00:00:07 Fashion industry supply chain intro.
00:00:34 Traditional fashion industry approach.
00:01:33 Fashion industry product lifecycle.
00:03:49 Dispatching and inventory management issues.
00:06:36 Ultra-fast fashion and quantitative supply role.
00:08:01 Decision-making and automation potential.
00:10:12 Fashion industry automation in production.
00:13:36 Warehousing, transport, sales in fashion.
00:13:53 Price elasticity, inventory liquidation, customer behavior.
00:15:18 Pricing decisions via supply chain optimization.
00:16:25 Waterfall process and steady supply benefits.
00:18:08 Handling demand spikes and celebrity influence.
00:20:47 Probabilistic forecasting and risk adjustment.
00:23:01 Real-time inventory management, future of social influence.
In an interview, Joannes Vermorel, founder of Lokad, discusses supply chain challenges in the fashion industry and the value of a quantitative approach. Vermorel critiques the industry’s traditional waterfall system and highlights the complexities of managing product life cycles, distribution, and inventory liquidation. He emphasizes the importance of automation, software-driven processes, and end-to-end optimization to expedite decision-making and compress lead times. Vermorel also suggests transitioning to probabilistic forecasting, gradual phasing between seasons, and small, frequent shipments. He acknowledges the impact of social media on the industry but notes that many fashion companies struggle to maintain real-time perspectives on stock levels or predict interactions with celebrities.
In this interview between Kieran Chandler, the host, and Joannes Vermorel, the founder of Lokad, the discussion revolves around the supply chain challenges in the fashion industry and the value of a quantitative approach. The interview begins with the host citing Edith Head’s quote about the importance of dressing and acknowledging the significant value of the global fashion industry.
Vermorel, a specialist in supply chain optimization, dives into the traditional approach employed by the fashion industry. He notes that the industry operates around the concept of collections, typically releasing four each year. Every process, from merchandising and purchasing to supply chain management, is an enormous waterfall system. This system starts with the composition of the assortment and ends with sales to liquidate the remaining inventory, preparing for the next collection.
Vermorel explains the standard product life cycle in the fashion industry, highlighting the inherent challenges. Each collection involves the introduction of new products, requiring a balance to prevent an oversupply for a particular market segment. The quantities to order from suppliers, often located in distant countries, need to be determined well ahead of the collection’s launch date. The suppliers then manufacture the goods, which must arrive on time for the collection’s start.
Further complications arise due to a myriad of nonlinear constraints. For example, suppliers might impose minimum order quantities (e.g., at least 300,000 meters of fabric per color), affecting multiple products simultaneously. Once manufactured, the goods are shipped, introducing additional nonlinear constraints, like the maximum number of cubic meters allowed in a container. Vermorel mentions the option of air shipping some products, despite the higher cost, to expedite their arrival when necessary.
Vermorel first delves into the balancing act required in managing product distribution to stores. Too many units at once can overwhelm store staff, while too few units can hinder the display of a new collection. The aim is to maintain optimal stock in central warehouses to meet dynamic demand. Failing to balance this can result in some stores running out of stock while others have an excess of the same product. He also mentions the issue of inter-store inventory redistribution, which is often too expensive for affordable fashion brands. He then segues into end-of-life cycle sales, which are intended to liquidate excess inventory and make room for new collections.
The conversation then pivots to the trend of ultra-fast fashion, which boasts lead times as short as a week from design to shelf. Vermorel explains that reducing lead times necessitates compressing delays, which is where quantitative supply chain can play a significant role. He underscores the importance of automation and software-driven processes to expedite decision-making related to purchase order quantities and other small decisions. By automating these processes, companies can cut down on time-consuming manual steps.
The founder of Lokad discusses the potential for supply chain optimization to help fashion companies evaluate sourcing options dynamically. These include assessing whether it’s worth paying more for faster production or choosing more expensive shipping options to get products to market quicker. Vermorel highlights the importance of end-to-end automation as a starting point to compress lead times.
When asked about the challenges in production and the role of automation, Vermorel makes an interesting observation. The fashion and textile industry was one of the first to be impacted by mechanical production during the Industrial Revolution. However, the industry has remained relatively manual due to the complexities of fashion manufacturing. He notes that while fabric production can be automated, fashion tasks like cutting and sewing are harder to automate.
He cites the steady progression in warehouse automation and its significant productivity improvements. However, with the physical processes becoming more automated, the clerks who make numerical decisions to drive production units and warehouses are becoming a dominant part of the workflow. Vermorel sees the next change in bringing automation to this decision-making process, while high-level strategic decisions, such as brand vision, would still require a human touch.
Vermorel begins by discussing the common practice among fashion brands of liquidating inventory at the end of a collection through sales. This creates an artificial demand, allowing the company to clear out stock. However, he also introduces the concept of raising prices when a stock-out is imminent, which, while not preventing the stock-out, could improve margins. This approach, while valuable, also poses a significant challenge as customers may delay purchases in anticipation of sales. This unpredictability of customer behaviour is one of the reasons Vermorel advocates for quantitative supply chain optimization. He explains that an automated, data-driven approach can efficiently determine whether the price of a product should be adjusted, eliminating the need for a massive workforce to manually track prices.
He further criticizes the waterfall model, where suppliers get massive orders at certain points during the year, causing stress on various levels of the supply chain. This model results in warehouses and stores dealing with large influxes of goods at specific times, which can be difficult to manage. Vermorel instead proposes a gradual phasing between seasons. This strategy includes frequent small shipments, which would be easier for the supply chain to handle, and would avoid the need for sharp discounting to clear out stock.
When asked about the challenges of forecasting for the fashion industry, particularly in accounting for unpredictable spikes in demand, Vermorel acknowledges that such “freak spikes” are statistical outliers and can’t be accurately predicted. However, he suggests transitioning towards probabilistic forecasting, which can account for the likelihood of such events. He gives an example from the sporting goods industry, where a brand might prepare t-shirts in the colors of different teams, but only print the logos after the championship’s outcome is known.
Vermorel also discusses the impact of social media, such as Instagram, on the fashion industry. He views the idea of brands pre-emptively stockpiling items based on anticipated social media trends as currently unfeasible, describing it as “science fiction”. Despite some brands’ success with guerilla marketing tactics, he notes that many fashion companies struggle to maintain a clear real-time perspective on their stock levels, let alone predict the outcome of interactions with celebrities.
Kieran Chandler: Today on Lokad TV, we’re going to investigate the supply chain challenges that affect this industry and understand why taking a quantitative approach means you can be a step ahead in what is a very complex marketplace. So, Joannes, a good way to sort of understand this industry is looking at how it currently operates. What is the traditional approach that these fashion industries are taking to the market?
Joannes Vermorel: The traditional approach is pretty much geared around the notion of collections. You have, let’s say, four collections a year, and your entire process, both for merchandising, purchasing, and supply chain, is a gigantic waterfall. You start by composing your assortment, figuring out the quantities you need, producing and sourcing those quantities, transporting them, distributing, and then repeating the process with the next collection. You’ll have sales to basically liquidate whatever remains of the inventory so that you have a clean slate to start over the very same process.
Kieran Chandler: To illustrate the problem a little bit more, I know it can vary from product to product and from different brands, but what does the standard product lifecycle look like for a fashion product?
Joannes Vermorel: The standard lifecycle starts with the assortment, where you compose your collection. The challenge here is that every collection introduces new products, at least for a sizable percentage of the products. Some products might be repeated from one collection to another, but it’s not supposed to happen for many years in a row, or it’s not exactly fashion. So, you start with the assortment, and you need to balance things so that you don’t end up with too many products for a segment of the market that is too small compared to what you can capture in terms of demand.
Then you need to figure out the quantities that you will order from your suppliers, who are typically located in relatively distant countries. If we’re speaking of supplying the European or North American markets, it’s typically produced in Asia, Eastern Europe, or South America. You’ll have to plan well ahead of time, compared to the start date of your collection, to place purchase orders to your suppliers who will start to manufacture the goods. You will need to plan for the timing it takes for your suppliers to produce the items so that they will arrive on time.
The big challenge is that at the assortment level, you can think product by product, but as soon as you enter the phase where you have to place orders with your suppliers, you have tons of nonlinear constraints, such as minimum order quantities. A supplier might say that you can only place an order if you have at least 300,000 meters of fabric per color that you order. That’s a kind of nonlinear constraint that will impact many products at once.
Then, things get shipped to you, which is the next stage, once they have been produced. You have another set of nonlinear constraints, such as container maximum capacities. You need to optimize your shipments to make the most of the containers. You can also decide to put some products on an aircraft instead of shipping them by sea. It’s more costly, but it typically makes sense to have a small mix shipped by air. The idea is that you want to put more urgent items in air shipments.
And then, items will land in your warehouses, and you will start to think about the dispatch to your various channels. Depending on the situation, let’s consider a case where the brand has its own stores. In this new
Kieran Chandler: On how many units should be shipped to every store, there are many nonlinearities to consider. If you send too many units at a time, the store staff can become overwhelmed, and the store will be a mess for a week until everything is unpacked. On the other hand, if you don’t send enough, the staff cannot make a good display for the next collection. How do you balance all these factors?
Joannes Vermorel: You have to balance many things, and it’s important to keep just enough stock in your central warehouses so that you can respond dynamically to the needs of your stores or different channels. If you don’t manage this carefully, you might end up with some stores running out of stock while others have excess stock for the same products. Unfortunately, in fashion, unless you’re selling very expensive products, it’s typically too expensive to redistribute inventory from store to store. That’s why sales events are there to liquidate excess inventory and make room for the next collection.
Kieran Chandler: Another trend the market is seeing at the moment is ultra-fast fashion, with lead times as short as a week between design and getting the product on the shelf. How can we manage to achieve such short lead times, and what are the key challenges to make this happen?
Joannes Vermorel: To achieve fast fashion, you need to compress all the delays involved in the process. Quantitative supply chain solutions can help a lot in this regard. First, you can reduce the time it takes to make decisions on purchase order quantities. Instead of taking weeks for your purchasing team to decide, you can have a software-driven process that generates optimized purchasing decisions based on the latest assortment, historical data, and statistical analysis, all in the same day.
The same approach can be applied to other small decisions. A large portion of lead time delays come from the time it takes for people to make decisions and the manual steps involved in the process. There’s a significant opportunity for automation and smart decision-making to help reduce these delays.
Once you have end-to-end automation and calculation in place, you can start considering options that further compress lead times. For example, you might choose to produce in Eastern Europe or Turkey instead of Asia, even if it’s more expensive, because it will get the product to Western Europe faster. To make this decision, you need a system that can tell you when it’s worth paying more for a faster production time. The same logic can be applied to transport options as well.
Kieran Chandler: Whenever you have something that can be transported by sea, there is the more expensive option of transporting it by aircraft. And again, it boils down to having a logic that can do the arbitrage between the two. You mentioned the word “automation.” Let’s talk about things from a production perspective now. Uniqlo recently announced they’re using automation in one of their warehouses. What are the real production challenges that you see in this industry, and how can you see automation changing that?
Joannes Vermorel: It’s interesting because textiles, I would say fashion and textile to some extent, have been an industry that was at the very beginning of the Industrial Revolution. It was one of the first industries that was really impacted by mechanical production. So it’s interesting they were there at the very start. But what happens is that, although manufacturing fabric can be heavily automatized, fashion itself, especially when it comes to cutting, sewing, and everything, is harder to have a high degree of automation. So this industry, to some extent, remains relatively manual, especially if you compare it to car manufacturing, where the plants are literally gigantic machines with very little human intervention.
Nonetheless, automation has been progressing, and when it comes to logistics and warehouse automation, it has been progressing rather dramatically. For example, in France a couple of years ago, La Redoute, which is a century-old fashion company, announced that their new warehouse had twenty times the productivity of the old one when it came to shipments. So there is a steady productivity improvement, and what’s interesting is that we are now reaching a point where you end up having proportionally a lot of people that are actually clerks who just have to come up with all those numerical decisions to drive all the production units and all the warehouses and to keep the workflow flowing.
Because the physical process is getting better and better and more automated every day, it means that the ratio of people that are those clerks who have to decide all those quantities is becoming dominant, not only in terms of headcount but also in terms of delays. And the next changes will be to basically bring the automation also to the decision-making process. I’m not talking about the very high-level strategic decision-making process, such as having a vision for your brand, that will remain completely dedicated to humans. But I’m talking about making micro-decisions on tens of thousands of SKUs on a daily basis.
Kieran Chandler: So we’ve spoken a little bit about production, warehousing, and transportation. I guess the final piece of the jigsaw is the end user and the actual sales process. A big part of the fashion industry is sales promotions and price elasticity. How does that affect things?
Joannes Vermorel: Price elasticity is the main mechanism by which fashion brands liquidate their inventory at the end of the collection by doing sales, thus increasing demand and liquidating inventory. Now, it’s interesting because you can also leverage the same mechanism but just the other way around, which means if you are heading for a stockout, there is no point in rushing into the stockout. So, you can raise your price a bit and still reach your stockout down the road, but you will have made a better margin through those products. Price elasticity is an interesting mechanism. It’s also a very challenging mechanism because the problem is that if you do a massive sale at the end of every collection, your customers get used to that, so they will postpone their purchase decision to the final moment because they expect the sale to happen. That’s very tricky, and again, your intuition can be misleading. That’s also one other area where a quantitative approach would be helpful.
Kieran Chandler: So, would you say that quantitative supply chain optimization makes sense? And if so, why?
Joannes Vermorel: Yes, indeed. One should consider quantitative supply chain optimization because it addresses a crucial question for every single product you sell across all channels. If you run multiple stores, each becomes a channel of its own. For every product, one needs to decide whether to adjust the price upwards or downwards. This decision needs to be made daily, although the answer might not always be to change the prices. Often, it might be best to keep them as they are. However, this process is labor-intensive if done manually, making it an excellent candidate for automated quantitative methods.
Kieran Chandler: So, are you suggesting that instead of sticking to collections, we should move towards a more gradual transition between the seasons?
Joannes Vermorel: Precisely. The waterfall process, or releasing everything at once, adds stress to your supply chain at multiple levels. Suppliers receive massive orders at certain times of the year, and they might struggle to keep up, especially if their other clients follow the same pattern. Warehouses then have to process large intakes at specific times of the year, which is also challenging. This leads to the shipment of large quantities at certain times, which can be difficult for stores to manage. When sales happen, we’ve all seen how chaotic stores can be. It’s a significant strain on the staff and the brand. Furthermore, you end up selling a lot, but at a steep discount, so the net result isn’t that great.
Contrast that with a scenario where you’re continually updating your collection to follow the latest trends. You have new products every week, items being discontinued weekly, and smaller, more manageable shipments. It’s easier to manage a steady flow than sporadic, massive spikes.
Kieran Chandler: One of the big challenges with forecasting for the fashion industry is accounting for those moments when a celebrity suddenly wears a particular brand, and sales spike. Can you ever really predict these freak spikes in demand? How can you respond to that?
Joannes Vermorel: Freak spikes are statistical outliers and, by definition, are hard to forecast. The good news is that it’s not just you; your competitors can’t predict these spikes either unless they’re the celebrity causing it. The only people who might know in advance would have privileged information. So, if a celebrity promotes a bag, for example, and the 5,000 units you had in stock sell out in two days when they were supposed to last the entire season, there’s no need for regret. It’s a good problem, at least the bags were sold. However, it wouldn’t have made sense to order 50,000 bags in anticipation of such an event. The risk would have been too great. So, there’s no regret in having a shortage in such a situation.
Kieran Chandler: You transition toward probabilistic forecasting, then you can add your bet. You know, you can start taking into account the likelihood that these sorts of things can happen. And by the way, sometimes in fashion, you can have some things that can be predicted. For example, if you’re into sporting goods, you do not know, and let’s say you are selling t-shirts that can be printed with the logos of various teams. Well, every year one team is going to win the national championship. So, you know, it’s possible some brands are doing that. They will keep t-shirts that have the colors but have not printed the logo, and you will print the logo at the last minute so that you have sporty shirts ready to be able to print more for the team that will ultimately win. Knowing that every year one team ends up winning, you know, that part is highly predictable even if you do not know which one.
Joannes Vermorel: The idea is that you have to transition from a mindset where you know the future to a mindset where you know the shape of the uncertainty, and you can adjust your decisions taking into account the risk and the financial consequences of those risks.
Kieran Chandler: What about channels such as Instagram, where you see somebody wearing a particular item of clothing? Would you say that the fashion companies have already pre-thought that through, and they’ll already have all the stock stockpiled, ready to go with that?
Joannes Vermorel: I’m pretty sure they don’t. I mean, yes, in a science fiction scenario, the brand would approach an artist and know the probabilities that the artist is going to wear the stuff that is being offered and have, based on those odds, prepared the goods so that they are ready to ship to their clients when the demand will emerge, and all of that coordinated in real-time. I believe it’s complete science fiction. Clearly, there are many brands who are doing guerilla marketing, trying to approach artists and celebrities, and marketing themselves quite successfully sometimes. But my own experience is that in fashion, for most companies, even having a clear picture of how much inventory they have at any point in time is a challenge. Yes, there is a system where you know how many units there are in a store, but that doesn’t give you the whole picture. What about the returns that you expect? What about the deliveries that you can expect from your suppliers? What about many things that are in transit within your own network? So, for example, just having a clear, real-time perspective on your stock levels is, I would say, already kind of ahead of most fashion brands nowadays.
Kieran Chandler: So the idea of having a system that is completely realistic, that can take into account the interaction with every single celebrity in every single country, is probably right now complete science fiction. They need to figure out tons of things. There are tons of things that are much simpler that can improve the situation. You don’t need to delve into social media unless you’re already, I would say, rock stars in quantitative analytics.
Joannes Vermorel: Okay, great.
Kieran Chandler: We’re going to have to wrap things up there, but hopefully, someone will send us a few shirts or something now that we’ve done an episode on fashion. So that’s everything for today. Thanks very much for tuning in, and we’ll see you again next time. Thanks for watching.