Improving Supply Chain at Zalando

Hey, welcome to another episode of Fashion Made. Today I have the pleasure to be joined by two individuals. I have Jan Willemking. Hello, hi. Hi, Jan. And Christian Beyer.

Hi there.

Hi. Jan and Christian both work for Z labels or Zalando. Which one do you prefer?

Both are fine. Yeah, we are basically from Z Labels, which is the private label company of Zalando. So we are— you can use both.

Fantastic. Well, welcome to the show.

Thank you.

It's very nice to have you here, and today we're talking really about the supply chain, AI, digitization, all of the amazing things that are happening in the fashion industry. It's quite an interesting topic, especially when it comes to supply chain though, because of course, it's not as digitized as some might have you believe.

That's absolutely right.

Yeah.

Yeah, I think for us, the starting point was to recognize that the front end of fashion, of fashion e-commerce, is highly digitized. I mean, you know it. I mean, we have apps, we have interaction every day, we have tons of data points, and we have, luckily, a lot of visibility on that end. So what consumers actually want, At the same time, Christian and I, we are leading the company that is creating fashion brands on an e-commerce platform called Zalando, and we see that the backend of all of that, so the fashion supply chain itself, the creation, the production, the quality control, the logistics, so making stuff happen is basically a world which is very, very analog. Yeah. And there's a big contrast, I think, We see that there's a lot happening, but status quo is it's not digital and it's not AI-driven and not data-driven at all.

Okay, and so what's your goal when it comes to changing this?

Well, our goal is that we actually want to use data to, well, inform decisions along our supply chain, right? And I think building on what Jan said, it's first of all about informing the design and the product creation in that supply chain. So we have the feeling that in our company, but also across the industry, much of, you know, what the designers, the product developers, the buyers actually decide is based on experience. It's based, you can also say to some extent, on gut feeling. We want to change that by really understanding the customers, what they want, and being able to describe a product along very many detailed product attributes. So for example, you know, what's the color, what's the material, what's the shape, what are the different trims we actually should take to create that product that's most relevant to our customers. I think that's where it starts, but then really it follows on throughout the value chain. For example, talking about logistics, when do you actually do an investment into putting something on the plane rather than shipping it in to save lead times? Is that a good investment, yes or no? And you can go further, for example, to sales. When is a good investment in terms of performance marketing to push visibility of these products on site? And I think this is how we think about enabling our entire value chain via better data insights that we want to derive not in a manual way, but by applying artificial intelligence or machine learning, if you want to.

Okay, yeah, maybe one analogy is when you think about the word stock, everybody thinks about in this industry about clothing or shoes sitting in a warehouse and that's my stock. I think we think about stock also in a different way when you think about stock trading. Then you're suddenly in the world of financial decision-making, and you're suddenly in the world of, I have a pot of money and I want to allocate this money into the stocks that are actually having a good trajectory and where I can actually have a high ROI. And ultimately, when you think about what happens in fashion, we have a pot of money, and luckily we in Zalando, we have also pot of data, and basically we are investing that money into stock. Stocks. I mean, in, in Zalabiz, we are doing 15,000 SKUs per season. So 15,000 little bets. Now, in the banking world, in the financial world, we have liquid markets, we have immediate reactions, and we have visibility of what's trading and why it's trading. That is something that we have in the fashion industry only with a time lag, right? Because when we decide to invest into a stock, into a product, We need to make it, we need to design it, then we need to get it from somewhere into our warehouse, then we need to list it, and then we need to basically start selling it. So we have a time lag of, in our model, of a few weeks to a couple of months. In the traditional wholesale model, even longer. I mean, we come from development cycles of 18 months to 12 months now, and that's already fast. But I mean, if you take the analogy of stock trading, And then you ask yourself the question, why the hell in a multi-trillion-dollar industry are we still accepting non-data-driven decision-making and very long lead times? Why should that be the case? And I think this is why we are now really thinking about the fashion value chain in a different way than coming from a wholesale model and now moving a little bit faster. We really think, how can we radically make a difference?

Yeah, radical change. What do you think Zalando is doing really, really well right now? In this area?

Yeah, I think what we have started to do quite well is to attract the right talent. I think this is actually where it starts because all the things we talk about, they are quite complex in nature and you need data scientists, you need data engineers, analysts to actually move us forward on that. We have been able to recruit actually more than 1,000 tech people working on these kind of topics. Not only on data science, but also on data science, which is a core investment area for the entire company. And apparently we're able to tell a story. I think we started telling that story that excites people out there and makes them want to join our company. And I think this, for example, is one area where I believe the company has been doing quite well in recent months.

Yeah, and I think that goes especially for the retail/retail technology part of Zalando, so where it's about buying stock from brands and selling them in a retail model. I think for us the complexity is even slightly higher because we have a longer value chain. But ultimately the logic of attracting talent that it's not common to the industry also holds true for our team. If you just think about our company within the larger company of Zalando.

Yeah, I mean, culture change, I guess, is a really important factor of this, and that's something I want to touch on as well. When you bring in all of these new data scientists, have there been people that have been working in ZLabels or Zalando for a long time and who have had to go through a period of change that have affected their day-to-day lives, and has that been difficult or have they been really good at it?

So I think when we take a look at our part of the business, I think data scientists are used to data and to data-driven decision-making. I think that's not a lot of change. I think the biggest change is we are— when you think about it, we are with Ted Labels, with our brands, you could say we are relatively analog fashion company that is bolted on a digital, highly scalable, technology-driven front end.

Okay.

So, and that actually means that we employ buyers, product developers, designers that come from other fashion companies. And we were— we didn't exist 10 years ago even. Zalando was starting, but Ted Labels was founded like roundabout 8 years ago. So we are super young company.

Yeah.

Which also means that we had to hire very, very fast and a lot of industry experts. Expertise, which means that we have people from, I mean, multiple different fashion retailers and fashion companies that joined us. And they have had their career, and they have had their experience, and they have had their learnings, and they brought all of that. At the same time, it is a traditional fashion mindset of a buyer who wants to sign off a collection based on what he or she believes is the right thing. You have designers who do trend research their way and rather look into maybe Instagram or do kind of store checks, but not looking really into granular data. You have designers who don't want to think about standardization because it takes away creativity potentially, and even though standardization might be a big, big driver of speed and flexibility. So I think the biggest challenge is that especially in the traditional professional groups in our industry, the technology and data capabilities are relatively limited, and that's the biggest change to be done.

Okay.

Yeah, and maybe just building to that, because on the one hand, I think it's truly exciting that you have all these different profiles. It's a creative industry, it's a great industry to work in, but that actually also creates that challenge Jan was talking about when you want to bring data science on board. So maybe just sharing what we have done in our company, which is a fairly small company of roughly 500 people. So I think the journey started actually with Jan and myself really thinking about the future of our industry, forming our picture on why data science is so important and also what we need to do on the supply chain around agility and speed. That's where we started and then actually first step for us was to take our top management on board. So we actually did a 2.5-day offsite where we then, you know, discussed these ideas and and also had these people speak up their mind and share their opinions on where the fashion industry goes with us. I think that was the first step. And then thereafter, we spent like 3 months, just like one day per week, a full day per week to discuss that and build roadmaps and do all of that. And when we were ready for kind of like then going the next step, we actually invited all of our leadership, which is around 100 people, to have these discussions. And again, followed up with many, many days where we intensely discussed and worked on different topics of the future of fashion and our company to prepare these people also to tell their teams and onboard their teams of what's needed. So overall, I'd say there was a process of 9 months roughly until mid of this summer in 2018. But we also, I mean, it was intense, it was great, super inspiring, energizing, but for sure it's the start of that journey. It's clear to us that that is a multi-year change management process if you want to really get to data-driven way of working and then the agility in the supply chain.

But I guess this is what you have to do, right? If you're trying to go through this bottom-up transformation, you've got to have all of these different stakeholders who are in all of these different departments, whether they're analog or or data scientists. And did you guys have champions within each department, or do you have one guy overriding everything? Was it you, Jan? I mean, how did you make it work?

So I think it was crucial to bring our leadership team on board and to go through the different kind of areas of our business. So like, let's talk with designers about data. Let us show— I mean, we invited—

How did that go?

Actually, in the beginning, obviously, it's like, yeah, not so good, right? But I think one observation was when we talk about data, the immediate association of many people is computer, then it's about mathematics and numbers. And that, especially in a very creative environment, creates fear. Secondly, I think when you think then about like what will data do to us and there's now this kind of crazy artificial intelligence that will take away my job, oh my god, so there's another element of fear. So there's a ton of fear. It's like I'm not valuable anymore, I was never really good in math, but I'm really good in like creating garments, so do I have a future here? So we had all these kinds of discussions and how we then treated it, we basically made it very simple to understand and feel what it actually means. So we brought in a company that is helping us right now also with the implementation of 3D construction of garments. So it's like not touch and feel anymore, but see it like in a 3D avatar, right? I mean, there's a couple of companies doing that now, and we deliberately decided to give them half a day of such a day and get people the ability to try out the software, to hear how from others how it works, to kind of Test it. And this was extremely important just to see, okay, that's not dangerous, that's actually fun. And the second one was about this kind of mathematical— I don't like numbers, ooh, go away. When you take a look at what data actually is, data is nothing else, at least for us, than oftentimes just the voice of the consumer expressed in a number. It's the voice of consumers. And who are we actually to judge what the voice of a consumer is by just looking into a store if we don't even listen? So when we truly want to listen to consumers, it is our obligation to look into data. And I think that was a clear thing. The voice of consumer is actually what data is. That opened also the, the minds, and I think also a little bit of the hearts of people who were data skeptics.

Yeah, yeah, yeah, I know it makes sense. You know, there's a lot of voices when you're trying to listen to a consumer, I imagine. When you're talking about data collection, there's a lot of data out there, right? And there's a lot that you can collect. So how do you guys go about deciding which data to listen to, or as you would put it, which voice to listen to?

Yes. Yeah, we thought about that, obviously, quite in detail. I think for us there's, there's a couple of different data sources, and specifically I'd like to mention two. First of all, We need to listen to our consumers. We want to listen to our consumers, that voice of the consumer Jan just mentioned. So, um, with our company, the, the, the private brand company of Zalando, we are in the fortunate position that we actually have access to all the data that's happening or that's, uh, collected on Zalando. And this is not only the data, sales data of our brands, but also of all other brands. Um, but it's not even only sales data, it's also all the on-site data. So the customer journey online of our customers. What do they buy? What do they click? What do they put into baskets? So what do they buy together? And I think that is one, one piece which obviously is totally vital. We have 23, 24 million customers on Zalando, and we want to listen to these customers. And once we do that right, detect the patterns, this will help us to get the right planning, dynamic planning, detailed planning that informs our value chain. I think that's the first bucket. But that only helps if we actually connect it and understand what's happening in our supply chain. So we need to know for our key partners, our factories, our suppliers, what kind of capacity can they offer at the moment? Do they have pre-bought materials and have these materials on stock? So are they ready to go with their production? What kind of lead times can they offer to us? What kind of quality levels are they able to achieve? And we believe only if we couple the, the insights from the consumers with the insights from our supply base, connect them, only then we can make the right decisions to develop in the end the most relevant product to the customer and order that in the right quantities and bring it in at the right timing. Because if we do that, then I think that builds and helps us achieve our vision. And that vision we basically formulated as creating fashion without waste, which might sound trivial. We believe it's not trivial at all, and we believe that's actually one of the big problems in our industry. But we want to work towards that goal of creating no waste, or at least a lot less waste in what we do.

I— yeah, you're absolutely right. We've been running this event for a while now, and I feel like some of the technologies really enable fast fashion, and fast fashion isn't necessarily sustainable fashion, let's be honest. So combining quicker lead times with better products, but also trying to be without waste, I think, is a very good goal. So yeah, it's certainly not a trivial topic, I don't think. With all this improved forecasting, what effect does that have on the early product creation process?

Well, I mean, one thing is that when you look into traditional merchandising in a retailer, what do they do? So first of all, they look into what did we sell, last season? What was our stock last season? Did we have an imbalance, yes or no? Did we have too much from X or no? And what is our trajectory for certain product groups, for certain price levels, for certain kind of shapes and forms?

Every—

everything done on a very— I would say ultimately on a high level. We're talking maybe about party dresses. Yeah, okay, maybe party dresses or loafers. Is the smallest unit that I actually look at, and I put a certain amount of money behind loafers or party dresses. And maybe you can then say, "Ah, I need to have them maybe in a different price point." Okay, that's my smallest unit that I look at. Now, if you think about briefing a designer and saying, "Hey, can you please do 15 party dresses?" Yeah, is it now the black one? Is it now the glittery one? Is it now the fancy one? Is it now a long one? Is it now a short one? Does it have a print on it? Does it have like this red zipper on it? Whatever it is. That is all up currently in most of the companies we are talking to, up to the designer and the buyer to judge it. And all the kind of decision-making is then driven by observation, which ultimately is this kind of meta-information that then a very good designer and a very good buyer absorb quantitatively, qualitatively, and then make up their minds and say, "Ah, I don't know, I think it's now—" black kind of shiny thing in short with short sleeves. That's exactly the party dress that we need to do. And I think that's in many cases good, but if we take a look into the success rate of how many of the party dresses or how many of our products in general that we do this way are actually really successful versus not, I mean, our hit rate with the consumer is also not 100%, right? And it's also not, it's also not 10%, but it's Maybe that's only 50-60%. So we are basically going into a casino and we're doing a well-informed decision, a professional decision, but we are losing more or less half of the time. Well, that is not really good, right? So can we increase the hit rate? Can we increase the precision? And our answer to that is, why don't we try to understand in more granularity and coupled with data, what the attributes are that consumers are looking for. What are the different kind of shapes and forms that they are looking for? So the little cell of a party dress suddenly becomes dissected into many, many, many subsets where we then say, "Okay, it needs to be attribute 1, 2, 3, and 4 in combination with A, B, C, D, and E." And that becomes ultimately a pretty good and granular product recipe which takes a lot of guesswork.

Out of the equation. Yeah. Is there a danger of becoming too clinical with this information and losing a little bit of gut feeling?

I think yes, because, I mean, data is the voice of a consumer, but it's also quite— it's always backward-looking. You can argue that there are segments of consumers in the— where you see in the data that they are actually leading the pack, so you also have the fashion pyramid in your data, and we can decide to, to, to, to, to overweight certain voices because they are leading the trend. I think it is possible to look a little bit into the mainstream future by looking into leading consumer segments, but at the same time, I think we are far from a reality where machines can make the best decisions on the exact look and feel of a dress. I think we are making exponential steps in that direction, but I don't think that in the next 3 to 5 years, it will be perfectly solved. So I think gut feeling and experience, especially in this early stage, are extremely vital. So it's not about substituting gut feeling, it's enhancing gut feeling with even more granular ideas of where then to focus your creativity on. Yeah, yeah.

And maybe just building on that, we not only want to better inform our teams, we also want to better inform our suppliers and factories. Because what we are actually doing in our company, we follow a mixed model. By that I mean we do own design where we create products from scratch, but we also do creative buying, basically meaning that we choose from existing collections that suppliers provide to us or show to us. So I think this entire data idea goes even further and stretches out then also to our suppliers. What we want to do with key partners to just integrate a lot better, share a lot more of that consumer-facing insights that we have, to also put them into the position to not only make gut-based decisions but really also look into the data to then develop their collections that really fit and are super relevant to the consumer that we see and want to serve on Zalando as a, as a platform.

Does you helping them allow them to help you back as well? Is it a two-way street?

I think it's very much a two-way street of integrating the data flow. What we can obviously do and want to do is share that information about our consumer, because normally, you know, factories, suppliers sitting in Asia, they don't really have access to the European consumer. So we want to share that data to put them into a better position to create products. But then vice versa, what we also expect and are working on is how do we actually get these insights I mentioned before around the supply chain. What are the capacities that they currently have? Do we need to pre-block capacities? The materials that they have on stock, all these different things. So it definitely needs to be a tightly integrated process exchanging data because we strongly believe in that data aspect, but it needs to be a mutual thing to get the best out of it.

And I mean, just to share a little bit of experience and observations here, I think there is a couple of suppliers out there, especially when you think about second-generation people in their 20s, beginning of their 30s, that are now taking over from the slightly older factory owner. I think these are sweet spots where you can suddenly start talking about these things, where you have people that understand that potentially have been educated in the US or in Europe and potentially were part of a consulting company or an investment bank, so they are not afraid of data, they're actually longing for data. And they are able to retain and recruit people that are similar to them. So that actually means suddenly you have suppliers, and it's not a lot, but you have few suppliers that are actually able to work with data and to interpret data and to have a dialogue. But there's also, I mean, there's also a large, vast sum and group of people out there who literally say, come on, just give me an order and make it at least 1,000 pieces and give me that cost price, and then I start working. I just want to be on the receiving end. And that's completely fine, but I think we are looking for people who are interested in this kind of joint learning and truly implement it.

Yeah. And then I think also worth to mention is we are also prepared and to go beyond that, not just only share data, but also really talk about how do we collaborate and also how do we share the gains from applying data. And the risks. The risks as well. That's good that you mention it. But then really possibly looking into a very different model of really sharing these, these benefits and these risks. To really share them and not leave them on one side or the other. And that is also something that we're now piloting with a couple of selected suppliers that are really up for the challenge.

I mean, surely the biggest risk— you talk about risks— surely the biggest one is you don't do anything and you get left behind.

Ja, at the same time, that is a very long-term thinking. If you own a factory and you have capacities to fill and people to feed, then I think the biggest risk is to not make enough money to keep your factory alive. So I think the time horizons of, of looking at it are very different if you're basically sitting there, you need to fill capacity. So, and that is exactly where it becomes interesting. So how can we make sure that we, in this kind of longer-term vision, create a setup that is comfortable enough for suppliers that have to have certain security and also efficiency in their production process to have a good business. Because if you go to a company that is typically geared to thousands of items per order, and we now tell them, hey, can you please do 200, and then if it actually works, we tell you in one week after it's online, we actually want to go for 1,500, 2,000, or even 10,000, are you ready for it? Most of them say, "Pfft, 100 pieces?" Shock. It's like the sticker shock. It's like, "I don't like 100 pieces. It's inefficient in the production. I don't like it." So we have worked with a couple of very open-minded suppliers to crack the formula how to actually make, for example, 200 pieces a reality. And that is something that is extremely important important because it goes into how a factory is managed, how a factory is set up, how the incentive of workers in the line are actually created so that even if they only, only do 200 pieces of a certain batch, they don't earn less money. So it goes really into the structure and the management of factories, and that is why we cannot do that with many, many, many partners, but we have a selected group of partners who are understanding this and who are willing to try it out. And to create, you could call it a mini factory within a factory, that is geared towards online and data-driven ways of working.

And I think what's interesting, these partners Jan mentioned are actually very different in nature. We do that with huge multi-billion suppliers. For them, we are actually not really relevant in terms of volumes, but it's exactly what you said and your question that you ask is they want to think together with us about the future and pilot the future and understand how this, how this all works out. So this is their interest in joining us. But we also have really relatively tiny factories with 300, 400 people working in that factory, sitting right in the middle of China. And there's this management, young guys, not even 30 years of age. They have done their MBA somewhere in the States. They come back and they really now also want to explore What's the future of that industry and want to make a difference in this industry? So it's, it's a versatile group of partners, also not too many, and those are the right ones for us to pilot all of what we've been talking about. Fantastic.

Well guys, I just want to thank you so much for coming in, sharing a bit of insight into all of this. You know, it's been really interesting. I think we've covered quite a lot in the last half an hour. So thank you very much for joining us. Cheers.

Thank you.