live map, with GPS telematics available as an optional add-on rather than a second mandatory subscription. The driver app is included on every plan — the platform page shows what is in it, and pricing what it costs.