Part of my workshop PCLS (Planning of Complex Logistics Systems) is a discussion of planning principles. One of these principles is: Define the load carrier early on.
In this article, I would like to provide an example that highlights why this is important. There are many more reasons, but this particular example is based on a recent project. I would like to illustrate with that project what can happen when you switch load carriers halfway through, and all the implications the load carrier has for the system.
Let’s assume the following situation. You are in the process of engineering a goods-to-person picking solution, and your customers are business customers. After order picking, your people will deliver the orders with your own fleet to your customers. That’s a fairly common situation and one that’s easy to visualize.
Okay, so the process is this: We do order picking at the goods-to-person stations. Then the order load carriers are transported on a conveyor to the goods-out area, where they are buffered in gravity flow lanes until the drivers pick them up and load the vehicles for delivery. Very simple process, very common, nothing special.
Let’s further assume that you would also like to introduce a picking robot at one or perhaps several of the goods-to-person stations in the system. And the picking robot vendors will not get tired of explaining to you that you can just switch out the person and put in a robot, and it’s super simple, and you just foresee the space for the later installation of the robot, nothing else you need to do. (There’s much more to this, and the complexity of integrating stationary picking robots is frequently understated and not properly understood. But that’s for a different article.)
We need to keep that robot in the back of our minds because that is what’s making some of the things more complicated here.
Let’s now assume that, for environmental reasons or cost reasons (it doesn’t really matter why for this example) the company building the automated system decides to switch from delivery in cartons to delivery in plastic totes. Okay, so let’s think this through.
The first obvious change induced by the change of load carriers is that we now have a returns process. Previously, with the cartons, you would leave them at the customers and the customers would dispose of them. Now you need to return the plastic totes.
This adds a little bit of effort and consumes a little bit of time in the delivery process. The drivers stack the totes when loading the vehicle, and they also stack the empty totes when collecting them from the customers.
The totes remain with the customer after delivery and are picked up with the next delivery. With many customers, that’s a significant number of totes that remain with customers at any given time. So that’s some tied-up capital. It’s not a big deal, but it’s something I wouldn’t simply ignore. At the very least, you’d like to keep track of the totes and where they went and whether they were returned (i.e., you need to have a function for load carrier management in your software. If you don’t have it, that’s some more money and/or development hours right there).
Also, some customers don’t have enough space inside their stores to keep the totes, or they simply don’t want to keep the totes inside their stores. So they keep the totes outside until they are picked up with the next delivery.
Which means that it can rain into the totes, leaves can fall into the totes, dirt can blow into the totes, and insects or other animals can get into the totes and will then be brought back to the warehouse.
Back in the warehouse, upon arrival after a delivery tour, the totes have to be unloaded and brought into the warehouse.
Now we’ve got stacks of totes that we have to buffer somewhere.
Since we want to reuse the totes, they need to get back onto the conveyor. Now, you don’t want to convey stacks of totes to the picking stations; you want to transport individual totes to the pick stations. So there needs to be some unstacking.
However, unstacking the totes and buffering them individually on the conveyor takes much more conveyor length for the same amount of buffer capacity than buffering stacks. You also don’t want to keep stacks of totes on the floor and feed them in the conveyor as totes are being consumed by the picking stations since that would require a person on stand-by. And you don’t want your delivery drivers to spend time unstacking the totes.
So we move stacks of totes onto the conveyor through a simple lift that picks up tote stacks from floor level and lifts them to the standard conveyor level. Then we buffer some stacks of totes on the conveyor before we run the stacks through an unstacking machine to separate the totes and bring them to the picking stations.
So far, so good.
So now we’ve added a returns process, we’ve added a machine for unstacking, and we’ve added some meters of conveyor for stacked as well as unstacked totes as a consequence of switching to reusable totes.
The next problem we encounter is that all the totes are labeled. A simple label is applied to each tote, showing the customer’s name and address so that the driver can see at first glance which tote belongs to which customer. Since each order contains multiple orderlines and there are frequently multiple totes for each customer, each customer order tote is numbered as “tote 1 out of X” so that both the driver and the customer can keep the totes together and know that the order is complete when delivering the totes. Yes, you could live without labels and track everything digitally, but simple labels are by far the most ergonomical solution.
Now, the cartons would be labeled too. But that wouldn’t matter because the entire carton remains with the customer, and the customer can decide to reuse the carton in some way, in which case it’s his problem that his name is on the label. Or he can remove the label.
But with the totes that are being returned, you have to remove the labels. Of course, you can have the driver remove the labels upon return, but then you have to unstack all the totes, remove the labels, and that takes time, and… well, stack them again for better buffer density. It can take hours every day, hours of removing labels, which is not a value-adding activity.
Of course, you can also delegate the task of removing labels to the pick stations, but that wouldn’t be very smart because the pick stations are designed to be the bottleneck. You want the people at the pick stations to be busy picking, not doing extra work like removing labels.
Moreover, let’s remember the picking robot. The fact that we are planning for picking robots now comes with important implications.
First of all, the picking robot cannot remove the label from the tote.
Second, the picking robot is dependent on receiving totes that are clean and ready to be processed. At manually operated pick stations, if a tote arrives and there’s a leaf in it, or it’s dirty or wet, the picker can simply take the leave out or remove the tote and take another one. The robot can’t simply do this.
So now we have to inspect all the totes that are being fed into the system.
In principal, the driver could inspect the totes upon return, but that would take a long time and he would have to unstack the totes. But again, unstacked totes on the conveyor take much more space. So we want stacks of totes on the conveyor until they are separated for transport to pick stations.
Which means there needs to be some automated inspection of the totes after unstacking, which is another problem that we previously didn’t have. Trusted colleagues with expertise in robotics and computer vision are telling me that’s not a problem, but then again there is no simple and cheap off-the-shelve solution available. So we’ve got to figure this out. Doable, but needs to be done.
Okay, so now we have two more additional processes and two more additional technical requirements that require machines of some sort. The totes need visual inspection through a camera system and some machine-learning logic. And we have to move dirty or otherwise unsuitable totes out of the system onto a separate “not-okay” lane on the conveyor. And we have to have a machine that removes the labels.
Now, it turns out that machines for removing glued labels from totes are not readily available, either. There are machines that can insert labels into a little pocket on the tote, if the tote has one, and there are machines that can take the label out of that pocket again. Companies like Caljan used to offer this kind of setup (though they closed their label printing business in early 2026).
But — and that’s a big but — if the customer has purchased reusable totes without pockets, then there’s no pocket. And then there’s no off-the-shelf machine for that process to be automated.
So we need to build our own, or use a simple robot in combination with a non-stick surface on the tote to remove the labels. It’s not rocket science, but it’s another engineering project.
So already now you can see that there’s a lot of extra work that comes with totes — especially with the additional introduction of picking robots — none of which you would have if you used cartons.
And yes, you would create a lot of cardboard waste. And cartons aren’t cheap, either. But considering all the extra processes, process costs, technology, that are now needed to take care of the totes, and the possible points of failure that come with it, I would think that most definitely cartons will be the more economical solution.
Well, we will still have some cartons in the system. So it’s not like the entire system is switched over to totes. We will have both totes and cartons. And while totes would make up the largest share of the throughput, there would still be 20 to 30% cartons in the system. Which means we still need the carton erector. We still need a carton closer. These cost components don’t disappear. We just add more cost components for tote handling.
And since the totes have lids that are mounted on the long sides, and the lids need to be closed, we now need a bypass at the lid-closing station because cartons shouldn’t travel through the lid-closing station. Or we need two routes: We send the totes through the lid-closing station, and we send the cartons to the carton-closing station, provided the cartons are not closed by the pickers at the pick stations. But latest with the introduction of the picking robot, closing cartons at picking stations goes out of the windows because the robot can’t do this. Therefore, you either need automated carton closing or you limit the robot to working on tote orders while carton orders are processed entirely at the manually operated picking stations.
Now, in the project this example is based on, fortunately the customer decided to use plastic totes from the very beginning. This was not a change made late in the engineering phase. However, the customer selected and purchased the totes before the system itself was planned, which came with some implications and complications for the removal of the labels and the unstacking of the totes. But clearly, if the switch to totes had been made late in the planning process, this would have become a very troublesome project.
None of this is a deal breaker. But I think the example illustrates nicely that the choice of the load carrier can have a significant impact on process design, cost, and the technical complexity of the solution.
This becomes especially relevant if you then add another technology — say, a picking robot — which is very limited in what it can do. It picks, and that’s it. It does not perform any of the other manual operating steps that a human picker would otherwise be able to perform, such as removing a label if you wanted to, closing a carton, inspecting a tote, or removing a dirty tote from the system.
And if you make the change to another load carrier later in the planning process — let alone after the planning process has been concluded and suppliers have been contracted — you are in trouble. The change is not simply going to work within the existing concept. You may have to go back and redesign substantial parts of the process and the system.
Therefore, define the load carriers early on. And be aware that a process with a different load carrier may not, and probably will not, look the same. Additional process steps and technical modifications will most likely become necessary.
This article is based on an example from my workshop, Planning of Complex Logistics Systems. If you’re interested in the workshop, check out the description, and don’t hesitate to reach out if you’d like to learn more.
