00:00:07 Nicolas Vandeput Introduction and Bridgestone operations.
00:01:05 Bridgestone’s complex multinational supply chain.
00:02:25 Lokad project: initial scope and objectives.
00:03:29 Lokad’s unified model and network perspective.
00:06:13 Tire business: combinatorial complexity and seasonality.
00:08:00 Balancing inventory and predicting demand.
00:08:47 Change management: adopting new stock targets.
00:11:49 Challenges of probabilistic forecasting and lead times.
00:13:52 Choosing Lokad for tailored probabilistic forecasting.
00:14:38 Programmatic approach for Bridgestone’s supply chain.
00:16:02 Fitting software solutions to unique supply chains.
00:17:01 Project implementation, initial results, adaptability.
00:19:11 Project evolution and maintaining agility.
00:20:33 Bridgestone’s IT infrastructure and operations impact.
00:21:37 Lokad’s project impact and key takeaways.
In an interview, Kieran Chandler discusses a supply chain optimization project between Lokad and Bridgestone with Joannes Vermorel and Nicolas Vandeput. The project aimed to develop a unified model to improve supply chain performance across Bridgestone’s complex European network. Lokad’s flexible and programmatic approach, coupled with close collaboration between both companies, resulted in a successful implementation. This streamlined processes and allowed local teams to focus on more pressing matters rather than competing for stock allocation. The project showcases the feasibility and benefits of a quantitative approach to supply chain management, highlighting the potential for future collaboration in other companies.
In this interview, host Kieran Chandler speaks with Joannes Vermorel, the founder of Lokad, a software company specializing in supply chain optimization, and Nicolas Vandeput, who discuss the recent project completed between Lokad and Bridgestone, the world’s largest tire manufacturer with 180 manufacturing plants and presence in over 150 countries.
Bridgestone Europe, the focus of the project, has eight manufacturing plants and 20 warehouses across the region, handling around 40 to 60 thousand different SKUs each month. The challenge is managing the complex supply network that spans from Latvia to Portugal, with different countries, languages, and seasonality.
The initial scope of the project aimed to address the issues caused by multiple teams in different countries managing their supply chains independently. This approach often led to local optimizations but failed to achieve a globally optimal solution. Sometimes, competition between countries for scarce resources even resulted in “wars” and subsequent shortages.
To address these issues, Lokad and Bridgestone sought to develop a unified model run from a single location, which would provide consistent stock targets for all teams involved. This required considering the entire supply network, including various stages and streams of products from production to end consumers, rather than optimizing individual locations.
Lokad’s approach to solving the problem involves a holistic, network-wide perspective that avoids merely shuffling issues from one location to another. Instead, it aims to address inventory issues throughout the network without creating new problems elsewhere. This is particularly challenging in Europe, where the economic market is both unified and fragmented across different countries.
The ultimate goal of the project was to create a unified model that would optimize the supply chain across the entire network, avoiding local inefficiencies and competition between teams. By doing so, Lokad and Bridgestone aimed to improve overall supply chain performance and ultimately better serve the end consumers.
Vermorel discusses the challenge of optimizing supply chain networks, stating that it is essential to focus on the end consumer and not just the intermediate layers of the supply chain. He highlights the complexity of the problem, with every decision made for a single SKU potentially impacting all others in the network. Additionally, the time-driven nature of supply chains adds to the complexity, as any change made for one SKU will have consequences over time for all other SKUs.
Vandeput emphasizes the importance of understanding seasonality in supply chain management, especially for businesses dealing with seasonal products like tires. He stresses the need to plan inventory levels not just for the immediate month but also for several months ahead to account for changing seasonal demands.
When discussing the change management process, Vandeput acknowledges the difficulty in convincing people to adopt new approaches to supply chain optimization. He shares that they started with a smaller scope by focusing on a specific subset of tires, proving their case before expanding to the entire Bridgestone company. He also highlights the importance of continuous training, communication, and demonstrating the effectiveness of the new approach through data and graphs.
Vermorel agrees with the counterintuitive nature of some supply chain optimization strategies, such as reducing stock levels at specific locations. By doing so, companies can actually liberate more stock for the rest of the network, including their own internal suppliers, and increase the probability of being able to serve customers better.
Vandeput explains that one of the issues they faced was a long-held belief that stock should be kept close to the customer. Through optimization, they realized that a larger portion of the stock should be kept at manufacturing plants, which allows for more flexibility in meeting demand across the entire network. This change required effective change management to convince stakeholders to adopt the new approach.
The project focused on addressing the uncertainty in the supply chain, including demand, lead times, and stock quantities. Despite predictable consumption patterns in the tire industry, the multinational network generates a significant amount of noise, which Lokad aimed to account for using probabilistic forecasting.
Lokad differentiated itself from other software in the market by offering a tailor-made approach, focusing on skew-by-skew probability functions rather than relying on normal distribution assumptions. This allowed Bridgestone to utilize Lokad’s technology to address their specific needs more effectively.
The interviewees explain that the key to Lokad’s approach was its programmatic nature, starting from a blank sheet and leveraging a highly predictive environment focused on supply chain problems. This allowed Lokad to craft a solution that fit Bridgestone’s unique supply chain requirements, in contrast to off-the-shelf solutions that require companies to adapt their problems to predefined templates.
The project focused on the parameterization of replenishment policies for every single SKU across the network, with Lokad serving as an analytical layer and not a replacement for existing ERP, WMS, or MRP systems used by Bridgestone. The project’s goals evolved rapidly at the beginning and continue to evolve, albeit more slowly, in response to changes in the company’s supply chain and economic drivers.
Vandeput emphasizes the importance of flexibility in the optimization software, allowing for continuous improvement based on feedback. He highlights that the project did not require changes to Bridgestone’s existing IT infrastructure, as Lokad’s solution only produces stock targets, which are then fed to existing systems. This setup allows for rapid deployment and adaptability, while maintaining the current IT landscape.
The interview also discusses the impact of the project on Lokad. Vermorel notes that the successful implementation of the large, multi-national project demonstrates its feasibility and the benefits of a quantitative approach to supply chain management. He mentions that the project has helped upgrade long-standing processes at traditional manufacturers like Bridgestone. By streamlining these processes, local teams can focus on more pressing matters rather than managing Excel sheets and competing for stock allocation.
Vermorel further elaborates on the detrimental effects of competition between replenishment teams, which can lead to suboptimal supply chain performance. He believes the project’s success lies in its ability to free up time for teams to work on problems that benefit the entire organization, fostering a more collaborative environment.
Vandeput reflects on the project’s success, attributing it to the close collaboration between Lokad and Bridgestone’s local teams. The cooperative approach has helped create a more efficient and effective supply chain model. He expresses enthusiasm for applying this experience to future projects with other companies.
Kieran Chandler: Today on Lokad TV, I’m delighted to say we are once again joined by Nicolas Vandeput, who’s going to help us take a peek behind the scenes of Bridgestone and understand the daily operations of a large multinational corporation. So, Nicolas, thanks very much for joining us again on Lokad TV. I think a nice place to start is discussing the project between Lokad and Bridgestone. Could you just illustrate the scale of the project, like how many SKUs and different locations are we talking about here?
Nicolas Vandeput: It’s really a huge project. In Europe, we have around eight manufacturing plants and on top of that, 20 warehouses, which is in total 40 to 60 thousand different SKUs that we have to manage automatically every single month, six months ahead. So yes, it’s really huge.
Kieran Chandler: Joannes, we’re talking about a multitude of locations, from Latvia all the way to Portugal. What complications does having so many different locations introduce?
Joannes Vermorel: The crux of the problem is that when you have such a complex network, it’s challenging to actually solve the problem. When you typically tend to tweak your network, you don’t solve the problem; you just displace it. So, you might solve an inventory issue in one country, but you just create another inventory issue somewhere else, or even exacerbate it. The real challenge is to not shuffle around the problems but solve them from a holistic, network-wide perspective. That’s very difficult, especially in Europe, which is both economically unified and fragmented, with different countries, languages, and seasonality.
Kieran Chandler: It would be useful for our viewers to understand the initial scope of the project. What was the initial idea behind the project, and how did it all start?
Nicolas Vandeput: Initially, a couple of years ago, we had supply chain management by different teams in every single country. Each team managed their supply chain differently, resulting in a lot of issues. Even if you reach the local optimum, you don’t necessarily reach the global optimum. There were even some conflicts between countries, with shortages on specific items leading to increased competition and fighting over resources. That was the situation at the beginning of the project. We had different teams with varying levels of maturity, and then we decided to have a unified model run from a single location to put everyone on the same level, using the stock targets.
Kieran Chandler: Let’s talk a bit about that unified model. Joannes, what was so different about the approach that Lokad took to this problem?
Joannes Vermorel: The perspective is to conceptually run and execute the entire network. It’s not about optimizing one location; instead, we think of the whole multi-echelon supply network, with different stages through which the goods, which are tires, flow from production to end consumers via various channels. When considering this network, you have to think of all the streams of products and the demand.
Kieran Chandler: Every single stage, except the end consumers, is basically things that you generate yourself. You see, you have to be very careful not to mix what is actual demand with what is the actual noise of your own process that just generates orders inside the network. So our approach was to have a complete modernization of this network, complete knowledge in the flows, which is easier said than done, by the way, and to align that with the endgame, which is ultimately serving the channels themselves as best as possible. It’s tricky because the mistake is to focus on what is happening in the middle of the network, which is, in a way, kind of completely irrelevant. For example, from an end client perspective, which we call the “best perspective,” it doesn’t matter if half of the intermediate layers in the supply chain are out of stock, as long as when you request your tires, they are there for you. So, the trick is that what matters is the perspective of the best. What is in the middle is not inconsequential, but it is very important. It’s only a means to an end; this is not the end. That’s where it gets very tricky, both from a data processing perspective and even from a pure mathematical perspective.
Joannes Vermorel: The way I like to think about this kind of problem is that there are lots of different layers and tiers. Some of the key difficulties introduced are the sheer combinatorial complexity and the sheer number of possibilities. As soon as you have many SKUs, you want to suppose we are talking about tens of thousands of SKUs. You have to think that every single decision made on pretty much every single SKU can potentially impact all the others. It’s fundamentally a quadratic kind of problem where, whenever you look at a SKU, it can have an impact on any other SKU. So you have like 60,000 SKUs times 60,000 SKUs if you want to investigate the ramifications of every single change that you’re doing. But it’s even worse than that because it’s time-driven. If you touch a SKU, it will have consequences over time for all the other SKUs. So, you end up with every single SKU having the potential to impact all the other SKUs over time, for up to six months per SKU. Obviously, we are doing things a bit smarter than that because that would be way too complex.
Nicolas Vandeput: Seasonality was also an important factor in the tire business. You have summer tires, winter tires, and you have to know that at some point, the season will end or another season will start. So it was really important not just to focus on one single month like what’s going to happen next month. We also wanted to know what should we have in two months, three months, or four months time. When you put inventory at a certain place, you have to know that next month might be totally different. So maybe you can have a bit more stock because the season is ending, or maybe you need a bit less stock because that’s the end of the season. It’s a really difficult problem, especially as soon as you start to look at it six months ahead.
Kieran Chandler: Let’s talk about the change management process now. When you’re coming to the powers-that-be and you’re explaining that you’ve got this new approach you want to take all under one company and take one singular approach, how did you approach that change knowing the inherent complexity behind it? Were people really interested in it?
Nicolas Vandeput: Change management is extremely difficult, that’s for sure. It’s one of the main challenges.
Kieran Chandler: To begin, can you talk about the process of implementing the new stock target model and how you overcame initial challenges?
Nicolas Vandeput: We started with a smaller scope, focusing on just a specific subset of tires. We tested it for a few months and proved that it worked. Once people started to understand that it worked for this subset, we could grow the project to cover the entire Bridgestone company. On the side, I also spent a lot of time training people and communicating with them. We showed many examples of counterintuitive results, like people who assumed that having a thousand tires in stock would be the right amount, but actually, 200 would be enough. We had to show graphs, numbers, and forecasts over and over again until people agreed that it would work with 200 tires.
Joannes Vermorel: I completely agree with the counterintuitive effect. When you say you’re going to reduce the stock, it actually means you’re liberating more stock for the rest of the network, including your own internal suppliers. If you keep less stock, it typically means that whatever comes before you might actually keep more, so if you do run out, you have a higher probability of being able to fulfill the demand. The same is true for every other location because they are well-behaved.
Nicolas Vandeput: Before we started the project, the main belief was to have all the stock close to the customers or the “bus,” as we call it. However, after running the optimization, we realized that a larger part of the stock should actually be at the manufacturing plants, which are further away from the clients. This allows more flexibility when a market needs extra tires. It was a difficult change for people who had believed for nearly a decade that the stock needed to be close to the market.
Kieran Chandler: So, you touched upon change management, and the project went live back in March 2018. What were the initial technical challenges you faced?
Joannes Vermorel: First, we had to implement a probabilistic vision of a multi-echelon network, which is quite complicated. Probabilistic means that the future is uncertain, not only in demand but also lead times and even the actual stock quantities at any point in time in the network. This quantitative supply chain approach was, at least for us, a bit unprecedented in terms of the sheer amount of uncertainty that needed to be properly accounted for.
Nicolas Vandeput: The trick is that the uncertainty doesn’t mean that you don’t know anything. The consumption patterns, from the tire perspective, are not that erratic. There are entire fleets of vehicles driving in ways that are quite predictable.
Kieran Chandler: Joannes, could you start by explaining what makes the demand pattern in Bridgestone’s supply chain so unique?
Joannes Vermorel: Of course. While the overall demand for tires is quite predictable and steady, there is a significant amount of noise generated within the multinational network. This noise can be accounted for with probabilities, which is one of the main reasons we chose Lokad.
Nicolas Vandeput: That’s right. Other software on the market makes strict assumptions, such as demand and lead time following a normal probability curve. But when you consider factors like demand, forecast, and lead time, normality isn’t the best fit for Bridgestone. We needed something more accurate and tailor-made, and that’s where Lokad shines with its ability to define skew-by-skew probability functions.
Kieran Chandler: How did you take that noise and translate it into a probabilistic forecast?
Joannes Vermorel: We have a whole algebra of probability in Lokad that we use with Envision. But at a more fundamental level, the key differentiator was our programmatic approach. We start projects from a blank sheet technologically, although we have many building blocks available. Instead of trying to fit Bridgestone into a template, we used our programmatic environment, which is highly predictive and expressive for supply chain problems, to craft a solution that really fits Bridgestone’s unique needs.
Nicolas Vandeput: I’d like to add that other software providers often come with a pre-defined solution, and you have to fit your problem into that solution. This can lead to losing focus. Bridgestone, like any other large supply chain, is unique in different ways, and that’s why we needed Lokad to start from a blank page and work together to construct a tailored solution.
Kieran Chandler: Let’s discuss the results of the project. How was it implemented, what were the initial results, and what changes were made as the project progressed?
Joannes Vermorel: At Lokad, we focus on decisions. Our goal is not to have a percentage of reports, but to output decision-like information that has a physical impact on the supply chain. In this project, we worked on replenishment policies.
Kieran Chandler: Can you discuss the output of the project completed between Lokad and Bridgestone and how it has changed over time?
Joannes Vermorel: The output of the project is primarily the numerical parameters governing replenishment policies for every single SKU across the network. Lokad is an analytical layer; we do not replace the ERP, WMS, or MRPs that Bridgestone uses. We focus on the parameterization of the replenishment policies. What has changed over time is the way we reflect the economic drivers of our goals and how we define our targets. This has evolved more slowly, but it is still evolving. As Bridgestone undergoes further transformations, I believe this will continue to evolve at some pace.
Nicolas Vandeput: Currently, we generate results for Bridgestone with a daily pipeline. However, they probably use these results on a weekly or monthly basis, as their entire data network needs to retune to the new strategy. Bridgestone’s supply chain is like a living organism, changing every month. We have to stay agile to accommodate changes in parameters, warehouse openings or closures, changes in tire routes, and other client behaviors.
It was important to me that our project remains agile, with a strong mathematical foundation but also an open layer for discussion with local planners. This allows us to receive feedback and make updates to our approach as needed. I cannot predict how the project will be in six months because if I already knew, I would have implemented those changes.
Kieran Chandler: What has been the impact of the project on Bridgestone’s IT infrastructure and daily operations?
Nicolas Vandeput: I’m happy to say that the impact on IT has been minimal. Our goal was not to replace their existing systems but to use Lokad to generate stock targets as an output that can be fed into their systems. Lokad acts as an external, intelligent machine to populate these targets, allowing us to move fast without going through extensive IT procedures. We can deploy our solution quickly, which has been a great advantage for the project.
Kieran Chandler: First impression when you think about it, the implementation time with this project has been extremely short thanks to many different factors, one of them being the fact that we don’t need IT infrastructure from Bridgestone. Joannes, what about the impact at Lokad? What have we learned from implementing such a large, multi-stage project such as this one?
Joannes Vermorel: First, we learned that it was possible. That’s great. We are very confident in our capability to execute, but still, it’s never a given. That’s one of the reasons why you have to move fast, because you want to fail fast if, by any chance, it would not work out. You don’t want to spend three years on a project, only to realize that the technology isn’t ready yet or it’s just not working. The fact that we have this quantitative supply chain is also a game-changing thing for more traditional manufacturers. I mean, not in the sense that Bridgestone isn’t at the top of their game in terms of technology for tires. It’s just that any company that has been around for decades has had processes in place for a long time, and it’s interesting to see that those processes can be successfully upgraded. The way I see it, we can free up a lot of time on many local teams to focus on things that matter more than crunching Excel sheets. If you take the old system where you have many countries, many locations, each one fighting for their own stock allocation – which was the case – it’s not that they are badly trained, it’s the by-design consequence of the process. If you set up your process with replenishment teams that compete with one another, then guess what? They will compete, and that leads to competitive behaviors that degrade the performance of your supply chain as a whole. It’s not that they are badly trained; it’s by design. One of the success factors of this project is to think that all those teams will have a lot more time to focus on problems they can actually solve, and solving those problems will be a win-win with the rest of the company instead of having to fight for a winner. Then, the next round, you win, I lose, etc. This is something that is much more beneficial for Bridgestone as a whole.
Kieran Chandler: Nicolas, I’ll leave the concluding comments to you. What are your key takeaways from this project, and what are the key results that you’ll really take back from this?
Nicolas Vandeput: I’m extremely happy that we could achieve this together with the team, and I’m talking about two different teams. First, Lokad, as this has been a mathematical challenge and a great experience with Lokad. But also, this has been a challenge with the team at Bridgestone. What has been extremely important to me in this project, and it’s still ongoing, is the fact that we could get feedback from the local team, the local planners, to help us design the best model possible. So, it has been really done hand-in-hand with the local team. It’s something that has brought all these people, who have been used to fighting for stock, together to design the best model. To me, that’s a real success, and I hope that I’ll be able to do that with Lokad and other companies.
Kieran Chandler: We’re going to have to wrap up there, but thanks very much for your time this morning. That’s everything for this week. We’ll be back again next week with another episode, but until then, thanks very much.