00:00:07 Forecasting in the luxury market and its unique characteristics.
00:01:45 Redefining forecasting for luxury items and addressing supply chain challenges.
00:03:15 Optimizing assortments in luxury stores and the role of store managers.
00:06:28 Quantitative optimization and transferring insights across stores.
00:07:52 Pricing and positioning of luxury products in the market.
00:10:22 Challenges in luxury product pricing and avoiding sales.
00:11:59 Addressing diverse markets and strategies.
00:13:29 Lokad’s approach to intermittent demand in the luxury industry.
00:15:11 Assortment and inventory optimization for luxury brands.
00:16:52 Importance of visual presence for luxury items and the cost of items being out of sight during transfers.
00:17:38 Forecasting demand in luxury brands through statistical analysis of past consumer behavior.
00:18:54 Steps luxury brands can take to improve their approach, focusing on quantitative optimization.
00:20:34 Closing thoughts.
Summary
In an interview with Kieran Chandler, Joannes Vermorel, the founder of Lokad, discusses the challenges of forecasting and supply chain optimization for luxury brands. Traditional time series forecasting methods are not suitable for the luxury market due to low sales volumes and intermittent demand. Vermorel highlights the importance of assortment optimization and inventory allocation, which requires determining which products should be displayed in each store. Lokad focuses on forecasting demand by analyzing the preferences of a brand’s existing clientele. Luxury brands need to embrace quantitative optimization, which has become possible due to breakthroughs in modern statistics and technologies like deep learning and autonomous vehicles.
Extended Summary
In this interview, host Kieran Chandler speaks with Joannes Vermorel, the founder of Lokad, a software company specializing in supply chain optimization. They discuss the unique challenges of forecasting and supply chain management within the luxury market.
The luxury market is characterized by high-value items like jewelry, precious stones, expensive bags, and watches. These items typically sell at a rate of one or two units per product per store per year. This low sales volume makes it difficult to apply traditional time series forecasting methods, as there is little historical data to base predictions on.
Instead, luxury brands must approach forecasting differently, focusing on factors like product introduction, pricing, positioning, and assortment optimization. The primary challenge for these brands is deciding which products to introduce, as they must rely on the creativity of their designers to create items that will resonate with their target audience.
Once a product is selected, luxury brands must optimize their assortments within each store. Unlike supermarkets, luxury stores are often located in high-value urban locations with limited space, making it impossible to carry the entire collection. These stores may only have room for a small percentage of the brand’s catalog, which means that each item must compete for valuable placement within the store.
To optimize product assortments, luxury brands must consider factors like diversity and the balance between high-priced showpieces and more affordable, accessible items. For example, a store might carry a €100,000 watch as a centerpiece, but the majority of sales might come from watches priced around €10,000. The goal is to create a compelling assortment that maximizes the return on investment while showcasing the brand’s unique offerings.
Vermorel explains that assortment optimization for luxury brands is a challenging task. Most luxury brands currently rely on store managers or similar personnel with lifelong expertise in creating a beautiful shop. While this expertise is essential, it does not scale well, making it difficult to replicate across multiple stores in various locations. Vermorel suggests that quantitative optimization, using statistical techniques, can help translate insights from one store to another. Although geographies and markets differ, there are still aspects that can be statistically transferred. The complexity of luxury brands, with their diverse assortments of products, requires advanced statistical tools to optimize their performance effectively.
Regarding pricing, Vermorel states that luxury brands approach pricing differently from regular fashion brands. True luxury brands never discount or run promotions because it would damage their image and perceived value. Customers who spend a significant amount on a luxury item do not want to see it discounted shortly afterward. Pricing for luxury brands aims to create sustainable, lasting value. For example, prestigious watchmakers have positioned their products as investments that may grow in value over time. The challenge for luxury brands is to set prices for new products that will help grow their market while increasing perceived value, allowing them to charge more in the future. This virtuous cycle of growth can lead to even more desirable products, further enhancing the brand’s reputation and success.
Luxury brands often avoid sales or promotions to maintain the exclusivity and high value of their products. To manage excess stock, they sometimes resort to destruction or recycling of materials. For instance, in the case of jewelry and watches, precious metals and gemstones can be recycled into new products. In the case of clothing or leather goods, destruction may be the only option.
Vermorel emphasizes the importance of getting the initial pricing right from the start, as once a product is sold, brands cannot reduce its price without risking customer dissatisfaction. Raising prices, however, can be a viable strategy to enhance perceived value. He also discusses the challenge of quantitatively optimizing the pricing and assortment strategy while considering the vast diversity of products, markets, and trends.
Lokad’s approach to addressing the complex and sporadic demand patterns in the luxury goods market involves leveraging statistical tools and techniques. Vermorel explains that when dealing with low sales volumes per product per store, having a large number of stores allows for more accurate statistical analyses. Additionally, luxury brands should consider the repeat customers who make up a significant portion of their sales.
Lokad has developed models that consider the relationships between stores, products, and clients in order to make better-informed decisions. These models help luxury brands make choices regarding the assortment of products to keep in each store and how to manage inventory effectively. The primary challenge for these brands is not deciding on the number of units to stock, but rather determining which products to offer in each store to maximize sales and maintain customer satisfaction.
Vermorel explains that assortment optimization is the process of determining which products should be on display at a given location. The main question is whether a product should be available at a particular store or not. Inventory allocation is another important aspect, as luxury brands may produce fewer units than the number of stores they have. For example, a brand may have 50 stores but only produce 20 units of a specific product. In this case, they must decide how to allocate these units among their stores.
A common pitfall of manual allocation is that brands tend to favor their best-performing stores, leaving others without the new product. This can create a self-fulfilling cycle where some stores consistently outperform others. Vermorel suggests that brands should be smarter in their allocation, considering the possibility of moving products from one store to another if they do not sell within a specific timeframe. The main cost associated with this approach is that the product will not be on display during the transportation period, which might affect demand generation.
Luxury brands must recognize that their stock generates demand, as customers are often triggered to make a purchase upon seeing the product. Vermorel emphasizes that Lokad is indeed forecasting demand, but in a very specific way. The company analyzes the preferences of the brand’s existing clientele to optimize the product assortment in each store. This is different from time-series forecasting, which involves predicting a continuous curve based on past data. In this case, demand is too intermittent for such an approach to be effective.
In order to improve their expertise and adopt a quantitative vision, luxury brands must first recognize that quantitative optimization is possible. Until recently, there were no suitable tools for number crunching in this sector, so brands had to rely on manual processes. However, breakthroughs in modern statistics and technologies like deep learning and autonomous vehicles have made it possible to deliver quantitative optimization for luxury brands.
Full Transcript
Kieran Chandler: Today on Lokad TV, we’re going to discuss a little bit more about this lucrative world and understand whether you can forecast for those items that you truly desire rather than actually need. So Joannes, we obviously know that the luxury market is very different from some of the more classic markets we work with. But what makes it so unique from a purchasing perspective?
Joannes Vermorel: There are many things that make luxury very unique. First of all, indeed, you are dealing with super high-value items. I mean, I’m talking about jewelry, precious stones, expensive bags, and watches, which are the core luxury items. Typically, just to give you some context, you end up selling something like one or two units per product per store per year. So obviously, this is not exactly the ideal candidate for, let’s say, time-series forecasting because when you’re looking at a product that is only going to be on the market for two years and you’re selling like one unit a year per store per product, it literally makes no sense to say we are going to have a time series that we intend to extend into the future. So if we mean the classical time series perspective by forecasting, then it’s game over; you just can’t do anything that is of practical interest. That’s maybe one of the key takeaways: we need to redefine in-depth what we are trying to do when we say forecasting as far as luxury items go.
Kieran Chandler: So typically, in a supply chain, we might know what products to stock and where to put them. But what is the approach that luxury brands take to that?
Joannes Vermorel: The first key idea is that the primary challenge is to even decide what is the next product that you want to introduce. Yes, you can say it’s all about the designer’s inspiration. Absolutely. I mean, if you don’t have the best designers to craft tremendously fabulous items that people are going to fall in love with and buy, you’ve lost; you cannot even be in this world. But once you have that, there are still small adjustments that you can do in terms of pricing and positioning to maximize your return. The first question is how do you quantitatively optimize what you introduce into your network? And then, as soon as you decide that you have this tremendously fabulous item that you’re going to push into your distribution channel, how do you optimize your assortments?
You see, if you think in terms of service levels or fill rates, luxury brands don’t make much sense, just because stores are typically located in incredibly valuable locations in metropolises. The store is typically way too small to hold the entire collection. So this is not a supermarket where you can have a 90% or 97% service level. At most, you can have a store that can hold, let’s say, 500 items on display, with a catalog of 5,000 items. So, if you think of it in terms of service level, you have a 10% service level, but it doesn’t make any sense to think of it that way. The way you should think of it is that all your items literally compete for those extremely valuable placements in the store. So there is a real problem of assortment optimization. Obviously, you need some diversity. You need to have products that are only there for the show. For example, you might have a watch that is above 100,000 euros
Kieran Chandler: You know, a piece that is like a masterpiece but really speaking, maybe most of the expensive watches that you’re selling are more like in the ten thousand euro range as opposed to the masterpiece that is very, very expensive. So let’s talk about that product assortment. I mean, we obviously know they’ve got very large stores which seem to have next to nothing in them, so how do they actually know what items they should put on display?
Joannes Vermorel: For most luxury brands, it’s still a completely manual task done by a store manager or someone very close to a store manager. It boils down to pure, lifelong expertise in crafting a beautiful shop, and I’m not denying that it’s very important and that this expertise does matter. The problem is that it scales relatively poorly. You might find a person that is like a genius at managing one store, and this person might actually be also managing a couple of other stores that are nearby, but it’s very hard to replicate that at scale into, you know, 40 countries, 40 large cities in the world. And then you might have people who are good at some point in their life, and then they are not so motivated anymore, they’re not so good anymore. How do you know if a store is performing super good just because it’s super well placed or because the store manager is the secret ingredient behind the success of the store? So I’m not denying that there is expertise involved, but then how do you make that super replicable and scalable? Our answer is quantitative optimization. The key insight is that if you have the right statistical techniques, what you see in one store can be translated to what is happening in another store. Not entirely, obviously, geographies differ, the market in South America is not the same as the market in Dubai, but still, there are things to be statistically transferred from one situation to another. What makes luxury brands so complicated is that the assortment of the products you’re selling typically is very diverse. So you have your 5,000 products and you have maybe your 100 stores, and it happens that there are not two stores that have the same product at the same time. So, again, classical statistical tools just do not work well or do not work at all for this situation. But with the proper tools, you can literally do quantitative assortment for luxury brands, which is really maximizing the value that you generate with all your assets, which are fantastic design on one side and those fantastic locations on the other side.
Kieran Chandler: And you mentioned a bit earlier about pricing. How do these luxury brands go about determining where they should be pricing their products and where they should be positioning themselves in the luxury market?
Joannes Vermorel: Again, pricing in luxury brands has a very different meaning and angle than what you will find in regular fashion brands. For example, if you have a true luxury brand, you never do discounts, ever. There are no promotions ever. Why is that? Because if you’re buying a twenty thousand euro watch, you do not want in the next six months or even years to see that this brand, this watch is now on promotion at fifteen thousand euros. That’s unthinkable. No, it cannot happen ever. So pricing is not about yield management. Pricing is about creating super sustainable, super lasting value for your items. We’re literally, and that’s what the most prestigious watchmakers are achieving, saying that when you’re buying a Rolex, this is not spending money; this is an investment. Chances are, ten years from now, it will even have grown in value, which sometimes actually happens, which is a bit insane because it’s not a productive asset, but
Kieran Chandler: Nonetheless, this is the reality of what they have achieved. So, really, the pricing should be understood as how do I take the price and, when I introduce a new product, choose a price that will help me both grow my market but also grow in a way that is steering my prices upward because people see more value in the products and are willing to pay more. So, and again, in turn, if people are paying more, then you have more resources to make the products even more incredible. That’s kind of the virtuous cycle of those extremely successful brands that manage to grow into things that were even more expensive while actually making the products even more desirable in the process. And you mentioned that there’s no sort of sales or promotions, so what do these luxury brands do to get rid of excess stock and have turnover of that stock?
Joannes Vermorel: Worst case, they destroy it. For example, if you’re dealing with special cases such as jewelry or watches, you can literally recycle the precious metals. The thing about precious metals is that they are 100% recyclable. Precious gemstones can just be taken from one piece of jewelry to another. So, basically, you recycle the precious materials and produce a new product. If you’re into clothing or leather stuff, you have a few options, but mostly it has to be destroyed.
Kieran Chandler: So, basically, what we’re kind of saying is these luxury brands have to gamble and get the initial pricing right straight from the start?
Joannes Vermorel: Absolutely. When you push a product to the market, you’re kind of stuck with this retail price because as soon as you sell the first unit to the first client, this client should never come back to the store and see that what they have just purchased is now on the cheap. Actually, quite the opposite, you might raise your price because then people come back and say, “Oh, I did a really good job. This bag was 800 euros, now it’s a thousand euros.” In this direction, you can do it, the other way, no, you can’t.
But fundamentally, the question is how do you quantitatively optimize your strategy? You can say, “Well, I’m just going to do guesswork based on a lifetime of expertise.” That’s good; that’s how the most successful brands have been doing it. But if they start competing against people who can combine a lifelong expertise plus tools to do quantitative optimization because, again, the problem is that as you grow and you end up with thousands of very diverse products, markets are changing faster. This is fashion; all sorts of things change in terms of trends. I mean, all of that is a lot of information to process. You have many different markets, and how can you expect to have a strategy that is consistent and capable when you’re both trying to address the market in South America, South Asia, North Europe, etc.? So, there’s a limit to what the human mind can do in terms of optimization.
Kieran Chandler: Okay, let’s move on to that numeric optimization then, and what is Lokad’s take on the industry? It seems there’s so much randomness, and sales demand is very sporadic. So, how can Lokad actually do anything here?
Joannes Vermorel: Indeed, demand is very intermittent. We’re talking about selling one or two units per store, per product, or something, sometimes a little more, but not much more. So, you need to have the proper class of tools that can extract information from this sort of situation. The key insight is that if you had only one store, and for
Kieran Chandler: You need a year. You will not be able to do anything statistically speaking, but chances are that you have, let’s say, 50 stores, so those products actually in aggregate, you are doing more and you can do things in statistics because you have more stores.
Joannes Vermorel: Also, you should not think of what you’re selling as one point in the time series. You know, the client, you have the identity of the clients. Usually, it might come as a surprise, but luxury brands are all about repeat customers. It’s not people who just buy one expensive watch in their life. No, the people who buy a lot of beautiful watches are going to buy one more watch every couple of years. So it’s like repeat customers even though it’s not like weekly shopping. The proper model that we have developed, for example, is leveraging the fact that you have a graph that connects the stores, the products, and the clients, and that’s a lot of information to leverage, statistically speaking, to make better decisions.
Kieran Chandler: What kind of decisions are we talking about?
Joannes Vermorel: There is a choice of assortment: what product to put, what are the products that you need to keep in every store knowing that it’s not about deciding whether you have like five or ten units. No, usually the problem is, do you have one unit of this product or zero? You know, it’s just because if you’re selling the product, you can restock the next day. It’s no problem. So the question is assortment optimization. Do I want to have this product on display at this location? And then there is the question of inventory allocation. Again, you’re producing a soft number of units. That’s usually frequently what happens is that you’re not even producing one unit per store, so you might have 50 stores, but you’re only producing 20 units for this product, so you decide how you allocate the units.
And then there is a curse when you do it manually, that you tend to say, “Oh, my best-performing store, let’s say my store in Dubai, is always going to have one of those 20 units.” But when you do that, it means that some stores are radically more appealing and obviously more successful than others. But then you might have your store on Champs-Elysées that is performing not nearly as well, but also the store is almost empty when it comes to allocating those 20 units to the 50 stores. The store on Champs-Elysées is always the store that is not served by this new product being launched on the market. So there are some challenges that you want to solve by being smarter.
And also, if a product does not sell in a store for a couple of months, maybe it’s time to move this product from one store to another. And again, because you have expensive products, transportation costs are not so much an issue. The real cost, actually, what we have measured, is that when you decide to move a product from one store to another, during the transportation time, the product is not on display anymore and thus cannot generate demand. And that’s also something that is very specific of luxury brands.
Kieran Chandler: Luxury brands have stocks that generate demand because people can see the product and it triggers their willingness to buy. So, are we not actually forecasting future demand, but rather optimizing product assortment and the goods that are being sold?
Joannes Vermorel: Actually, we are forecasting the demand, but in a very specific way. We take the perspective of all the clients who have already bought from the brand’s network and statistically analyze what they are looking for. This allows us to optimize, in every single store, what should be put on display to satisfy the desires of the client population. The forecast is still a statistical forecast in the sense that you’re making a numerical statement about the future, but it’s nothing like a time series forecast, like the weather forecast, where you have a time series that goes up and down and you’re trying to predict the curve. The demand is too intermittent for that, but it doesn’t mean that you cannot extract any information from what you’ve observed in the past.
Kieran Chandler: To conclude, what steps can luxury brands take in order to work on their expertise and improve on that, while also taking into account a quantitative vision?
Joannes Vermorel: I think the first step is to start considering that quantitative optimization is even possible. Most luxury brands have been doing things manually, mainly because there were no tools capable of doing number crunching that worked for them. However, the world is changing, especially in terms of technology. There have been breakthroughs in modern statistics, the sort of things that power deep learning and autonomous vehicles. It may seem unrelated, but it’s actually connected, as it’s the same breakthrough in fundamental statistics that are at play. Suddenly, it’s possible to deliver quantitative optimization, which means a lot of things, such as rebalancing your items within your store network every single day. This can be achieved in a way that is beyond what can be done manually, as if you can spend an almost infinite amount of human thinking hours for every single product, considering the entire history of every single client that has ever visited the store and the subsequent demands.
Kieran Chandler: Great, let’s leave it there. Thanks for your time this morning.
Joannes Vermorel: Thank you.
Kieran Chandler: Thanks for joining and tuning in. We’ll be back again next time with another episode, but until then, thanks for watching.
