00:00:04 Silos’ role in decision-making.
00:00:42 Silos’ impact on corporate structure.
00:02:00 Complexity and fragmentation in supply chain decisions.
00:04:06 Fragmentation in supply chain decision-making.
00:06:28 Practical example: Vertical fragmentation in fashion.
00:08:01 Problems of vertical fragmentation in businesses.
00:09:04 Horizontal fragmentation case: Diapers in supermarkets.
00:11:56 Cooking analogy to understand silos and fragmentation.
00:12:22 Silos’ prevalence and the need for productivity.
00:14:41 The value of productive teams and Lokad software.
00:16:00 Challenges in large-scale data processing.
00:17:07 Forecasting issues and probabilistic forecasts in supply chains.
00:19:00 Non-linear behavior and constraints in supply chains.
00:19:50 Prioritizing staff productivity in supply chain solutions.
00:22:00 Enhancing supply chains.
In this interview, Joannes Vermorel, founder of Lokad, discusses the concept of ‘silos’ or divisions within a company affecting supply chain decisions. He observes that these vertical and horizontal divisions can both benefit and complicate an organization’s operations. Fragmentation within these silos, he asserts, may lead to inefficient decision-making and potential business losses. To mitigate this, he recommends utilizing data analysis and forecasting for more informed decisions. Vermorel also dismisses the idea that increased productivity equates to job losses. Rather, he suggests that leveraging technology, such as Lokad’s software, can enhance the value of employees by enabling them to manage more complex tasks and make better decisions.
In this episode of Lokad TV, the discussion centers around the concept of ‘silos’ and their impact on decision-making within supply chains. Host Kieran Chandler questions Lokad’s founder, Joannes Vermorel, to elucidate his view on ‘silos’.
Vermorel expounds that silos emerge from the division of labor and a ‘divide and conquer’ approach in corporate environments. Silos can manifest both vertically and horizontally in an organization, creating a matrix of vertical and horizontal divisions that can each become a silo.
The conversation then shifts to the intricate nature of decision-making within supply chains, especially for multinational corporations operating across diverse countries and cultures. Vermorel posits that routine decisions, such as procurement decisions, become highly fragmented and complex, needing to be made for every product, location, and channel, often leading to thousands or even millions of decisions daily for larger companies.
According to Vermorel, fragmentation is a designed phenomenon, stemming from vertical and horizontal divisions in an organization. He uses an example of a fashion company to demonstrate the potential inefficiencies caused by this fragmentation, both vertically and horizontally.
Next, the topic turns to horizontal fragmentation, exemplified through the supermarket industry and the product of diapers. Vermorel asserts that fragmentation within a supermarket setting could lead to business losses if critical products run out, causing customer dissatisfaction and potential revenue losses.
Vermorel suggests that the root of fragmentation lies in the need for growing companies to manage an increasing number of tasks with limited personnel, necessitating a division of labor and subsequent fragmentation.
Drawing an analogy to a chef in a grand restaurant, Vermorel advocates for high productivity as a way to manage complex tasks efficiently with a small team. He cites Lokad’s software platform as a tool designed to enhance the productivity of supply chain scientists and improve decision-making.
In the episode, Vermorel also emphasizes Lokad’s transition to cloud computing in 2010. The shift to the cloud enables Lokad to handle large datasets swiftly, speeding up decision-making and enhancing overall productivity.
The conversation then covers forecasting, another critical aspect of supply chain optimization. Vermorel explains the challenges with forecasting, and suggests a forecasting engine capable of automatically extracting seasonality and addressing all possible futures probabilistically as a solution.
Moreover, Vermorel discusses challenges regarding non-linear behavior in supply chain management, such as minimum order quantities imposed by suppliers. He says Lokad employs a non-linear numerical solver to handle these constraints and improve productivity.
The conversation then addresses potential fear of job loss due to increased productivity and automation. Vermorel refutes this fear, asserting that productivity improvements should be seen as enhancing the value of employees rather than making them redundant.
The episode ends on a note emphasizing how technology, particularly data analysis and forecasting, is transforming the supply chain management industry. Vermorel underscores the benefits of adopting these technological advancements, which he believes not only optimize operations but also enrich the workforce, enhancing their productivity and decision-making capabilities.
Kieran Chandler: In today’s episode, we’re going to be talking about decisions and introducing the concept of silos. This might be a new concept for many of our viewers. It’s not about doing something better in particular, but about having the framework in place to do things better in general. Today, many supply chain consultants are charging absolute fortunes, pushing operational decisions that dramatically change the way our businesses are structured. So, Joannes, before we get onto whether these operational decisions are needed, perhaps you could start by clarifying what you mean by ‘silos’.
Joannes Vermorel: Silos are like an emergent property of some very old ideas that are based on the division of labor and a ‘divide and conquer’ approach to the corporate environment. By that, I mean companies grow and organize themselves, and a simple way to set up the organization is to have both vertical and horizontal splits. A vertical split in a fashion company, for instance, could have one division for shoes and another for shirts. The horizontal split could be a division for forecasting and another for replenishment. This forms a matrix - a divide and conquer strategy, and every box in this matrix can become a silo.
Kieran Chandler: Given the truly international way we’re conducting business these days, it’s easy to see how supply chains can become very complex. Let’s take the example of a large multinational operating across many different countries and cultures. Decisions can easily become fragmented. What sort of decisions are we actually talking about here?
Joannes Vermorel: In supply chains, there are many different types of decisions. You have high-level decisions, like when to enter a new market or sell in a new country. These decisions aren’t usually fragmented because they require approval from the CEO and the board. The fragmentation happens with the more routine, mundane decisions that are made thousands or even millions of times a day. We’re talking about decisions regarding when to buy, how much to buy, where to buy from, where to keep the stock, whether to move or liquidate stock, whether to lower or increase prices. These decisions need to be made every single day for every product, location, and channel. For companies of a significant size, it’s literally millions of decisions that need to be made every single day. That’s where the true fragmentation lies in these mundane supply chain decisions.
Kieran Chandler: We’re talking about basic supply chain decisions, such as how much stock needs to be replenished in a given day. However, these decisions are still being made by one department, like the purchasing team. So why are these decisions becoming fragmented?
Joannes Vermorel: That’s interesting. You can have an illusion that, just because there is a software or team at the end of the chain making a decision, the decision is centralized. But the reality is that the decision is fragmented by design. The fragmentation comes from the vertical and horizontal design of the decision process. Let’s delve into how this fragmentation plays out.
Kieran Chandler: For instance, one approach would be to have quarterly budgets for divisions. In this scenario, you have a purchasing budget for the division handling shoes and a quarterly budget for the division managing shirts. This arrangement puts a constraint in the middle and fragments the decision-making process, although it may not be immediately apparent.
Joannes Vermorel: Absolutely, and if we look at it horizontally, we might think about forecasting versus replenishment. In many companies, you have one team that determines the forecast, which is essentially the average projected future demand. It’s not how we do it at Lokad, but in many companies, that’s their approach.
On the other hand, you have a different team dealing with replenishment and determining the safety stock. In many systems, your decision, which is what your order quantity should be, is essentially a sum of half of the demand plus the safety stock. This decision is built on many parts that have been produced by possibly many different teams. That’s how you end up with fragmentation, even though in the end, someone just issues a purchase order. But the process leading to this purchase order is highly fragmented.
Kieran Chandler: Okay, I think I understand, but it might make more sense to our viewers with a practical example. Can you give us an example of vertical fragmentation, and explain how businesses are actually losing money because of it?
Joannes Vermorel: Sure. So, let’s revisit our hypothetical fashion company that is producing shoes and shirts. Why does this company lose money because of a quarterly budget? In fashion, it’s all about trends and being fresh. If you want to be fashionable, you shouldn’t be wearing things that were in style last year. You need to stay on top of the latest trends.
If you are operating with a quarterly budget, by design, you’re a quarter late for everything and every single trend. What does that mean in practice? The item that’s becoming the super hot trend right now doesn’t have a budget big enough to meet the demand because you can’t perfectly anticipate future trends. There’s always an element of surprise.
If you can’t purchase enough of the trendy items, you’re not going to be able to sell as much as the market demands. Conversely, the division that’s selling something that’s not so hot, they have an excess budget, but their sales are suffering. Do you think they will willingly give up their budget? No, they’re already struggling with sales. From their perspective, they expect to maybe turn the situation around. However, the company as a whole is losing because of these vertical splits. It’s always one quarter late, and that plays out differently in various areas, but the problem remains the same.
Kieran Chandler: So, the main drawback of vertical fragmentation is a company’s lack of agility and slow response to changes in the marketplace. I suppose that means they need to forecast further into the future. What about horizontal fragmentation? Do you have an example of that as well?
Joannes Vermorel: Oh, yes, good question. Let’s consider the case of supermarkets selling a wide range of products. For this example, let’s specifically look at diapers. Diapers are quite interesting. If you’re a young parent, diapers are mission-critical. If you run out of diapers with a newborn infant, you have a significant problem on your hands. It’s a sensitive product that’s quite expensive for many people and, as I mentioned, it’s mission-critical. Now, let’s look at how fragmentation affects this scenario.
Kieran Chandler: Could you explain why we face specific issues in supply chain management for certain products, such as diapers in a supermarket setting?
Joannes Vermorel: Absolutely. Diapers are mission-critical items in supermarkets. If your supermarket runs out of diapers, customers, particularly young mothers, will be significantly affected. Their resentment could not only cause you to lose the current sale, but if they resent you enough, you could potentially lose their business for the next five to ten years. It’s an incredibly asymmetric problem where one missing item can cause you to lose a decade of business with a customer.
Kieran Chandler: How does this affect the company’s decision-making process?
Joannes Vermorel: It’s interesting because the people who understand this problem the best are not the forecasting or replenishment teams, but the marketing people. They understand the customers’ needs because that’s what they study all day. Despite this massive asymmetry in the problem, no one is really accountable for it, and a purchasing decision is made without factoring in this super asymmetrical cost. This fragmentation of decision-making can lead to customers resenting the company because the decision was not smart enough. This is how companies can lose a lot of money.
Kieran Chandler: That’s a fascinating perspective. So, it’s not about the decision itself becoming fragmented, but how the decision emerges. Is it like starting with great ingredients but having a bad recipe?
Joannes Vermorel: Exactly, no matter how good your ingredients are, if your recipe is not good, it won’t result in a great meal.
Kieran Chandler: This brings us to the problem of silos that have existed for decades. Why are companies still struggling with this issue?
Joannes Vermorel: Companies are still struggling because the solution, while simple, is not easy. You need a massive boost in productivity. Silos exist because as a company grows, it can’t deliver everything it wants with a small number of people. So, you create divisions to manage the workload. However, if you had tools that significantly boost productivity, you wouldn’t need to have these divisions as fragmented as they are.
Kieran Chandler: You mentioned earlier about the analogy of a chef…
Joannes Vermorel: Yes, and that’s a perfect example. In high-end restaurants, chefs are incredibly productive, working so fast that it’s almost like watching a TV show. Why is speed so crucial in cooking? It’s because of fragmentation. As the saying goes, too many cooks spoil the broth. So, in the end, there is just one recipe, one chef, and a few assistants. You cannot have 200 people working on the same dish; it’s not going to turn out well. That’s why a small team with massive productivity is crucial. It’s an entertaining sight to see them working so quickly, but there’s also a crucial lesson in that.
Kieran Chandler: For supply chain, the main idea is that if you want to make great supply chain decisions, you need to have a few people who have high productivity in making those specific decisions. This ensures that the resulting decisions are of high quality. And by the way, this is exactly what Lokad is about as a software platform. It’s a software platform dedicated to the productivity of supply chain scientists, as we discussed last week. We don’t have data scientists, we have supply chain scientists, which is way cooler.
Now, I’m glad we talked about food because I’m currently addicted to Chef’s Table on Netflix. It’s a passion of mine at the moment. However, getting back to productivity, you mentioned that it’s the one thing holding back great tacticians from making great business decisions. So, what is hindering productivity and how can it be improved?
Joannes Vermorel: Well, it’s a simple problem, but a complex one at the same time. You could say it’s death by a thousand cuts. It’s not just one thing that drags your productivity down, but many. Let’s list a few, shall we?
First, you have to deal with lots of data, potentially terabytes. The question is, can you process this vast amount of data in minutes, or ideally seconds? Usually, you can’t, and that’s due to a lack of big data processing capabilities. For instance, if you need 20 minutes to process one terabyte, that means the person doing the calculation is just sitting and waiting for 20 minutes until the results come out. So, computing performance is a crucial factor for productivity. You need something that operates lightning-fast because that’s what it takes to be productive.
That’s why, in 2010, we moved Lokad to cloud computing because it offered access to tremendous computing resources to solve this very problem.
The next aspect is the forecasting angle. The productivity can be quite low here because you have many edge cases. For example, you might need to make special adjustments for products with limited history, products that haven’t been launched yet, products with less than a month of sales history, products about to have a promotion, and products that had stock-outs in the past. If a data scientist has to manually tweak all these edge cases, productivity is incredibly low.
Furthermore, you shouldn’t have a forecasting engine that requires you to manually adjust for seasonality. It should automatically extract this from the data.
Additionally, we also need to consider probabilistic forecasts. Companies can waste a lot of time on “what if” scenarios. The issue is that if you start to list all the scenarios, it becomes an endless stream. This process takes a massive amount of time. The solution is to have a forecast that addresses all possible futures with probabilities.
And finally, there are problems related to nonlinear behaviors, like minimum order quantities from suppliers. They might say you need to order at least 500 units, for instance, which is a constraint. The way we address this at Lokad is by having a nonlinear numerical solver.
So, enhancing productivity involves having a comprehensive software solution that delivers productivity at every single stage. So what you’re basically saying is that with the right software, a supply chain scientist has the tools to make better decisions without having to use lots of extra resources. That sounds like a good strategy for companies that don’t want stock-outs in the future but don’t want to double their headcount to prevent it.
Kieran Chandler: If it really does boil down to just a few key members of the supply chain, what are you going to do with the rest of the supply chain staff? Are you going to fire them?
Joannes Vermorel: That’s very funny because every time I’ve seen productivity discussed in my professional life, there is the journalistic take which is, “Oh, productivity is improving, we are just going to fire half of the company.” Guess what, it’s not exactly what happens.
Think about it from the perspective of the CEO. Unless you’re an actively dysfunctional CEO, how do you see your people? You’re training them, they have economic value. They’re economic assets for the value of the company. Your company is worth something because there are valuable people who make the company work. It’s not just a machine, it’s a group of people.
So, the most rational answer to “we have a tool to make people more productive” isn’t “let’s throw half of the people out of the window”. That’s wrong. It’s not about discarding half of your assets. Your company wouldn’t suddenly be worth more.
What’s most likely to happen in a non-dysfunctional company, where people are actually doing valuable things, is that they are going to use those supply chain scientists to come up with better decisions. But these scientists don’t operate in a vacuum. They need better data, and it takes a lot of effort to collect data more reliably, to collect more data quantitatively.
For example, you might want to collaborate more with your suppliers. You might want to negotiate your minimum order quantities. By doing so, you can make your life and your supplier’s life easier, potentially resulting in a lower price. Not because you bargained with your supplier, but because you lowered the price of what you’re about to buy in the first place.
The same thing can happen on the customer side. By understanding your client better, you can improve the service, which is not just about making things cheaper. Sometimes people are willing to pay for things that are more expensive because they get more out of it. The ‘more’ can be faster delivery, more reliable delivery, even a different class of product properties.
I believe that with the productivity we can deliver from better supply chain scientists, companies will be able to spend more time on things that make more sense, as opposed to spending time on tweaking numbers on Excel sheets which barely makes any sense at all.
Kieran Chandler: Well, I’m afraid we’re going to have to wrap things up for today, but thanks for a really interesting discussion. We’ve talked about fashion, restaurants, even diapers. It’s definitely been entertaining and hopefully, it’s given our listeners a better understanding of both vertical and horizontal fragmentation, and how it can affect the world of supply chains. Silos is a very interesting topic given the way that they resist change. So that’s all for this week, but we’ll be back very soon with another episode. Until then, we’ll see you very soon. Bye-bye.