weekly log · issue 5 · the five clocks of a truck
2 h 34 min in the yard, but not all of it is ours: the five clocks of a truck
Cronos splits every truck's time into five segments and grades the warehouse only on its own: a 1 h 47 min inbound median against a 45-minute target.
- Cronos splits each truck's time into five segments that add up to the total, to the minute.
- The warehouse is graded on three: waiting for a door, at the door and paperwork. Before the appointment and after we finish belong to the carrier.
- No appointment means everything counts against us: when in doubt, the minute is ours.
- With 30 days of real data we're still in the red against a 45-minute target, but now we know where: waiting for a door, and paperwork.
Every warehouse has an argument that never ends: the carrier says we kept them in the yard for three hours, and we say they showed up before their appointment, stayed for lunch after loading, and the delay isn't ours. Both sides are partly right. And as long as nobody can separate the minutes, the argument goes to whoever shouts loudest.
This week I want to explain how we solved it in Silodisa AI with Cronos, the agent that measures every truck. Its job isn't to say how long a truck took. It's to say who owns each minute.
One truck, five clocks
Our warehouse system already stored every truck's timestamps: when it arrived, when it pulled into the dock, when loading or unloading started and ended, when the paperwork was ready, and when it left. With those times, Cronos splits the total into five segments, in order and with no gaps:
- Arrived before its appointment. If the appointment was at 10:00 and the truck arrived at 9:00, that hour belongs to the carrier.
- Waiting for a door. From the moment it's the warehouse's turn until the truck pulls into the dock. Ours.
- At the door: loading or unloading. Ours.
- Paperwork. Outbound only: from the end of loading until the documents are ready. Also ours.
- Stayed after we finished. From the moment the warehouse finished until the driver leaves. The carrier's.
The five add up exactly to the total time on site, to the minute. A watchdog checks that every time the screen reads data: if one day the parts don't add up to the total, a warning shows at the top of the screen.
Warehouse time is the sum of segments 2, 3 and 4. It's the only thing we grade. The total is still shown, but as a reference.
When our clock starts
The most important rule is when the warehouse clock starts. Here's where we landed:
- If the truck arrived before its appointment, the clock starts at the appointment time. But if we put it at a door before then, it starts when it pulled in. Time we gain counts in our favor, never against us.
- If it arrived on time or late, the clock starts when it arrived.
- If it has no appointment, the clock starts when it arrived. Everything counts against us.
That last point was deliberate. If an appointment is blank, set at exactly 00:00, or more than 24 hours from arrival, we treat it as "no appointment." It's the prudent call: when in doubt, the minute is ours. I'd rather have a number that punishes us too much than one that gives us free time.
What 30 days of real data showed
From August 28 to September 26, the warehouse handled 292 inbound trucks and 318 outbound trucks. The numbers:
- Inbound: median total time on site, 2 h 13 min. Warehouse time, 1 h 47 min.
- Outbound: total time, 2 h 34 min. Warehouse time, 1 h 39 min.
- Outbound trucks arrived on average 46 minutes before their appointment.
Our target is 45 minutes of warehouse time to unload or load a truck. We're still in the red there, and far from it: 62 of 292 inbound and 36 of 318 outbound trucks came in within target. I won't hide it: it's one of the numbers that will take us the most work.
But now I know where the problem is. Inbound, almost all warehouse time is waiting for a door: the truck is already in the yard, it's our turn, and there's no door. Outbound, two things weigh: waiting for a door and paperwork. Neither gets fixed with faster forklifts. They get fixed by assigning doors earlier and closing documents while loading.
One truck, taken apart
A real example. The inbound truck on delivery ticket 1730626 took 51 minutes: 12 in the yard, 1 waiting at the door, 17 unloading and 21 closing out, with a single forklift. Unloading, which is what everyone pictures when we talk about "truck time," was a third of it. The rest was waiting and paperwork.
That's why Cronos doesn't just give a number. The screen shows one bar per truck with the five segments in color: gray for the carrier, color for the warehouse, with a dotted line at the appointment and another at the target. By default they're sorted by the most warehouse time.
And in the yard, live, every truck inside has its own clock: green up to 45 minutes, yellow up to 60 and red after that. If it arrived before its appointment, the clock says "before its appointment" and has no color. If we're done, it says "finished, waiting for the driver to leave." That way nobody gets nervous about a minute that isn't theirs.
What we don't know (and say so)
Measuring well also means owning up to what's missing:
- Every timestamp is entered by hand, to the minute. 100% have zero seconds. They work, but they're not an automatic clock.
- The system doesn't record when a door was assigned. We know which door, not when.
- Inbound, there's no record of when stamped paperwork is handed over. So there, the clock stops at the end of unloading.
- Timestamps don't always line up. In 30 days, 7 of 604 trucks had a time earlier than the one before it. One case: unloading until 14:53 and departure at 14:51. Those trucks show "doesn't add up" and aren't graded. We never compute with times that don't line up, and we never make one up.
When a load's start or end is missing, Cronos estimates it from pallet scans (the median is 36 scans per truck), only if it falls between arrival and departure, and the screen marks it "estimated." In those 30 days it wasn't needed once.
And as with everything else, an independent auditor redid the math on its own, without reading the code: 1,238 of 1,238 figures matched.
What Cronos does with it
Cronos doesn't recommend renting forklifts. That was agreed from the start: in 30 days, 547 trucks were handled with one forklift and only 4 with two, never at the same time. The bottleneck isn't equipment.
What it does do, every day, is put the heaviest suggestion in "What do I do today?": if the median wait for a door goes over 20 minutes, it names the two busiest hours of the day; if many outbound loads had their goods staged after the truck arrived, it flags "goods not ready." And if 30% or more of total time belonged to the carrier, it leaves an informational note to discuss with the carrier, with data and without a fight.
Why this matters
BIEN y a Tiempo (done right and on time) isn't measured on the carrier's clock or on ours. It's measured on the right clock for each minute. Once you separate the minutes, the conversation with the customer and the carrier changes: it's no longer "you're slow" versus "you show up early." It's "these 25 minutes waiting for a door are ours, and we're going to bring them down."
We're still in the red on truck times. But for the first time, the red has a first and last name: waiting for a door, and paperwork.
In your operation, who owns the minute a truck arrives before its appointment? I'd love to hear how you count it. Write to me.
Miguel Marengo Canales · Silodisa
Comments
comments · I answer every one. If you disagree, even better: say it with numbers.