Using Automation in Design to Unlock Time | Clothing Tech | PI Apparel

Alright, so we're Clothing Tech, and I'll confess first that I know very little about garments. I've learned a lot actually in the last 5 years, so I know something, but I would say I'm still learning. But I have a few ideas and things to present that I think you might find interesting about how we can actually go faster. So clothing tech is a joint venture with Hexagon, which is a large, about 25,000-employee company that supplies technology to automotive, aerospace, electronics, and civil engineering, construction, all kinds of industries. We started this about 5 years ago now to try to bring some technology that we had and also create some new technology that would be useful to the fashion industry. And so we've taken a very different approach. Our FashionR software is a 3D CAD system, and I use the word specifically CAD because what we have right now in the industry is actually not CAD. It's computer-aided visualization because what we're doing is you're visualizing garments, but we're not very few people actually designing garments in 3D. And even when you're using a 3D software, you're actually changing the pattern in 2D and visualizing in 3D. So CAD is computer-aided design, which means that you should be designing in 3D, which is what most industries do. So we're not actually doing the same thing as the automotive industry, the aerospace industry. So we are, came from that perspective of different industries and saying, okay, how do they work? I've been in hundreds and hundreds of manufacturing facilities around the world, from cars to iPhones to all kinds of airplanes to you name it. I've only been to one clothing manufacturing facility, but I'm working on that. I'm headed to India in a little bit here to see a few more and work with more partners there. Some observations coming from an outsider. And one of the things is that very late to 3D. Most industries started 3D journey in the 1980s, maybe in some even in the '70s. So aerospace was one of the first to convert to 3D. And so I've been around since those days. Don't ask me how old I am, but I do remember going into design studios where they, used drafting, paper drafting for cars and things like that. Now you don't see any of that ever now. It's all done in 3D CAD. So, and for a large industry, it's a little surprising. Well, it was surprising to me anyways when I started investigating this industry. The product definition is not digital. Not sure if you know what that means, but basically a computer can't understand the product you're trying to make. A person, an expert has to understand it, has to be there to understand it. So you need to communicate with another expert who can read your tech pack and say, "Oh, I know what to do here. I know what to, oh, I have a question. Can you clarify this?" But most of the communication about product definition, Product design is expert to expert and also fairly manual. So there's very little digital definition of the product. Yes, a tech pack is a digital document, but it's not digital information. It takes a person to read it and say, okay, I understand what's being asked for here. And so this is also surprising to us from an outsider. It's like, why not define the garments, the product? Digitally so that you could produce with a robot if you wanted to, for example. A lot of inefficiencies because it's expert-extra-driven. So you have an expert who's talking to a designer, talking to a technical designer. A technical designer's talking to a pattern maker, to the factory about refining the specification. The pattern maker and the— is talking to the sample maker. The sample maker sends the sample back, whether it's virtual or physical. The sample goes back to the designer. All of these is sort of serial and expert-to-expert communication. And it makes for inefficiencies because you're having to communicate and describe things in a way that is not always completely clear because it's sort of like with arrows, do this here, that kind of communication, which is visual, but it's not very understandable if you wanted to automate, for example, which is the last point that we saw. Coming into this industry from an outsider's perspective, there's almost no automation. There is automation in the supply chain, but it's early on, like going from fibers to yarn, from yarn to fabric. This is highly automated, knitting machines. All of this is fairly, you don't see a lot of people doing that work. But from that point on, going from that yarn to, sorry, fabrics to now products, Pretty much all manual. Some little more and more automation in the factories, but not a great deal. So automation, when you're doing the same thing over and over again, it seems like if I go into a store and I've been into, you know, and I see the garments, they don't look that different than the garments from last year, the year before. There's some similarities, right? The garments are similar. So we're not really, changing things drastically from one season to the next. So it could be automated. There's nothing really preventing that from being automated, but it's about how you set it up and how you think about the problem. So typically 3D visualization sits in the place where we are between sort of pattern— usually there's a pattern before you start visualizing it. So somebody's gone through and done pattern making. And you have tools which have been around for 20 years, generally, 20, 30 years, some of them. So fairly mature tools that you use to take the technical design and the pattern and visualize it in 3D. And it takes an expert to do this. It requires a special skill set to do this work. And it creates a sample. And so this sample and reiteration is kind of how most brands are working today, or at least trying to work. A lot of them still use physical samples depending on the, depending on the category of the product. But for like certainly for some categories, there's a lot of 3D visualization. From clothing tech, we wanted to do something different. We wanted to go from design to manufacturing. And so this is our focus, a tool to design or completely define the product and have it be used throughout all the way through to manufacturing. So this part of the process makes us quite different in approach and also in how we use the software. So let's talk a little bit about how this could work. So there's a couple of things that we've talked to. You know, there's a lot of, there was a survey done I think by Calypso last year And they ask a lot of brands, why did you buy 3D? Why did you invest in DPC? And the answer was generally to go faster. 96%, I think, said that they wanted to go faster. They invested in DPC to go faster. And so the second follow-up question was, how many of you made significant changes to your calendar? And the answer was 4%. Very few had actually made significant changes to calendar. Most people could say, oh yeah, I shaved off a week or a month or something, but a month out of 13 months, it's not a big, it's not gonna change the way you do business basically. So how do we go actually go faster as an industry? How do we, forget about technology for a minute, how do we actually go faster where we wanna go from idea to start a manufacturing and be able to do this in not just one week faster or one month faster, but 60% faster, 70% faster. So 12-month calendar down to 2-month calendar. How do you do that? What would we have to change and how could that work? So there's 3 things that we noticed that would help with that and make a significant difference. One is digital product definition. Where you're defining the product digitally. What that means is you're defining, and I'll explain a little bit more in detail in a minute, but you're basically describing in a way that a computer can understand all of the relationships and the details of the product. What is this product to manufacture it, to visualize it in 3D, to cut it and pattern and all of those things? What are all the data that's necessary to drive that process? And is it digital? Right now, in almost all of the processes I've looked at, it's not digital, or at least a substantial part of it is not digital. And so this would be a change, but the advantage of digital definition is that you can drive automation because you have now data that is in a format that a computer can understand, which means that repetitive processes can be automated. Working in parallel and in tandem is another thing that, you know, this phone, I've worked with Apple and I've been to some of their factories and we've worked on a couple of automation projects there. This phone, it's a one, one-year engineering cycle. So every year they come up with a new phone. They start to finish, they design it, they get it all ready and they're ready to start production after one year. And they're very rigid on their calendar. This is probably 5,000 to 10,000 times more complicated. Of a product than this jacket or these slacks or shirt that I'm wearing. It's substantially— so why does it take the same amount of time to do both? It shouldn't take the same amount of time. It should take much, much less. And so this is sort of some of the ways of thinking about it. One of the reasons why they're able to go so fast is that they're able to work in tandem. So they divide up the product and they work in tandem. They don't all go like, okay, I did my part, now you do your part. And then they hand it off to somebody else. You do your part, and then you do your part, and so forth. When you do things in serial, in steps, it's gonna always be slower. And so this is one of the other things to say, well, could we, how could we enable that so that you could actually work in tandem, in parallel to each other, so the technical designer and the patternmaker are both working at the same time on garments and not necessarily waiting on each other. Or the sample maker, et cetera. All of these things should be able to be done much faster. And then of course automation, which we talked about. There's a lot of automation potential in the industry. A lot of things that are the same from one style to the next, from one season to the next, and so forth. So this is the things, this is what we focused on when we developed Fashioner, which is an end-to-end, starting with concept all the way through start of production, a tool that allows you to define and automate many of the steps in that process. So, so see, these are some of the areas that we've focused on in terms of automation. So we have projects, but not in all of these. Mainly what we focus on at this point is the design process. What do we do? How do we get the construction, the reusable finishes? So typically, when you take a new style, what do you change? What changes from one season to the next for a style that— I mean, maybe you're making a brand new creation that's never been made before, but that seems somewhat rare from my perspective. So you're taking a style, and you're usually changing fabrics and colors. You're sometimes changing the shapes. So we want boxy, or we want form-fitting, or we want the shape, maybe we want longer, shorter, midriff, I don't know, all of the different shapes that you can think of. So shape changes often between seasons, and also finishes or features change. So I don't want collars anymore. I want just collar stands now in my season. I don't want, the placket should be narrower, you know, it's just the finish should be constructed differently and things like that. So finishes, the features of the garment change. There are also potential to automate marketing imagery. Once you have a complete database, when you have a database of a digital definition of your product, you can convert it into a 3D, and I'll show you how we do that. We automatically convert. All of the definition of the product and convert it into a 3D visual. And there's no manual sewing and things like that. It's very straightforward. It's also possible we could automate the manufacturing of the garment because same information is available to drive a robot to tell it how to sew the things together. Nobody is pushing to do that right now, so we haven't focused on that area. Too much, but it is possible when you have a digital definition of the product, you could potentially automate. You cannot automate from a tech pack. A robot's not gonna understand how to do a tech pack, what you want done with a tech pack. You have to go from something more specific, more concrete where they can actually understand it. So let's talk about the definition. So we've talked a little bit about the idea of digital information. And so we're trying to create a 3D, but what we really want to do is define the product well enough that there's no miscommunication possible. So when we talk about, you know, when you send a tech pack to a factory, do they sometimes have questions? They come back and say, "I'm not sure about this detail. Could you explain this a little bit more?" Maybe they get it wrong and you have to go back and correct them afterwards because they didn't understand something about it. And for us, that process means that there's something wrong because you shouldn't have that. If you defined your product, everything should be, all the details should be available. There shouldn't be questions because you've answered all the questions in the specification of the product definition. So if we talk about this, look, a different way of viewing this is that it's like one set of data that has many different views on it. So you have views of the cloth simulation. How does the cloth simulate drape? All of that is part of the process. The stitches, how do we sew it together? What seams, what goes, gets sewn to what? What are the details of the pattern? What shapes are we gonna actually cut? The folds, because we're gonna fold and press and create seams. There's finishes, some areas where you could think of in other industries we call them subcomponents where you assemble them separately. Like for example, I've been to a bra factory where the bra strap is made in a different assembly line than the rest of the bra because it's easier to automate the bra strap. So they create a separate assembly line for that and have steps for the— and they just have a box full of bra straps that they sew onto the garment in the main assembly line. So finishes can be considered like that. This is a part of the garment that is, it's like an inset pocket on a pair of jeans or a pair of slacks. You have the pocket bag, it gets folded, it gets placed, it gets stitched, it gets, you know, there's a certain set of steps that happen there. Or patch pocket, how it's sewn on with a pocket machine that is automated. And so you can set up that information, these kind of drive that directly from the data. Production steps, operations, order of operations, and costing. The bill of materials, the 3D visualization, the manufacturing, all of the equipment setup, and the tech pack, of course, is just a human-readable view of that. So what are some of the advantages of this? Well, one is that you have no— right now, if you take any one of your garments, any one of your products, and you say, okay, what's the different views of it that I have now? You have a DXF file probably somewhere. Of that, you probably have a tech pack as well, and you're probably gonna have sketches, and you're probably gonna have a 3D rendering or some kind of 3D visualization, maybe if you're doing 3D. So you're gonna have these views of the product. My guess is that we wouldn't have to look too hard to find that there's some discrepancies between the documents. They're not always matching up. Sometimes the tech pack gets changed. We approve this, but with these changes, but the pattern that you have never gets updated. Maybe the factory does some updating or vice versa. Maybe the pattern changes, but the tech pack or the points of measure don't get actually updated in the tech pack and so forth. All of these things happen because they're different sets of data. They're actually separate documents and there's no connection between them. There's no connection. If you had a style line and a garment, Yeah, you're cutting that pattern differently and you're adding an operation to assemble it, but that may not be reflected in your tech pack. It just may be added somewhere else. And so this is guaranteed that you're gonna always have the same view because it's only one definition of the product. All of the views are just different ways of visualizing or gathering data out of that definition. So you have the single source of truth. When we create a 3D, it has all of the details of the 3D. So what is the details of a 3D? Well, if I have a rendering of my garment, does it have seam allowances built into it? Generally, no. Generally, people leave off seam allowances, but you can put them in. It's just more work. And so, but when we convert it automatically into a 3D version, visual, It has seam allowances because the garment has seam allowances, and the seam allowance has to be folded in wherever they go. They go inside the collar, like this seam allowance here goes inside the, you know, gets folded inside the collar, and so forth. So you have a 3D representation that is consistent with the rest of the data, and you also have the, you're manufacturing ready because there's no ambiguity about the operations. In fact, the same costing data that is used to plan the steps is available to the manufacturers and vice versa from you to them if you set it up that way. So the manufacturing is ready to go once you're starting. Now, so that's one point about digital information. I think there's a lot of advantages of defining the products digitally and the problem is the tools we're using right now don't allow for that. They're just, they're not set up for that. But our tool, we tried it from the beginning, from the very beginning we said okay, What is the data that we need and how is it connected? What are the pieces and how are they, what's the relationships between them? And how can we make that so that it's easy to base as a basis for automation? So the line planning, typically you're gonna go through and so design sketching and so forth. Lots of room, come on in. So you're gonna go through and design sketching and so forth. Lots of room, come on in. Yeah, so you're gonna do design. Typically it's a drawing or some kind of, whether it's an Illustrator or some other tool. Then you're gonna sort of technically design it to figure out, okay, where should the seams go? Where should, gotta get into this thing. Where should the opening, where do we have to have a zipper or whatever? So somebody's gonna go through and take the design and make it into something that could be produced. Technical design and tech pack is gonna be created with points of measure. When I've visited factories, they call it a thin tech pack because it doesn't have all the definition of the garment, but it has enough that they can get started with the pattern making. Then there's pattern making, and then there's sample making, virtual or physical, and then, you know, review, and if there's changes, continue on. And I think you're all pretty familiar with this process. It's very serial, expert, expert-driven. So working in parallel, how does this work? Well, the designer uses assets that are pre-created. So the designer doesn't have to know about seam relationships. You don't have to put a designer in front of a very technical tool that's intended for a different audience and say, "OK, design your new style in this tool," where you have to do pattern making, sample making, and you have to also be designer, and you also have to know something about 3D assets and digital fabrics and digital trims and all of these complexities that they may not care about or know about. So the designer wants to look at the visual typically. And so we wanted to make— we have a different— our software, we've designed it around the idea of different roles. So the designer role is more or less drag and drop and property driven. What do I mean by that? Well, I wanna change the waist of a garment. I wanna change how the circumference of the waist. How do I typically do that in most systems? Well, I go back to the pattern and maybe there's like a, there might be 4 pieces that make up the waist. So I have to go through and adjust the shape of those pieces, add a little bit of volume to it and do that and then revisualize in 3D. And that's the typical process. Well, Again, that's not necessarily in the wheelhouse of the designer, and they're not so interested in that. They just want it to be 1 inch more. So we've tried to make this as simple as typing in what we call parametric CAD. So you change the shape of the garment by parameters, and the pattern making is done in behind for you automatically. So that means that the designer can drag the garment onto the avatar, from a DXF file with no sewing, no stitch relationships. They don't have to have any knowledge besides just how do I mouse and click and drag and drop. Once the garment's on there, I can say, well, I wanna change this about, or I wanna adjust this. And so there's, again, the tools we talked about. You wanna be able to change fabrics, prints, and colors. You wanna be able to change the shape of the garment, and you wanna be able to change the finishes. I want pleats now. I want a smocking right down the middle here gathered back in. These are the things that the design is intended on. So there's a role for the designer. These require tools. So if I create a smocking tool in our software, it's pretty technical because I have to know about relationships. I have to know how much I'm gathering. I would need to know how many lines of sewing I'm doing down that smocking. What's the start and end point of that? So it's relatively complex process, but the technical designer, the technical role could do that in advance and have the tool already created. We actually create many of the tools out of the box, so they're already available. But if you need a new tool that does something maybe slightly different in terms of how you change the shape of the garment, I wanna add some lace around the waist, neckline and I want it to start here at the shoulder, high point of shoulder, and go down and drape down and so forth. Okay, you may have to make a new tool for that. It is technical to create the— 'cause you have to know how to assemble a garment. You have to know how is this gonna be produced in the factory because we integrate costing into everything we do. And so you need to be able to explain all of the steps so that it can be done correctly. But once that's done, the designer's just dragging. So you're eventually gonna build up a library of tools. How many finishes do you have? How many different ways do you have of doing plackets? Well, you might have 10, but you won't have 100, most likely. I hope not 100. So same with cuffs, just talking about shirts for a second. Cuffs, how many cuffs do you have? Well, you're gonna have shapes of cuffs, but in terms of the assembly details of how it's assembled and so forth, how it's manufactured, Probably not that many, maybe half a dozen, maybe a dozen, I don't know, something in that order. So you can see how eventually you'd have all the tools you need. You wouldn't necessarily need to be great creating a lot of new tools. It just becomes, it basically componentizes the garment. So everything is drag and drop. I just drag this finish on and say, well, let's see what it looks like without. Hmm, I don't like that that much. Let me undo that. Let me drag a different one on, you know, and you can, iterate, in one day you can not only create the 3D design, but you can create 15 versions of that design with slight changes to it. Like I want to adjust this and adjust that and then compare them all side by side and say, mm, these two don't look so good. Let's cancel them, get them off the table. These ones we'll focus on and so forth. You're not waiting for a pattern maker. You're not waiting for somebody to— they've already done their work. The tools are there available. You have the block. That's what the pattern maker has provided. You have the block, you have the finishes, the features that you need, and it's just a matter of, oh, let's add a side slit, let's change the length, let's put a style line down the front. All of those things are done in 3D with simple Bezier curve, like in Adobe Illustrator. So it's very easy to do. So you have a design role, you have a technical role, which is about creating, how do I construct this part of the garment? Some things are very technical. and you have a pattern maker. Now the sample maker is a little bit different because other than physical samples, you have a virtual sample from the beginning. There's never a time when you don't have a virtual sample because you're defining a product. We could always represent that in 3D and we do. So every change you make, 2 seconds later you see it in 3D. Make another change, 2 seconds later you see it in 3D. So review, still a manual process. Is this a good idea, a bad idea? You still have to go through that decision-making process. But this is dramatically faster than the other. This is the— this is one way of combining the technologies that we've added to the table, a mixture of technologies that we've brought to the market that allows us to go faster, not just take off one sample out of the 4 samples that I'm doing, but actually go from idea to start of manufacturing in half the time or less. So let's talk about the timeline of things. You have all these process typically, I don't know, depend on the brand, the calendars can be a little bit different, but there's usually quite a few weeks. Some calendars are much shorter than this, some are even longer than this. But there's often a lot of planning and design before even design starts, there's demand planning and so forth. Then you have the plant design itself, then you have the virtual sample, physical sample, costing, all of these different things. You have to eventually issue a purchase order, purchase the raw materials, all of these things have to happen. These are all steps that have to happen. But we feel like we can shorten that substantially just by taking a lot of the serial nature out of the thing and also by defining it in a digital way, we can shorten this cycle. So this is, you know, depending on what your calendar looks like now, we can sit together and talk about, okay, how could we shorten that? What steps? Now you still are gonna make physical samples, I don't think anybody can claim— and people may try to claim it, but I don't believe it myself— that virtual samples are as accurate as physical samples. They're getting better. They're good. They're useful. But they're not as accurate as actually having a physical sample and actually trying it on and actually measuring it and wearing it. So you're going to probably still need physical samples at some point in the process. But you might need a lot fewer, and you might also need less time for the early decision-making process. Because if you know already that a garment is— the style you're interested in making is going to be too expensive for your target price, stop spending time on it. You would know that with this process in 1 or 2 days. You would know that. So this one's not worth continuing. Let's go on to the next one. So I'm just going to show you a little bit about how this works, how the idea of the drag and drop and all these things that I've talked about. So I'm going to play this video and you can see. So we're going to start with an idea. So this is the designer sketch, let's say. And they wanted to create a few different iterations of this top dress. And what block are we going to start with? Well, we actually started— in our software, the block is not super important. But it does have to cover the right part of the body. We can't pair it— take a pair of jeans and make a t-shirt out of it, or vice versa. They have to cover the right area of the body. So we actually created these with a t-shirt and then changed the pattern and adjusted the shape of things and the finishes, the features of the garment, and created these— those garments with this basic block. We exported it to production, and we actually have produced these garments. So this video will show a little bit about our software's perspective of it. So we're going to start again with a— we're going to import a garment. This is just a DXF file. It's not been preassembled or anything. It's just a DXF file. We bring it in and drag it onto the avatar. And it gets sewn automatically and placed on the avatar. There's no skill required much here. So we're gonna go through and, so now the design is sleeveless, so she took off the sleeves. And then she's gonna go in and change the openings and put facing on those armholes, 'cause she wanted facing. One of the styles had facing on the armholes. This has actually been improved slightly. You do both facings at the same time, but she's doing them one at a time, left and right. But it's, It's fully automatic. Somebody has defined what a facing does, how it's defined, how it's created, where it's placed, how it's stitched. All that's been defined, and she's just reusing that tool that's been created. And now we're going to go in and change the shape of the armhole because she didn't like the way it looked. The designer said, that's not what I want. So she's going to go back here and using again Bezier curve tools, just simple, very similar if you've ever used Adobe Illustrator, you're familiar with these tools, and then it's going to cut that new opening. Are we out of time? Oh, sorry. What's that? Okay, so we'll finish up. So you can watch this as we talk. So let's just ask if there's any questions before we get kicked out of here. So— Anybody want to ask any questions while this video is rolling? She's going to create all 3 styles. Yes, please. What's the process of creating? Sorry, if you— the process of— of creating stuff like finishes and prints? Yeah, so it's very similar to physically creating them. We have a tool inside of our software that you take the piece, if you have to fold it, create a new piece, place it, stitch it. It's similar to that process of doing that, yes. [Speaker:AUDIENCE] And so once the designer decides that's it, saves the pattern generated? [Speaker:PHILIP ROSE] The pattern is generated all the time in the background as you're doing it. Yeah, always changing, yeah. It's a DXF file. It could be exported as a DXF file, yes. Yeah. Any other questions quickly before we— Yes. [Speaker:AUDIENCE] Does this work with any kind of garment? [Speaker:MICHAEL STERNBERG] So we have to— we focus right now on the 4 styles— tops, skirts, pants, trousers, and also dresses. We're adding intimates, which will be released within this month, so bras and underwear, et cetera. And also we're working on outerwear. Any other questions? So does this take into account the shrinkage in the— Sorry, we have no more time. Yeah. They're going to kick us out. Sorry, but you can ask me afterwards. Thank you.