00:00:00 Welcome & Housekeeping
00:01:30 Evolution of logistics and supply chain terms
00:03:31 Emergence of software in supply chain
00:05:30 Logistics execution vs. supply chain decisions
00:08:09 Modern logistics: software provides routes
00:10:25 Supply chain vs. operations distinction
00:18:08 Mainstream conception of supply chain vs. logistics
00:23:11 Companies failing to distinguish supply chain and logistics
00:25:57 Supply chain optimization and automation
00:28:50 Future automation of truck drivers
00:31:28 Air France example: large-scale automation investment
00:33:45 AI automation: misconceptions and realities
00:36:29 Logistics: cost reduction through automation
00:41:20 Aerospace example: financial opportunity in decisions
00:45:01 Potential conflict: logistics vs. supply chain excellence
00:47:21 Cost of downtimes in various sectors
00:52:05 Bill of resources: parts, people, tools
00:54:27 Importance of automation in supply chain
00:57:52 FIFO: not always financially optimized
01:03:55 Mechanization progress in logistics
01:05:20 Disappearance of blue collar jobs: distant future
01:08:39 E-commerce example: blue collar vs. white collar
01:12:12 Mechanize decisions for capitalistic investment
01:14:07 Conclusion and final thoughts
Summary
Conor Doherty and Joannes Vermorel delve into the distinctions between supply chain and logistics. Joannes traces the historical evolution of these terms, noting that logistics, originally a military concept, focuses on execution, while supply chain management involves decision-making. The advent of software in the late 1970s further separated these roles, with logistics handling the implementation of decisions generated by supply chain algorithms. Joannes illustrates this with examples like route optimization and the traveling salesman problem, emphasizing that modern supply chain management relies on dynamic, real-time tools to enhance efficiency and responsiveness in operations.
Extended Summary
In a recent episode of LokadTV, Conor Doherty, Head of Communication at Lokad, engaged in a thought-provoking discussion with Joannes Vermorel, the CEO and founder of Lokad. The conversation centered on the critical distinctions between supply chain and logistics, a topic of increasing relevance as automation continues to transform industries.
Joannes began by tracing the historical evolution of the terms “logistics” and “supply chain.” Originally a military term from the 19th century, logistics referred to the management of troop movements, shelter, and supplies. This concept was later adapted for civilian use, particularly in the context of operations research after World War II. Over time, the term “supply chain” emerged to describe the broader, more complex decision-making processes involved in managing the flow of goods and services.
Joannes emphasized that while logistics focuses on the execution of decisions—often involving blue-collar workers—supply chain management is concerned with the art of decision-making, typically handled by white-collar professionals. This distinction became more pronounced with the advent of software in the late 1970s. Before software, supervisors made all decisions, but the introduction of software allowed for more complex and dynamic decision-making processes, leading to a clear separation between supply chain and logistics.
Conor pointed out that even within logistics, software plays a crucial role, prompting Joannes to elaborate on the nuances. For instance, while a logistics director might oversee truck drivers and ensure vehicle safety, the actual route optimization is a function of supply chain management. The logistics team executes the decisions generated by supply chain algorithms, which are designed to optimize routes, load trucks efficiently, and ensure timely deliveries.
Joannes further illustrated this point by discussing the traveling salesman problem, a classic optimization challenge. In modern supply chain management, software solutions handle such complex problems, providing logistics teams with pre-determined routes and schedules. This division of labor allows for more efficient and effective operations, as logistics personnel focus on execution while supply chain professionals handle the analytical and decision-making aspects.
The conversation also touched on the role of software in dynamic decision-making. Joannes highlighted how real-time tools like Waze can suggest alternative routes based on current traffic conditions, exemplifying the kind of automated decision-making that characterizes modern supply chain management. This capability ensures that operations remain flexible and responsive, reducing the likelihood of errors and inefficiencies.
In summary, the discussion underscored the importance of distinguishing between supply chain and logistics, particularly in an era of increasing automation. While logistics is about executing decisions, supply chain management involves the complex, often automated processes that generate those decisions. This separation allows companies to leverage specialized skills and technologies, ultimately leading to more efficient and effective operations.
Full Transcript
Conor Doherty: Welcome back to Lokad. Today I’ll be talking to Lokad founder and CEO Joannes Vermorel about the critical differences between supply chain and logistics. As you’ll hear, this is a very important point for companies, particularly as the world trends in the direction of increased automation. Now, as always, if you like what we do at Lokad, consider subscribing to the YouTube channel and following us on LinkedIn. And with that, I politely invite you to sit back, relax, and enjoy the conversation.
So, Joannes, welcome to the new studio, the Black Lodge. How do you feel?
Joannes Vermorel: It’s quite nice. I mean, fun fact for the audience is that it’s the first time that we are not actually sitting in the kitchen just in front of a couple of appliances such as like two fridges, series of microwaves and whatnot. So, exactly, we are for the first time, we have our own private space. That’s very nice.
Conor Doherty: That’s actually quite fun. Now, if you think about that in terms of little Easter eggs, if you go back over the seven years, there have been a few heated exchanges between you and some guests. So the context being that directly behind the camera there are people making their lunch, people making a coffee, etc., etc.
Well, in any case, the studio is not finished yet, but we can’t let that get in the way of important business, which brings me to today’s topic: supply chain does not equal logistics. So, Joannes, 30,000-foot view, why are we here?
Joannes Vermorel: The terminology is quite complicated in the field. The thing is that what people have been calling logistics until probably, let’s say, the 1970s is what people are now calling supply chain. So there has been some gradual evolution of the meaning of the terms. We had logistics emerge as a military term in the 19th century, and it was actually a French word that refers to the specialty of the LOI. That was literally a military term, and at the time, the problem was to find shelter for your troops.
It was theorized by two generals, one French and one from Switzerland, and they theorized the idea of organizing troop movements and taking care of shelter, food supply, and whatnot. That was pretty much the start of large-scale organization, synchronization of large organizations. Fast forward a century and a half later, that was pretty much logistics. Then we had operation research as a field, which became very important, I would say, in the aftermath of the Second World War. Out of that emerged supply chain and logistics, which took different paths.
So, when you say what those terms mean, it really depends on which decade you consider. Nowadays, if I have to really summarize, supply chain is the art of decision-making. So it’s really about making decisions, and logistics is really about the execution of those decisions. There is really a split: supply chain deals with white-collar people and decision-making processes, while logistics deals largely with blue-collar people and makes things happen once the decisions have been made.
Conor Doherty: Historically, at least up until very recently, supply chain and logistics were treated as more or less synonyms. So what led to that divergence, as you said, with supply chain focusing more on white collar and logistics blue collar?
Joannes Vermorel: It was primarily the emergence of software. Until, I would say, the late ’70s, the only entity capable of taking any kind of decision was a person. That would be the supervisor who was supervising the people doing the stuff, who would also call the shots and make the decisions. So, until the ’70s, the idea that you would segregate the two didn’t really make sense. But as soon as you started to introduce layers of software, the separation became more and more obvious.
First, the complexity grew enormously. Since the late ’70s, supply chains have probably multiplied the amount of product references and variants and everything by a factor I would estimate probably of 100. The situation is vastly more complex than it was 50 years ago. Again, that is really software making that possible. We now have warehouses that can contain up to 100,000 distinct articles. This is a much bigger complexity than what used to be.
As a result, dealing with the complexity and dealing with the decision-making processes became a skill of its own that is very analytical, very data-oriented, where people are using instruments, even if the instrument is as crude as, let’s say, an Excel spreadsheet. An Excel spreadsheet lets you deal with thousands of products, and it requires some specialized skills as well.
That’s why there was a divergence between the analytical skills, what we are calling supply chain today, which involves forecasting, setting inventory parameters, and whatnot, and pure execution. Pure skills are around the physical execution, which is, for example, managing the truck drivers, making sure they are on time, making sure that nobody’s drunk, making sure that everybody is driving safely, etc. So the two things really took different paths, I would say, in the world.
Conor Doherty: True. That said, if you operate purely in the logistics space, I mean, you still use software, there are still decisions that have to be made. So could you drill down a bit on that distinction then?
Joannes Vermorel: So if you look at, for example, logistics and you look at the route that the trucks would follow, even if some companies still think of that as a function of logistics, I think of that as a function of supply chain. See, the logistics director, the person who is supervising the truck drivers, who is making sure that the trucks themselves are in good shape, that they are safe, etc., they get a piece of software that will give them the route that they need to follow, and this is it.
So you see, this is not from the logistic director that you would expect a certain refinement from the route optimization algorithm. You see, the fact is, as I said, Logistics executes the decisions that have been generated for them, and thus the generation of those decisions belongs to the realm of supply chain. Then, the execution is the realm of Logistics. So yes, there are decisions, but I would say the logistic director doesn’t call the shots in the algorithmic optimization that goes into establishing the route. If this route is ineffective, they would probably request some other party to deal with it; they would not be dealing with it themselves.
Conor Doherty: Well, that reminds me of a discussion we had not so long ago with, I think, Meinolf Sellmann, where we talked about the traveling salesman problem. So to take that very concretely, if you’re talking about optimizing routes, I’ll let you explain better than me the traveling salesman problem. Can you delineate the supply chain decisions there and where it ends, and Logistics picks up the slack in terms of decisions?
Joannes Vermorel: So Logistics doesn’t pick up the decisions. There are no decisions to be made on the logistics side. Decisions are already made; it’s purely about execution. That’s a modern take. Fifty years ago, people would not have analyzed a problem like that. From the logistics perspective, you already have a piece of software, given to you by some third party, that gives you the routes. It’s a given. Also, what you should put into the trucks is a given, and it’s up to this third party to make sure that when they suggest putting something in the truck, it fits. It’s also up to those third parties—there can be multiple ones—that if they give you a route, the route is correct, and the suggested time frame is also feasible, etc.
You see, the perspective would be that supply chain deals with all the decisions, from short-term decisions like which route to take next to long-term decisions such as capacity projections for the next five years. It’s just a matter of time horizon. But all of that, from the short-term to the long-term, are pure analytical processes. So it’s stuff that can happen in a piece of software, irrespective of the actual execution. Naturally, the models and calculations need to be adequate with regard to real-world constraints, but nevertheless, it’s really the decision plane, and Logistics is really the execution plane. You’re given a route, and now there must be a driver fit for driving the truck and commencing the route.
The same applies to intra-Logistics inside a facility. And when you look at this separation, the more meaningful distinction nowadays is most likely Supply Chain versus Operations. Operations involve the person who supervises all the blue-collar jobs of the company. Logistics is a class of those jobs, but you also have others, such as production workers who operate machinery in a static position, as opposed to moving things around.
Conor Doherty: I don’t want to jump too far ahead, but you did provide me with a nice transition for a point I want to hammer down on. You talked about the physical constraints. So you have a route, and Logistics makes sure the driver is there or picks the driver that will execute that step. That sounds similar to what companies like Lokad do when it comes to scheduling. You take parts, tools, and people, and decide, for example, to put this part over there with this tool, and have Joannes do it because he has the right accreditation, skills, and availability. All of that is supply chain decision-making, which we provide.
Joannes Vermorel: Yes.
Conor Doherty: So, where does Logistics fit into that? Because it sounds like Supply Chain’s done it all.
Joannes Vermorel: No. You see, if we go back to the pre-software world, the only person who could make those decisions was the supervisor on-site, close to the person executing the decision. In that situation, you couldn’t split the responsibility. The person supervising the truck drivers was also calling the shots. It’s only because we now have software, which is networked, that we can distribute decisions.
Your supply chain may be distributed across many locations, but layers of software connect everything. Geography becomes irrelevant because the speed of light is fast enough to transmit information almost instantly. With networked software in place, you can decouple monitoring the execution of tasks and making the decisions. Supply chain makes the decisions, including anything planning-related. Deciding what time your truck drivers should arrive, how many you need, what you should load into the trucks, or if you need special equipment—all of that is supply chain.
What is not supply chain is ensuring people don’t kill themselves with forklifts, that equipment is used appropriately, that employees aren’t sick, and that morale is good. These are process-oriented tasks, not decisions. For instance, the speed limit of a forklift in a warehouse is a decision made once as an engineering matter. It won’t change for the duration of the warehouse’s operation. That’s Logistics’ domain, but I wouldn’t call that a decision. It’s just an established process that doesn’t require ongoing decision-making.
Conor Doherty: So, to collapse all that, supply chain is subject to great uncertainty. Is your position that Logistics is not?
Joannes Vermorel: Yes, in the sense that people say, “Oh, but there is so much uncertainty, and things vary so much.” Yes, conditions vary, and the planning given to you varies. However, the way you’re supposed to execute does not vary. How to drive a truck safely doesn’t depend on the delivery. There are safety policies in place, like speed limits and braking rules, that remain invariant. These operating processes are always the same, regardless of the day’s plan.
What requires specialized skills and whatnot, but I’m digressing. The point is, the challenge in logistics is to maintain complete adherence to your policies at all times. That’s the main challenge. And this is very difficult. The challenge with supply chain is to come up with good decisions that are adequate despite the fact that everything is constantly changing. So, these are very different perspectives.
Conor Doherty: This makes sense as you describe it to me, but what I’m curious about is how radically does your position here differ from the mainstream conception of supply chain and logistics decision-making processes?
Joannes Vermorel: I think companies have been gradually converging to this sort of understanding over the last two decades. The process was very much empirical. Companies realized that as supply chain became more prominent, it brought with it an increasing number of software tools. Even spreadsheets are used to extract data from many systems. So even with crude analytics done with spreadsheets, you still have tons of instruments.
The reality is that if a logistics director spends a lot of time on the ground in warehouses, they may not develop the skills required to crunch all this data and develop analytical skills. Companies, empirically, realized they needed people more focused on analytics. Conversely, those doing analytics often had very few people to manage, especially blue-collar workers, which is a completely different skill compared to managing white-collar workers in a clean, safe office.
Companies have gradually separated blue-collar management under logistics directors and white-collar management under supply chain directors. However, there is still some confusion in companies that haven’t fully removed analytical tasks from the logistics director’s responsibilities, leaving them to handle analytics they aren’t suited for. Instead, all analytical decisions, from the short-term to the long-term, should be under the supply chain director. This includes everything from long-term objectives to real-time decisions, like those made in milliseconds when driving robots in an automated warehouse.
Conor Doherty: You touched on digitalization and the software skills required. What would be the difference between a supply chain director and a logistics director in terms of computer savviness?
Joannes Vermorel: My view is that a logistics director can know almost nothing about computer systems. They just need to be savvy enough to read the planning and other basic performance indicators. But they won’t be expected to program or handle anything more complex than basic percentages to monitor team performance.
In contrast, a supply chain director holds a highly analytical position. While it’s still possible for someone without programming skills to hold this role, I believe that in the future, programming will be a basic requirement. If you want to do non-trivial analysis and crunch numbers, you’ll need to know how to program.
Conor Doherty: Are there any examples of companies that haven’t implemented the kind of distinction you’re talking about? Or companies that treat supply chain and logistics as synonymous or as one department?
Joannes Vermorel: Yes, it’s still frequent when we talk to prospects. Some old-school companies still use the term “logistics” when the logistics director is essentially a de facto supply chain director. The problem is that this person ends up with divergent job requirements—managing teams of blue-collars on one hand and refining forecasts on the other, which is too demanding.
In other cases, logistic directors may change their title on LinkedIn to supply chain director, but their skills remain mismatched. Many companies now have both a logistics director and a supply chain director, but they haven’t fully rearranged responsibilities. Some short-term decisions, like route optimization, remain under the logistics director, even though they should be handled by software-savvy teams in supply chain.
The correct way to organize a company, in my view, is to group skills that make sense together. Supply chain decision-making requires people who are very savvy with software, while logistics is more about managing people. These are very different mindsets.
Conor Doherty: When we talk about supply chain optimization, a lot of that, at least for Lokad, rests on automation. If you embrace that concept, there’s theoretically an upper limit to how much you can optimize supply chain decision-making, but it’s very high because automation eliminates manual processes. On the other hand, logistics, as you’ve said, is almost exclusively a physical enterprise. To what degree can you optimize logistics compared to supply chain optimization, financially speaking?
Joannes Vermorel: Mechanization of blue-collar work has been ongoing for two centuries, maybe three, but it progresses much more slowly compared to software advancements. In the last decade, warehouses have become increasingly robotized, but it’s a slow process. It started with warehouses handling a limited diversity of small products because they were easier to automate, as well as painful environments like frozen food storage, where nobody wants to work all day at minus 20°C.
This process will likely take 40 years from start to finish, with the starting point around the early 2000s. Automating truck drivers is another area that hasn’t really started yet, but it will happen. My guess is that large-scale automation of truck driving will begin before the end of this decade, but it will take another two or three decades to complete due to the complexity involved. You can remove the driver, but you still need someone to load and unload the truck.
That’s a problem that can probably be solved because at some point there will be systems automated to load and unload the trucks. So you see, and the same thing has been happening in factories. The stuff that was easy to automate has been automated decades ago. So what remains as manual tasks are the stuff that is quite difficult to automate. So you see, this is the situation. So I think on the logistics side, the goal is still to automate everything, and the process will continue probably for most of the 21st century. It’s still ongoing, but the pace is very much set, and people expect it. It will keep progressing a few percent per year for the foreseeable future.
So I would say this is a given, and people expect that. There is no grand surprise, and again, it has been steady for a very long time. Nobody’s surprised anymore when some stuff gets automated. Everything gets gradually automated, and sometimes, for example, a warehouse gets replaced by a new one, and you need 10 times fewer operators. But on a grand scale, the process is slow and steady.
What on the software side, which affects more of the supply chain decisions, I think the situation is very different. Unlike the physical space, it is much more a matter of technology rather than a matter of upfront capital investment. One of the reasons why not all warehouses are being robotized immediately is because the activity is extremely capital intensive. I mean, we’re talking about hundreds of millions of euros to make a large warehouse completely automated.
Conor Doherty: You mentioned an example before to me off camera about Air France, the one roof something.
Joannes Vermorel: Yes, for example, Air France Industries has a one roof initiative where they want to connect essentially two large buildings to make sure that all their MRO activities can be carried out in a large unit. Having just one roof simplifies everything because that means parts are never exposed to the outside, getting cold, stuck, falling off, whatever.
Conor Doherty: Yes, plus there are tons of processes. As soon as a part leaves your facility, it needs to be readmitted with very strict criteria. So it just complicates everything. It’s easier.
Joannes Vermorel: But yes, if you want to invest in connecting two buildings that are already very large and want to add something like a 200 million euro roof, we are talking about tens of millions of euros of investment just to make that happen. Things take a lot of time, and companies, even if they are inclined to do it, their resources force them to pace themselves. Most of our clients in aerospace nowadays are gradually investing in automated warehouses, but it takes time because it’s very expensive.
Unlike, let’s say, Amazon, the return on investment is not as spectacular as it is for an e-commerce business like Amazon. So it takes time. In software, the thing is that until the technology is there, people struggle to automate at all. Once it is there, deployment can happen much faster because there is not that much investment to be made. Yes, there are investments, but they are inconsequential compared to what needs to be done on the physical front.
Conor Doherty: Yes, this is going back a while. It was last year. I can’t remember what article we were talking about, but we discussed the difference between automation in software and how quick and rapid that can be once the technology exists. If it’s software-based, it can proliferate quickly compared to just getting a robotic hand that can replicate the dexterity of a human hand. That’s still difficult, if not yet discovered. People have a misconception about AI automation—it’s everywhere. In certain sectors, yes, and in very specific areas, yes. For example, software that generates decision-making, yes. But the ability to nimbly load, unload, tie knots, things like that, not yet and probably not for quite a while from what you’re saying.
Joannes Vermorel: Yes, if you want to manipulate objects, we already have tons of technologies, but they all have their limitations. You have systems that are extremely fast and precise, but they’re not adaptive. So the part needs to be in the exact correct initial position. That’s what you have in the automotive industry—robotic arms that are extremely fast and precise but not intelligent. The input of the robot needs to be perfectly placed.
Then you have systems that can deal with uncertainty, but they are slow and not very strong. All of that is progressing gradually, but when you do the math, people are still cheaper. Every year, the spectrum of operations where machines are cheaper extends. That’s exactly what I was describing with blue-collar work being gradually mechanized. The process is still ongoing.
For example, France is still losing about 1% of its farmers every year, and the production of food in France is growing by about 1% every year as well. So every year, we have 1% fewer people, we produce 1% more, and we do it with 1% less land. When you look over the course of a century, that’s enormous progress, but it’s slow and steady, and nobody expects any kind of major breakthrough.
Software is very different, and yes, innovation can, I would say, proliferate much faster because the amount of investment is much lower.
Conor Doherty: Speaking of investment in terms of gauging the return on investment, if you invest in your supply chain decision-making software, you can use certain financial metrics to determine whether or not that is having a positive impact. If logistics is a purely or at least predominantly physical enterprise, how do you gauge the impact? Do you use the same metrics? Do you use financial return on investment for logistics and for supply chain, comparing the two? So, logistics is… how do you know it’s getting better? Excuse me, let me rephrase. How do you know it’s getting better?
Joannes Vermorel: Yeah, how do you know it’s getting better? So, logistics, the idea is that you have a mission that is given to you, and it is not acceptable to accomplish the missions in ways that would endanger people. It’s a no-go. So, you have the mission that needs to be accomplished with full compliance to, you know, whatever would pass as sanity. And now it is just a matter of cost. Can you do it cheaper? That’s it.
If somebody orders a product online, the question will be, how much does it cost you to ship this product from your warehouse and have the product arrive on the doorstep of the client within this time frame? So, the progress in logistics is really about cutting cost through automation. That’s it.
Through supply chain, the question is much more open because, again, supply chain is not… it is a very open-ended challenge. There is no upper limit on how much you can improve your decisions. That’s what I said earlier. It’s a completely different game in the sense that, for example, how many variants should you introduce? That would be a supply chain question.
You know, you have a product, you can have more variants to please more people, but is there any limit on the number of variants? Well, every variant that you introduce creates some extra overhead, and there are diminishing returns, so there is a balance. But the amount of questions that can be asked, like the prices, whether you should steer them up or down, etc., is open-ended. I’m not saying that there is no absolute limit in what you could expect from better supply chain decisions, but it is something where there is no clear limit in how far you can go.
And the questions are much more open-ended. Fundamentally, you can start considering things that you were not considering—more suppliers, more alternatives, more options, more price schemes, and whatnot. There is no clear limit. Again, in the realm of logistics, the missions that are given to you are much more narrow and closed. If the goal is to move a piece from point A to point B, this is it. You can do it cheap.
But ultimately, as part of the game of logistics, changing completely the strategy of the company itself, of rethinking, for example, the way you’re going to deliver stuff to your clients, is not part of the game of logistics. An example of that would be, let’s imagine you have a retail store in fashion. So, you have a fashion store. You could, for example, consider that as a matter of supply chain.
Say if someone, instead of buying the last unit of the store, you give them a discount, and they will receive the article sent to them via classic e-commerce. So, imagine you have a person that visits the store, but when this person is about to take the last unit that is left in the store for a given article or size, instead of having this person go on their way with the article, they get a discount to just have this unit sent to them.
Conor Doherty: Why would you do that?
Joannes Vermorel: Well, you could…
Conor Doherty: Doesn’t that look like a stock-out?
Joannes Vermorel: Yes, exactly. So, that would be a way to mitigate the stock-out and also potentially increase the assortment, because then you could afford to keep much fewer units. You see, that would be the sort of thing supply could explore. But from the perspective of logistics, this is not the game that is being played. The decisions have already been made. It is about executing what is given to you.
Conor Doherty: I like that theoretical example you gave, and it actually reminded me… I just jotted it down while you were talking. I think it was you who mentioned it in one of your lectures, or maybe I heard it from a supply chain scientist. It was an example from an aerospace client of how they received their daily recommendations of, you know, buy this, buy that. I’m going to simplify it—it was buy these two engines. It wasn’t as enormous as that, but just buy these two parts.
And it was flagged as an incorrect decision. Why would we do that? We don’t need them. And the algorithm had generated that decision because the price to buy those engines new had dropped below a certain point, making it economically rewarding to hold on to them to resell at a later point. So, the decision was not of need.
Joannes Vermorel: Yes, it was one of financial opportunity.
Conor Doherty: Yes.
Joannes Vermorel: And that’s exactly what happens when aircraft get dismantled. You can have a few extra parts or tons of extra parts that flood the market, and temporarily, mistakes are being made by your peers. Something that should have been sold at the price of, say, 100 is sold at the price of 30, and that’s an accident. So, you grab the opportunity because you’re immediately making a profit.
So, yes, it’s open-ended. Open-ended supply chain is a very open-ended job. That’s also why, for example, when we go back to the difference between a director of logistics and a director of supply chain, when you’re playing a super open-ended game, there are some qualifiers when you’re thinking about employees that don’t really apply.
For example, excellence. In the world of logistics, excellence is clear. You want to be fully compliant with your process. If you do that, you won the game. You’re excellent. That’s it. But in the supply chain world, excellence is ill-defined. It’s so open-ended that how do you know you’re even close to the best you could do?
That’s why teams that focus on excellence make sense if you’re very operational with blue-collar workers because if they do everything by the book exactly as it’s done day after day, congratulations, you’re perfect. We cannot expect anything more from you. But in the supply chain side, this is a very different game. It doesn’t make sense to congratulate people as if they had achieved perfection.
Yes, you can congratulate people, no problem, but because this game is completely open-ended, every success is just another milestone for the next one that will be even better. That’s why it’s a very different perspective. It doesn’t make sense in the supply chain world to have an employee of the month, for example.
Employee of the month only makes sense if you have clear targets where it is possible to be Mr. Perfect and have done everything exactly as asked. In supply chain, no, it doesn’t make sense.
Conor Doherty: On that note, the way you describe that, again, if I were to summarize, in logistics, there is a theoretical perfect game. You can play the perfect game, no errors, no mistakes, nobody dies, as you might say. But it occurs to me then that there is an inherent tension between the pursuit of excellence in logistics and the pursuit of excellence in supply chain.
For example, I give you a schedule for production. Here’s a production schedule—you need this, you need that, that person at that time, go there. And then logistics says, well, actually, that machine has downtime. I need to repair that because I want to maintain my perfect game in terms of safety. If I let you proceed with your production, which is a supply chain mechanism, it might negatively affect my logistics record of safety and maintaining protocols.
So, is there not a conflict there where the pursuit of one kind of comes at the expense of the pursuit of the other, or can…
Joannes Vermorel: No, not really. I mean, okay, the supply chain needs to have decisions that factor in all the constraints of the real world. That is an immense challenge because, for example, the state of repair or disrepair of the machines needs to be taken into account, and that information might be missing from the systems and whatnot. Fundamentally, this is the job of the supply chain to accommodate that. And if they don’t know the state of repairs of a machine, they need to have the decisions with some sort of buffers to take into account the fact that there will be unplanned overhead and generate a planning that can still be feasible by logistics or other teams of blue-collars once they discover gradually all those overheads.
But you see that we are talking about how the plan must be feasible, and that takes into account planning for uncertainty. Again, that’s very much supply chain. Now, from the logistics perspective, they would say in their case the calculation is very different. For example, we have a machine that breaks down one day per year. What will be the cost to have a machine that breaks down one day per decade? Maybe it’s not worth it, or maybe you could have a second machine as a spare. There might be some calculation involved, but we see that in terms of order of magnitude of decisions, it’s much lower.
And if you start having fancy budget allocations and whatnot, I would say it goes back to supply chain. We’re talking about a complex decision-making process that needs to be tackled from a very analytical perspective. Again, we are back to supply chain. The compliance from the logistics side would be, “Do we operate the machine in a way that doesn’t generate premature breakdown?” And if they do, they comply, and they have done their job excellently.
Conor Doherty: It’s pure coincidence because I read this recently for something else I was writing. So I’m a bit familiar with the actual cost of downtimes in certain sectors. There was a report by Siemens called The True Cost of Downtime from last year, and it estimated, depending on the vertical, the cost of downtime. On one end, you had fashion or FMCG, where the cost was about $39,000 per hour. On the extreme end of that spectrum, it was north of $2 million per hour in automotive if there were unplanned downtimes.
And because everything is interdependent, like if one thing goes down, it affects production elsewhere, causing a knock-on effect. It’s not isolated. You have direct and indirect costs. So when you talk about how supply chain can factor in the cost of a machine that might or might not be down, how do you reconcile or factor in the potential enormous financial loss of unplanned downtime into an optimized schedule when you’re getting close to a situation where it’s possibly just time for repair?
Joannes Vermorel: The technical term is stochastic optimization. Stochastic optimization is just optimization under uncertain conditions. That’s why it gets super technical. And that’s why I think we need to segregate an analytical position, like a director of supply chain, from a non-analytical position, such as a director of logistics. It is already very difficult to do the sort of data crunching that can be demanded from the director of supply chain.
The idea that someone on the logistics side, who deals with blue-collar workers, has to also handle very elaborate optimization techniques like stochastic optimization to account for uncertainty is not a very reasonable proposition. It also means we need to have a very broad view of what a decision means from a supply chain perspective. A decision might involve alternative plans for every breakdown situation. It’s not just “I give you a schedule, and you’re done.” It could be “I have a system engineered by the supply chain team, and if at any point something goes wrong—a machine breaks down, an operator is sick, a part is missing, or there’s a defect—then it gives me an alternative path to follow.” This is scheduling optimization.
Sometimes, even with stock, let’s imagine a warehouse that must serve a series of stores, but there isn’t enough in the warehouse to serve all of them. The remaining stock in the warehouse is too low, so you know there will be stockouts in many of those stores because there’s not enough inventory. Nevertheless, you need to split the inventory somehow. Should you send the bulk of the inventory to one store, or should you evenly spread it, or do something else? You need to accommodate situations that are slightly defective, and that is part of the decision-making process. This is not just about deciding in ideal situations. No, conceptually, you have the entire decision tree of alternative paths that may have to be taken if things are not exactly as planned.
Conor Doherty: This is something I may talk about at a later date with Simon Schalit, the COO of the company. I talked with him recently about this, and he explained what you were talking about. Just to put some terms on it: the bill of resources. You have the parts, the people, the tools, and for any given process on any given day, you provide the client a sequence of actions. You need this and this at this time to complete the entire process. That’s generated overnight, for example, and then the following morning, something changes—Joannes, with his key certifications and qualifications, is sick. Let’s say there’s a 1% chance of that happening. The entire generated sequence needs to be regenerated to reflect that the original state of events is no longer the case. Maybe you can fill in a bit more detail there.
Joannes Vermorel: If we go back to the pre-software world, you can see that the reason why the supervisor of blue-collars was also the decision maker was because stuff would come up all the time, and they needed someone who could make decisions when deviating from the plan. But if you have software, then the software can make those decisions dynamically on the fly for you. That’s exactly what, for example, route optimization does. Let’s say you’re using Waze, and it informs you in real time that a street is blocked. It will suggest an alternative route. The decision being made is the exact itinerary you’re taking, and this is revised continuously based on the latest information about traffic, streets, etc.
When I say supply chain deals with a decision-making process, I’m not necessarily talking about things that are static. It’s most likely going to be a piece of software that can automatically revise decisions according to the latest situation. That’s why I emphasize automation is key—if it’s not fully automated, it means that, in terms of reactivity, if you have to go through a person, it’s going to be slow.
Conor Doherty: Like multiple people, most likely.
Joannes Vermorel: Yeah, I mean, we are talking of, if you have to consult with someone, we’re talking of what, half an hour response time if they’re available, etc. So it is very, very, very slow. So the only way to realistically have the decisions managed by the supply chain is to automate them completely. Otherwise, you end up back to what was done before, which is the supervisor, logistics, logistics on the side just improvising. And again, I would say it is better than doing nothing, but it can lead to all sorts of relatively bad decisions, and especially situations of non-compliance, such as, for example, where a driver would end up driving more hours than they’re allowed, and then an accident may happen. It’s very difficult to, you know, off the top of your head, readjust the planning in real time so that you preserve all the environment you want to preserve. That’s really what software is good for.
Conor Doherty: Alexey, who you know, Alexey Tikhonov, I’ve spoken with him before in Dijon, he has a nice phrase for what you’re just describing there, which is, “Any alternative solution is often a low bandwidth solution to a high-dimensional problem.” Which is, as you just said there, all the immediate consequences, the near-stream consequences, the far downstream consequences, the contingencies, like if I send you there, I can’t send you here, the opportunity cost this implies. Humans genuinely, I think it’s unreasonable to expect any human or even a group of humans, in real time with potentially $39,000 to 2 million per hour on the line and penalties, contractual penalties, and compliance issues, to collaborate and just go, “That’s the best solution to this problem.” That’s unreasonable in my opinion.
Joannes Vermorel: Yeah, I mean, in practice, people just have heuristics.
Conor Doherty: FIFO, for example.
Joannes Vermorel: Yeah, FIFO, exactly. First in, first out. Again, it’s okay. And I would even say that crafting the superior version of those heuristics is also a supply chain problem. You see, it would also normally be the mission of the supply chain to hand over the heuristics so that if the entire software system fails, here are your super simple heuristics that you need to adopt to keep the flow going. But we shouldn’t be under any illusion that those super crude heuristics are going to be very good. They are going to be better than freezing everything, but they are not going to be very good, they’re not going to be very efficient, and they will lead to predictable problems.
Conor Doherty: Yeah, well, again, to take the example of FIFO, and correct me where I’m wrong, but just to sketch that out: why? Because anyone listening might ask, “What’s wrong with FIFO?” As a side conversation, correct me where I’m wrong, but if you had, let’s say, MRO, you have two engines, engine A and engine B. Engine A comes in before engine B, requires more or less the same parts, same expected repair time. Engine A comes in first, so first in, first out, I work on that first. But to complete the repairs, I need another part that is not yet available. Whereas if engine B, if I were to repair that first, even though it came in second, it can get back into operation much quicker.
There might be financial implications to not getting engine B out the door as quickly as possible. Again, that’s a very black-and-white scenario, but it illustrates that I have a heuristic—it’s better than nothing. Repairing one is better than not repairing any, granted, but is that a financially optimized decision to take? Arguably not if you’re trying to maximize profit or return on investment.
Joannes Vermorel: Yes, and again, that’s why you need to separate those functions. If you’re already dealing with blue-collar workers, it is an immense responsibility. Being able to do real-time financial optimization in your head is just kind of nonsense. It’s not feasible. The best you can expect from people who are not superhumans is just basic heuristics that they can follow. The rest, anything more elaborate than that, needs to be done by people who can devote all their intellect to analytical processes. There’s no workaround. And the thing that has really changed is that software makes it possible.
I mean, it made it possible decades ago by just making the information available to people who are at a distance. So suddenly, you don’t need to be in the middle of the warehouse to know how much stock is left and what are the pending orders of the day. You can be at a distance at a desk and still have access to all the relevant information. That’s what I would say software enabled in the late ’90s for all companies. That’s how people started to isolate those functions because then people could work on spreadsheets, even if the software was not providing any intelligence.
So we’re still talking of a pre-automation era where we have just segregated people, and we can have the white-collar people in a different place. But nowadays, we can have the superior version where we can just automate the whole thing. So the white-collar people are not making the decision-making process anymore. They’re engineering the numerical recipe, having it executed automatically. And even from a risk management perspective, this is a superior approach. The thing is that if you depend on people, then mistakes will be made again and again. So in a way, if you want decisions to be very safe, then you can do that just like engineering is done in aviation, which is you have a lot of sequential reviews of the work. And once you have five stages of incremental reviews, you can be very confident in your outcome. But the problem is that it’s super slow.
That’s why, for example, when you want to put a new aircraft on the market, it takes a decade because of this super slow process of reviewing everything again and again. The problem with the supply chain is that you need decisions to be made quickly. So having a multi-stage process for reviewing decisions tends to create more overhead than the problem it’s trying to solve, which is the occasional incorrect decision. You need to make it fast. And if you have automation, the thing is that if you have a numerical recipe that has a defect, you fix the defect, and then all the decisions you generate from now on are free of this defect. That makes the process much more capitalistic.
That’s why, at Lokad, we really favor this “automate everything” approach. It’s not just for productivity. It’s about having an accretive process where every hour you invest makes the numerical recipe better, and every defect you identify can be fixed once and for all, as opposed to training people and realizing that they make mistakes, retraining them, and fine-tuning your training until you have a residual amount of mistakes that is very low—but still, it’s never going to be zero.
Conor Doherty: And also, that level of training leaves when the people leave. If they move on to another job or retire.
Joannes Vermorel: Yes, the numerical recipe will keep living with the company forever, while people will eventually leave.
Conor Doherty: Well, it occurs to me that at the start of the conversation, you drew a very sharp distinction between supply chain and logistics. And over the course of the conversation, we’ve touched on ideas of how automation is very clearly present in the software realm, which is the supply chain decision-making, and it’s in progress.
Joannes Vermorel: You see, that’s the thing. As far as logistics is concerned, we’re talking about the mechanization of the physical stuff, and it has been a work in progress for the last three centuries. People don’t even realize how much we have nowadays. The amount of mechanization in those fields is just gigantic compared to what it used to be.
You see, just compare the capacity of a modern truck with what people had with an early truck a century ago. A modern truck carries way more, is much more reliable, and it’s much easier to operate and everything. So even if we’re talking about vehicles, because a century ago people already had trucks, but they were not the trucks that we have today. The progress is very significant, and I would say logistics, just like in manufacturing, even more so, is heavily mechanized, but we still have people in the places where mechanization is just very, very difficult.
Conor Doherty: True, and well then my question becomes, that suggests that the definition started off as synonymous, diverged, but if automation is inevitable in both, will there be a convergence where it will just become one giant department? Supply chain and logistics just becomes known again as, I don’t know, operations research or supply chain or logistics, but like one of those terms and it does everything?
Joannes Vermorel: I mean the question will be the disappearance of the blue-collar job, and this is just science fiction, you know, for now. It may come at some point, and indeed if you were to consider a world where, let’s say, Tesla has succeeded with their Android robots and anything that a human operator can do, a machine can do cheaper and faster, then yes, blue-collar jobs would disappear, and the idea of having a position to manage teams of blue-collars would disappear as well.
You see, I believe that it’s a relatively distant future. I’m not sure if I will live long enough to see it because the challenges are just enormous. What is clear is that the domain is still progressing, you know, progressing very nicely, but we’re talking about a multi-century process on this front.
And there will probably be things that need to be changed in terms of infrastructure. People may not even realize that for the trucks that we have today to become so efficient, you needed, for example, to build warehouses that have a loading dock that is like one meter high and standardize that. Collectively, we had to adapt the entire infrastructure to take advantage of those very, very big trucks with appropriate loading docks.
That takes a long time. So whatever automation will come, they will probably need to revamp entirely things that I cannot even fathom yet, the infrastructure, and that will take time. But yes, at some point, if we remove entirely the need for blue-collars, the position of logistics director will disappear. And I believe that we already had in the past entire divisions of people that were already removed.
Nowadays, for example, most companies used to have an entire department dedicated to sorting mail. That is entirely gone. Occasionally, at Lokad, we receive a letter like once a week, and someone from the admin team just drops a letter to the person, but that’s it. This is gone. So I can envision a future where logistics is fully automated, and then indeed it will be a pure matter of analytics and engineering, but we’re still kind of far from this situation.
Conor Doherty: Well, by contrast, you’ve spoken in very strong tones, I’d say, about, and I’m quoting you here, “the extinction event coming as a result of automation in the supply chain space.” So sketch out why you see it as being quicker in the supply chain space.
Joannes Vermorel: If I were to draw a picture, I see, for example, e-commerce companies that have 500 blue-collars to deal with the stuff on the ground in their warehouse and organize their shipping and return logistics, and 500 white-collars to deal with the flow and management of it. That’s literally the supply chain running with a thousand people, half of them being blue-collars.
My take is that a decade from now, I do not see situations where those 500 blue-collars will be significantly reduced. Maybe, if they go crazy with automated warehouses and everything, they might be able to halve this number of people. I’m talking about people who are already extensively mechanized. However, as far as the decision-making processes are concerned, that is an area where going from 500 people to five is completely feasible.
And Lokad is already doing that for some of its clients. We now have clients where we have close to 1,000 employees consuming our decisions, but the decisions are generated for them, and Lokad is doing that with just a few supply chain scientists. So that really begs the question: okay, we had those 1,000 white-collar employees generating the bulk of decisions, but those decisions have now been mechanized. So obviously, the client doesn’t want to go to zero, but the idea of going from 1,000 to maybe a team of 20 is not unreasonable.
So that’s where again, that is software in action, exactly. I suspect that we will see an evolution that will be much more drastic for teams of supply chain workers who are now in many companies as extensive as the number of people on the ground. The idea of having as many people dealing with spreadsheets as you have on the ground physically dealing with the…
I think it was a little bit of a technological absurdity. We have so much super clever automation on the physical front, and for some odd reason, we were lagging behind on the software front. Now we are just catching up with the sort of productivity that should be expected on the decision plane.
Conor Doherty: Well, in terms of bringing things to a close, you mentioned earlier thinking capitalistically. So what are the potential capitalistic opportunities in supply chain based on the information you’ve mentioned today?
Joannes Vermorel: In supply chain the decision plane is to mechanize your decisions. Why? Because until you have mechanized everything, the process is not capitalistic. You’re not in a capitalistic setting. If you invest one man-hour, is it something that is consumed to generate the decision of the day, or is it something invested in making all future decisions better?
Conor Doherty: Well, this is about the improvement of the numerical recipe.
Joannes Vermorel: Exactly, and that’s why the practice of supply chain has been until very recently something that was not capitalistic. It was just Opex, you know, operational expenditure. You need this many man-days every day to generate the decisions your company consumes. That’s it. You can train people, but there are limits that were reached decades ago, and those people cannot be improved further by training, only very marginally because large companies have been training people for decades.
So you had reached the steady state decades ago, and where automation completely changed that is once you mechanize the decision-making process, then any man-days invested becomes like a capitalistic investment that will pay dividends indefinitely. That’s why in terms of return on investment, it’s incomparable because automation is literally a money-printing machine.
Conor Doherty: You’re making an asset essentially.
Joannes Vermorel: Exactly. There are limits, though. At some point, you do not see obvious ways to further improve your numerical recipe, and you get diminishing returns. So it’s not an asset that can have unlimited returns because you may stall in your capacity to make the recipe better. But the upper bound is still significantly higher than people believe. For companies that have not mechanized decision-making yet, the gap is absolutely enormous.
Conor Doherty: So, think in terms of money, yes. Well, Joannes, I don’t have any further questions. I picked your brain quite a bit today. I’ve missed these conversations, but again, thank you very much for your time, and thank you very much for watching. We’ll see you next time.
