00:00:00 Introduction of Dan Scharneck, Supply Chain Director at Trek Bicycle
00:00:29 Trek Bicycle’s mission and differentiation
00:01:50 Trek’s custom bike business and configurability
00:03:33 Logistical complexity and parts of Trek bikes
00:04:51 Trek’s collaboration with Lokad and combinatorial complexity
00:06:56 Shift to mass production for cost efficiency
00:08:10 E-commerce, supply chain, and customization in industries
00:10:41 Reliable delivery time and inventory level challenges
00:12:29 ERP limitations and forecasting with configurable products
00:15:20 Bottleneck issues and technological obsolescence in bike production
00:18:34 Introduction to probabilistic forecasts and bottleneck risks
00:20:26 Transition from traditional ERP to Lokad
00:21:59 Daily dealings with probabilistic forecasting and decision-making
00:23:44 Applying probabilistic forecasting to different tasks
00:26:03 Cost factor, supplier management, and inventory rebalancing
00:29:10 Incremental approach of probabilistic forecasts and recomputation
00:30:52 AI’s role and long-term performance of probabilistic forecasting
00:32:40 Future inaccuracy and impact of configurator changes
00:34:25 Dealing with uncertainty, evolution, and pre-existing complexity
00:35:34 Trek’s sport events following and value creation
00:38:06 Simplifying Lokad’s role and advice for configurability
00:39:07 Managing complexity and importance of a reactive supply chain


In a dialogue moderated by Conor Doherty, Joannes Vermorel of Lokad and Dan Scharneck, the Director of Supply Chain at Trek Bicycle explored the intricacies of supply chain management for customizable products. Vermorel emphasized the combinatorial complexity of supply chains, noting that companies are using them as a differentiator for customer experience. Scharneck agreed, highlighting the importance of delivering custom products promptly and reliably. He acknowledged the challenges of managing inventory with traditional ERP systems, leading Trek to adopt Lokad’s more adaptable solution. Vermorel explained that Lokad uses probabilistic forecasts to assess the risk of bottlenecks, providing a more powerful perspective on supply chain delays. Both agreed on the importance of managing supply chain complexity and the value configurability offers to customers.

Extended Summary

In a recent discussion hosted by Conor Doherty, Joannes Vermorel, the founder of Lokad, and Dan Scharneck, the Director of Supply Chain at Trek Bicycle, delved into the complexities of supply chain management, particularly in the context of customizable products. Trek, a company that prides itself on providing an excellent customer experience, offers a custom bike business, Project One, where customers can choose from a wide array of options to personalize their high-end bikes. This, however, presents a unique challenge in managing the supply chain, given the multitude of parts and configurations involved.

Vermorel, a specialist in supply chain optimization, highlighted the combinatorial complexity of supply chain management. He noted that while the number of combinations can be large, it doesn’t necessarily reflect the actual complexity of the supply chain. He further observed that companies are increasingly using supply chain as a differentiator for customer experience, offering more flexibility and configurability to customers, a trend seen across various industries.

Scharneck concurred with Vermorel, emphasizing the importance of delivering custom products quickly and reliably. He mentioned that Trek aims to deliver a bike anywhere in the globe within 30 days. However, he acknowledged the challenges they faced in managing inventory levels using traditional ERP systems and forecasting software, which were not designed for the complexities of a custom business. This led Trek to seek a more customizable solution, which they found with Lokad.

Vermorel explained that traditional tools can provide a classic forecast, but this becomes complex when dealing with configurable products with near infinite options. He pointed out that forecasting at the part level doesn’t account for bottlenecks and that increasing the service level to 99.9% results in a massive increase in stock, which is not feasible in the bike industry due to technological obsolescence. Lokad, on the other hand, uses probabilistic forecasts to assess the risk of a part becoming a bottleneck and for how long, providing a more powerful perspective on supply chain delays.

Scharneck admitted that transitioning from a traditional ERP to probabilistic forecasting has been a challenge, and there is still a need to educate people within the business about this new approach. However, he noted that probabilistic forecasts with Lokad have led to better decision-making and are helping to get things back on track after the pandemic.

Vermorel further elaborated on Lokad’s holistic forecast, which allows for multiple dimensions of decision-making, including modulating orders in progress and rebalancing inventory between locations. He also highlighted that Lokad can recompute new probabilistic forecasts daily, which results in incremental changes rather than drastic shifts in planning.

In conclusion, both Vermorel and Scharneck agreed on the importance of understanding and managing the complexity inherent in supply chain management, particularly in the context of customizable products. They emphasized the need for good systems to manage this complexity and a reactive supply chain that can respond to the custom business. They also underscored the value configurability offers to customers, and the importance of demonstrating long-term performance and results to increase the adoption of AI and machine learning in supply chain forecasting.

Full Transcript

Conor Doherty: Welcome back to Lokad TV. I’m your host, Conor, and as always, I’m joined in the studio by Lokad founder, Joannes Vermorel. Joining us remotely today is Dan Scharneck, the Director of Supply Chain at Trek. He’s going to talk to us about forecasting demand with configurability. Dan, welcome to Lokad.

Dan Scharneck: Hey Conor, hey Joannes, good to speak with you guys.

Conor Doherty: All right, Dan, so before we get into the nitty-gritty, let’s talk a little bit about Trek. A lot of people have heard of Trek, I’ve heard of Trek, some of my friends have, but what exactly is it that Trek does?

Dan Scharneck: Well, we’re a bike company, we sell bikes. We actually view ourselves as a hospitality company. To us, it’s all about giving that experience to the customer. So whether you go to a Trek retail store, or you go to a Trek event, or you buy one of our products, it’s all about providing incredible hospitality. We think by extension, we’ll get more people on bikes, and that’s a good thing for Trek, a good thing for the planet.

Conor Doherty: And what exactly is it that separates buying, let’s say, a high-end Trek bike from buying any other high-end bike? I mean, it’s not just the hospitality, I presume.

Dan Scharneck: Yeah, well, so I think that’s where the business that my team is responsible for, the Project One aspect of Trek, comes in. That’s our custom bike business, and it’s six to fifteen thousand dollar bikes, so very expensive high-end bikes. We offer a wide selection of choice, whether it’s paint colors, all the parts can be chosen on the bike, so you really get to make it yours. And there really is no other bike company in the world doing it at the scale that we are, at that high-end price point.

Conor Doherty: And that’s the configurability that you’re talking about, the optionality that a customer has when they decide to buy a high-end bike?

Dan Scharneck: It is. So, I mean, there’s really kind of two buckets of configurability we offer. So, paint and color choice is kind of the first one, and a lot of people want, if you’re spending that kind of money on a bike, you want it to be yours. So we offer the standard, here’s six really cool colors, we have curated designs from our designers that you can choose, or you can go all the way to just have a color palette and pick which colors go at which part of the bike. So a lot of choice and time can go into the configurability of color. The part selection, so the parts that go on the bike, is where my team spends, and we work with Lokad on, a lot more of the complexities.

So, for those of you who don’t ride bikes and maybe tuning into this, you know, it’s very common for riders to spend two, three hours on a bike ride, all-day events, sometimes it’s multi-day events, so many hours spent on a bike, you really want it to be comfortable and fit you. So, one example I give is, one of our flagship road bikes, there’s like 36 different handlebar options. And it’s not that they’re all that different, but there’s different sizing depending on your shoulder width, you could have different stem length depending on how flexible your hips are, how long your torso is. So all of that is really important to have a bike that is fit for you. And we just think that if you’re a customer spending that kind of money, you should get it straight from the factory that works for you. So that’s the kind of configurability that we offer in the Project One part of the business.

Conor Doherty: Well, it occurs to me, in terms of a business model, that must increase the logistical complexity significantly. I mean, again, it seems pretty obvious, but compared to let’s say just buying a regular high-end fixed bike and a configurable bike that might have, I don’t know, I mean, how many parts are on a regular Trek bike?

Dan Scharneck: Around 150 to 200, depending on if it’s an e-bike or a non-e-bike. But you’re right, the complexity of buying all the parts, waiting for the customer to tell you their configuration, and then delivering that in a relatively short lead time, is a challenge. And that’s where we’ve reached out and started working with Lokad for the last couple of years, to help us make better decisions with all these choices, spending our dollars more efficiently, and just making sure we can actually provide that hospitality to our customers. But you’re right, it’s significantly more complex than ordering 50 black bikes at lead time, putting them on a boat, they show up at the warehouse. And then there’s a market for those kinds of bikes, but at the high end, we really think configuration is the way to go.

Conor Doherty: Well, actually, Joannes, did you want to say something? I’ll kick it over to you now. In terms of the combinatorial complexity that Dan just described, I’m going to presume it’s significant. I mean, 200, 250 to 200 parts, what, 36 separate handlebars? I’m going to presume combinatorially that’s more than 50 or 60 different combinations. So, what’s the order of magnitude there we’re talking about?

Joannes Vermorel: So, I mean, the numbers become very quickly gigantic, but also slightly meaningless because if you have even if you have like 10 options where each option has like 10 possibilities, you get 10 to the power 10, you know, number of combinations. But numbers become so large, it’s a bit meaningless because it doesn’t also truly reflect the actual complexity of the supply chain. So it grows, you know, in a pure combinatorial fashion, it grows completely out of hand, although it is still possible for companies without having billions and billions of inventory to actually cope with that. So, it’s, I would say, we should not be confusing the raw combinatorial explosions, which gives you absolutely gigantic numbers, with the actual growth in terms of supply chain cost, which are growing quite steadily, but not nearly as fast as the raw combinatorial explosion would suggest.

If you go back a century in the past, the idea of having a lot of configurability was much more the norm. If you go back again a century in the past, there were much more options on cars, much more options on tons of pieces of equipment, things you would not suspect. So, you would go to a clerk and they would note exactly what you want, and they would deliver it, it would take maybe six months or a year, and they would deliver the thing that is fairly customized.

In order to make things massively cheaper, pretty much all industries went into a short list of products, and this is it, and you pick a quality point that can be fairly high, let’s say like Apple with the iPhone, so that’s a high quality but still very, very few choices, or it can be very cheap and again very few choices. So, it was not just a matter of quality or no quality, it was more like at a given quality level, by having very few options, it was making it so much easier to scale, and thus that was the dominant strategy.

Now, at the beginning of the 21st century, companies find themselves again with the idea that if you can give more flexibility to your customers, you can deliver a lot more value. And that’s interesting because I think that’s also part of the general trends of the 21st century, which is to use supply chain as a differentiator in terms of consumer experience. One way to do that is to do it e-commerce, so e-commerce is just, well, you can come to our store, or we can send the thing to you. So that, basically taking care of the last mile delivery, is a way to replace a customer problem by a supply chain problem, you just deal with the last mile, the delivery. But configurability is also kind of the same thing, you essentially add a supply chain complexity in exchange for a better customer experience.

I have seen that for Dell, literally conquered the world by having very, very customizable setups for computers, they have been doing that for more than two decades. Companies selling expensive pieces of furniture, like sofas, you know, pick the things, you know, piece by piece if you want, and a corner, a shape, materials, and whatnot, and you end up with 20 different settings for your five thousand dollar sofa. And so, even car manufacturers nowadays tend to gradually reintroduce configurability, although they had, it’s a market that is so insanely driven by prices that it’s very difficult to open up this route. But a lot of, let’s say, the expensive German cars, you would pick a lot of options and then they would just say, well, if you want your Mercedes, it’s going to take you a year to be delivered. So, you can pick all the options and then it takes forever to be delivered. But in some markets, people are willing to wait that long. In some other markets, let’s say bikes, people have expectations that they will be able to enjoy the new bike for, let’s say, the summer, so it should not take a whole year for the products to be delivered.

Conor Doherty: Thank you, Joannes. I’ll turn back to you, Dan. Is that more or less the pain point or what exactly was the challenge that Trek had in terms of delivering the configurable bikes?

Dan Scharneck: Yes, I think Joannes hit the nail on the head. Offering the choice is really good, but then there are two really important things. One is delivering it in a reasonable time frame. We aim for 30 days or less anywhere in the globe for a bike. In a lot of markets, the summer riding season is pretty short, so even 30 days could be a long time. And then you need to reliably tell them when they’re going to get their bike. Both of those things, I think we’re improving on with a lot of the work we’re doing with Lokad. There are a lot of things out of our control that happen in the supply chain, but yes, I completely agree. Fast or relatively fast delivery for a custom product and reliability are key.

Conor Doherty: In terms of determining, let’s say, inventory levels for all of these parts, historically, how did Trek manage that?

Dan Scharneck: This gets into a little bit like how we started looking at maybe there’s a better way to do things, which we ultimately found with you guys. We at Trek have an ERP system that runs the whole business and then we had a forecasting software for finished goods planning. Neither of those things are really created for the complexities inherent with a custom business. We were using the tools we had, right? We were using the ERP system and bolting it on with the forecasting for finished goods. Anytime we tried to do something for the custom business, it was custom work back into an ERP system. Most companies or experiences you guys have had, no one wants to do that. They do not want to customize a standard business. So, trying to find something that was customizable for our business led us eventually to Lokad. I think that’s been the last couple of years actually at this point as we’ve been building out the tools in the process.

Conor Doherty: Thank you. And actually, Joannes, on that point, when Dan talks about the brittleness or maybe the limitations of an ERP for this kind of complexity, I’m sure you can expand on that quite a bit.

Joannes Vermorel: Indeed, ERPs in general, with a transactional database at their core, are just not the sort of tools that are appropriate for most analytical processing. This is a general consideration that is quite independent from the specific problem that Trek had with regards to configurability. It’s not going to do much in terms of analytics.

Where the problem is really magnified is that what you can typically get in sort of tools would be a classic forecast. So, you start with your finished products and then you expand according to your bill of materials. But what do we have here? The notion of finished products to be forecast when you have configurability becomes relatively elusive. If you take it the naive way, then you have a near infinite amount of options. The vast majority of all the possible configurations are never going to be sold ever.

The alternative is to look at time series from the parts angle and it works a little bit better. But the problem with forecasting at the part level is that it doesn’t tell you anything about bottlenecks. You do your forecast but then because a bike has something like 200 parts, the odds are super high that even if you have super high service levels on all parts, like let’s say 98% service level, for a bike with 200 parts, that means that there are four parts on average that are missing every single time.

If you approach the problem saying well instead of having like 98% service level we want to have like 99.9%, the reality is that to get those last percent of service level, the amount of stock just completely explodes. And that’s not a good move. So you end up having a mountain of stock. And in the bike industry, there is technological obsolescence. There are new technologies coming out. So when you have a part, it’s not something that is going to last decades. It’s going to last maybe one season, two seasons, and then there will be the next generation that is slightly better.

So, bottom line, if we go back to this ERP, you can’t forecast the finished goods. You can forecast, time series-wise, the parts, but then you end up with something where if you end up putting very high service levels, even if you go for something like 99% service level, which is exceedingly high, you still end up with, on average, for all of your bikes, a few parts are missing. And then, 99% is not a good place to be when you have a bit of technological obsolescence. That would mean that every single year you would generate large quantities of dead stock.

In the specific business of Trek, we have to think that Trek is selling very high-end bikes, very high quality. The people who go to Trek, they don’t chase promotions. So the idea that you could just liquidate your unsold parts by doing promotions, it kind of conflicts with the idea of being prestigious, being on the very high end.

What Lokad does here is we are using probabilistic forecasts. The probabilistic forecast is the way that gives us the idea that for every single part, we can assess what is the risk of this part becoming a bottleneck and for how long. So the question is, is this part just going to be short of one unit or short of one unit for this amount of days? The interesting thing is that we can suddenly see the risk not from a one-dimensional perspective as in how many units are missing, but how many units/days of delays am I creating in my supply chain, which is a different way to look at it but a much more powerful way.

Conor Doherty: That’s a lot of information.

Dan Scharneck: Yes, it is. But it actually segues wonderfully into how exactly Trek transitioned from a traditional ERP to everything Joannes just said.

We talk a lot to other parts of the business about what we’re doing with Lokad and I get a lot of head nods and kind of blank stares and not a lot of questions. So it means they either really understood everything we’re doing or they have no idea what the path we’re going down with probabilistic forecasting is. I think it’s the latter. We do have to still educate even people within our own business that this is a new way, a different way of doing things.

In business, Joannes is right. We were doing time series forecasting at the part level and we often did have a 98 percent service level on all the parts. This is kind of the worst place you want to be because you have all this money tied up in inventory, your customer can’t get the bike they ordered and just no one’s happy in that situation.

I think we’re seeing better decisions being made by the team. We’re seeing the results starting to get things back on track. Recovering from the pandemic has obviously made it a little bit harder and will take a little bit more time, but having different decisions based on probabilistic forecasts with Lokad has been huge for us.

Conor Doherty: Following up on that, when you see you kind of alluded to the day-to-day dealings with probabilistic forecasting. Without going into incredible and possibly even confidential layers of detail, what does that look like on a daily basis when you’re dealing with inventory levels on high-end configurable bikes?

Dan Scharneck: We try to break it up because even with Lokad, you’re getting a lot of decision recommendations on a daily basis and we have a team of people that are in charge of the forecasting and buying. They can’t get to it all every day, so we break it into a weekly, monthly process.

The really good thing with Lokad is that it floats to the top like, “Hey, here are the biggest opportunities or risks, things you should look at.” And then you just keep working through the list and it changes every day. We get new data daily, which is really important. We take new orders from around the world every single day, so you do want to have that data pipeline refreshed.

I think just having the team be able to focus much more time on evaluating the business decision recommendation has been a huge change for us. We used to spend the majority of our time just trying to come up with a decision we think we wanted to make and a lot of it was Excel based. So, big time shift of the team versus trying to generate the numbers themselves first. Now, it’s generated in Lokad and now we’re able to evaluate it, review it, work with suppliers and spend a lot more time on value-added activities.

Conor Doherty: Thank you. And do you want to, I mean we at Lokad, we apply probabilistic forecasting to a wide array of different tasks: retail stock allocation, prioritized inventory replenishment, configurability. In terms of let’s just even take those three separate topics, how complex is what we do with Trek compared to retail stock allocation or prioritized inventory replenishment?

Joannes Vermorel: What is done with Trek is, I would say, mid-scale in terms of our own scale of difficulty. It’s not as complex as let’s say Aviation where, let’s say you have aircraft engines that just an engine is going to have like 15,000 parts and then an aircraft itself is like 300,000 parts. And then you have to deal with the fleet plus the fact that your client has aircraft parts in 200 airports worldwide. So that’s that’s, and the parts fly through aircraft as well all the time. So they are very mobile because they can use the spare capacity of their own fleet to move stuff around cheaply, which they do.

Nevertheless, it is a lot more difficult than let’s say fashion where you don’t have those sort of dependencies. You do have cannibalization but it tends to be soft relationship between products as opposed to hard dependencies between the parts where if there is one that is missing then you’re stuck. So that’s a mid-level complexity which is again quite a challenge because we are talking of a business that even if there are high-end bikes, the prices are still quite competitive.

An aircraft engine, just to give you a scale, an aircraft engine every blade of the angels that you can see spinning would be something like fifteen thousand dollars a piece. Just one blade. So obviously when every single piece and the keyboard in an aircraft cabin is like twenty thousand dollars. So when every single part counts in ten thousand dollar plus, you can incur crazy supply chain overheads and it’s still okay. In the case of Trek, they are still quite competitive. I mean they are the high end but nevertheless, this is not the same order so that’s why I say yes it’s middle complexity but the cost factor is much more prevalent than what you would observe let’s say in Aerospace.

Another factor is that through holistic forecast you can approach the problem through multiple dimensions. For example, ordering is just one of the type of decision that you can take. You can also, without going into the secrets, Trek has a close relationship with many of its suppliers and thus orders can be slowed down or accelerated. So that’s another type of decision. It’s not just a quantity, it’s that when you have a good relationship with a supplier you can modulate also orders that are already in progress to accelerate them a little bit or slow them down a little bit. So that’s also another dimension where it’s not just about a fire and forget, there is a bit of ongoing supplier relationship management.

And then when you have also multiple locations like Trek has, there is an element of inventory rebalancing between the locations that happens. So again that’s interesting because when you look at it from the lenses of the classic forecast it’s just time series, you’re supposed to know the future. All those things don’t even make sense. I mean if you know the future why would you ever rebalance? You know the future so you just have to orchestrate the execution, you don’t need to discuss with your supplier what is in progress, you don’t need to rebalance your stock, you just have your one perfect plan and you just orchestrate it.

But when you go into those probabilities where things can gradually evolve, where stuff as a situation evolved with every extra order, you can have something that gradually bubble up to the top and become the new hyperative things. It’s not like you have a change and then everything is completely reset with completely different priorities. But it means that everything that happened in your supply chain is going to nudge the rank level, the rank of every single thing that you have in your outline. Some stuff are going to move up toward the top or downward toward the bottom and with the full probability forecast you have a much more incremental expense as opposed to something where you refresh the forecast, you get different forecasts and then all your planning has to be redone from scratch with the new forecast.

At Lokad, we can recompute new probabilistic forecasts and then very frequently people when we say we recompute new forecasts on a daily basis, they say “Oh, it should drive every practitioner completely crazy because suddenly all the planning is done completely differently.” But the reality is that when you go into those sort of prioritization sort of view, when you recompute the forecast it just means that there is every day the ranks of everything just shift a little bit up or down and the changes are nowhere as drastic as in the classical perspective when you recompute your Classic Time series forecast.

Conor Doherty: When you’ve dealt with a classical perspective as Joannes just described for so long and then you introduce probabilistic forecasting, a purely financially driven perspective, embracing the uncertainty of future demand, how exactly do you sell those concepts to people who are hearing them for the first time? They’re not like you, they don’t understand these things immediately. How does that conversation go?

Dan Scharneck: I was trying to allude to it a little bit earlier. We do talk about it internally to other parts of Trek and I still think there’s a pretty big barrier to understanding. The Classic Time series is what you learn in school, it’s what you were taught when you had your first job. It’s really what’s rooted in supply chain forecasting, purchasing business.

I do think with AI being everywhere now over the last few years, whether it’s good or bad, people tend to be more perceptive to “Oh, you’re doing a machine learning based thing. I’ve actually kind of heard about it. I think what I’ve heard about is what you’re doing.” So I do think that’s helped. But it is going to be showing the results, it’s going to take some long-term performance of the business and then I think that’s how we’ll start seeing more people picking it up internally to Trek or other industries we talk to and they’re interested in what we’re doing as well.

Conor Doherty: Joannes, is that how you explain it when you’re not at work, you know, just at a dinner party? Is that how you sell this to people?

Joannes Vermorel: Yes, I mean the gist of it is that as soon as you give up on the idea that you know the future, it becomes a lot more complicated. It would be so much easier if the future was just the symmetric of the past. If you could just have time series for the past, which is the correct way to look at the past, you can look at those time series for the past. They are what happened, there is no uncertainty. There might be a clerical error here and there where somebody miscounted a part in the historical records, but the odds of this are super low. Companies that have correct processes with barcodes and whatnot, the sort of clerical error rates we are talking about is less than one over a thousand. So essentially, the clerical errors are almost non-existent, so your time series are almost 100% accurate historically speaking. And then you go into the future, and then it becomes massively inaccurate, like 50% inaccurate. And also, there are even contingent on decisions that have yet to be made.

For example, this, and I’m not disclosing things that are secret because you can see that through the Project One configurator online, but I’ve checked this configurator over the month and it changes. There is novelty, things are not presented exactly the same way. So it’s not like there is a configurator and it’s an immutable offering of Trek. It’s something that in itself, every month there is something that comes to this configurator or some stuff are phased out as well.

Obviously, if Trek decides to modify their offering, so what is being proposed in this configurator, it is fairly evident that it will have an impact on the future demand. But if you look at the future through the lenses of the time series, where do those evolutions take place? The answer is nowhere, they can’t even exist in those transitional time series.

Thus we have uncertainty, but we also have another problem which is we have the contingency on decisions that have yet to be made, such as the evolution of the configurator.

It is unfortunate, but as soon as you say well we are going to deal with this uncertainty, we are going to deal with those evolutions of the configurator, yes it becomes more complicated. But I would say it’s the price to pay to deal with the actual complexity that pre-exists Lokad as opposed to pretending it doesn’t exist.

I really differentiate when I say it’s more complex or complicated. Is it a vendor that has not made enough effort in his technology so the vendor has a super complex technology, but it might be just accidental complexity. There is no value, it might just be Lokad may have just over-engineered everything and so everything is overcomplicated for no good reason.

Or is it like essential complexity? The complexity is in the business, it pre-exists Lokad. Uncertainty exists whether we want it or not. For example, the biking industry went through a boom during the lockdowns and then there was the exit of this boom. Lokad didn’t cause that, it just is what it is.

The evolution of the configurator or something, I know that Trek follows very closely a lot of sport events like Tour de France in France. And so there is this sort of evolution of the offering that kind of matches those sports events. Again, this complexity pre-exists Lokad.

The way I typically present that, that’s the long-winded answer to your question, we didn’t shoot first, you did. I mean, this complexity is kind of on you, but also it is a good complexity because it’s complexity that creates value for the customers.

Having an offering that evolves with the sport events, I believe that biking fans, racers, they love this sort of things. It’s relevant in this market. And same thing, having lead times and whatnot, it is again the clients, they love that. But it creates a lot of complexity and that has to be addressed.

And thus, probably forecast is just a mirror of this pre-existing complexity and as opposed to just pretending it doesn’t exist, we face it. And yes, it’s a bit complicated, but I think it’s not that complicated.

To be honest, it is a bit, but at the end of the day, I believe that the list of priorities, it’s not something where you need a PhD in math to kind of understand what we recommend. At the end of the day, it’s a prioritized list of what you should order, the touch points you should have with your suppliers, the stock rebalancing that needs to happen most urgently.

Underneath, the calculations are very complex, but the reality is people have even forgotten about that. When you just do a multiplication of floating point numbers in your computer, there are fairly advanced things that are happening under the hood. For example, very few people would have a clear view of what is the role of the mantissa when you do the multiplication of floating point numbers. Most people would just say, what’s the mantissa? Well, it turns out that it’s one of the key ingredients if you want to do multiplication of floating point numbers. But the reality is that it’s the technicalities and as the end user, you don’t really care, you just get the end result and that’s it.

Dan Scharneck: I just tell people we save them money.

Conor Doherty: That’s what I say. All right, well Dan, I’m mindful of your time, so as this customer here, we like to give the last word to our guests. Is there any advice you would give people who are struggling with configurability or anything you really want to say at all on supply chain?

Dan Scharneck: Yeah, I mean, I’ll put a little plug in for Trek. If you’re a bike retailer and you’re looking to sell configured bikes, you should become a Trek retailer and sell Project One bikes.

If I guess outside of the biking industry, I mean, we’ve been doing this custom thing for a while, so I would say the first thing is really understand what value it adds to your customer depending on the product you’re providing. We have a really good product team here at Trek. We’ve also at times offered things that didn’t go over as well with our customers, so I think it’s just really important to understand before you go down the path, what does configurability offer.

The second one is, I really do think it’s important to have good systems to help you manage the complexity, and that’s for us, that’s become Lokad now. We didn’t have it when we started and just as the business has grown, it’s really become a necessity for us. And whether it’s Lokad, which we feel strongly it is, or it’s someone else, you just need to have a strong foundation of your systems.

And the last one is, you really do need a supply chain that can be reactive and can be set up to be responsive to a custom business because you are going to get it wrong. That’s kind of the nature of it, nothing’s ever 100% right. And I think if you do those three things, you can have a pretty good chance at a good product in the configured space.

Joannes Vermorel: I agree.

Conor Doherty: All right, well I think I’ll draw things to a close. Joannes, thank you very much for your time. Dan, thank you very much for yours and thank you very much for watching. We’ll see you next time.