Virtual vs Physical: The Truth About Color Approval in Fashion

Hello, how is everybody? Ryan Stanley, good to be here. So we have a couple of questions that we did beforehand. If anybody has any questions, put them on Slido. But some of the first questions that we figured out together was, can digital color workflows fully replace physical approvals, or is a hybrid model still the only safe approach for ensuring accuracy and confidence across stakeholders? And I think that question was to Gerd and Doug.

Okay, well, I think it's— it depends very much on how far you want to take it. I think it's possible to go very far. There are, of course, materials that are more complicated than others, and that brings challenges that you need to address, maybe with also different measurement devices, different ways of moving forward. So what we think is obviously there's a huge percentage you can take with traditional color measurement and something that is done over years and years and years in dye houses and production facilities. And maybe if we are going to something that is more complicated, it's also getting into consideration that it's not only color, but also the appearance of the material. So there is the material reactions that you need to see. And this is something where obviously you are— you need a system that gives you more an image and not only a color value. And if you have a combination of both of those, then basically I think we can go 100% or Let's make it 98%. Never say 100%. And it's a matter of building trust into the system, right?

It's—

you probably wouldn't start with 100%, but the more you do it, the more you experience the advantages. And there are clear advantages. The farther you can take it.

Yeah, I mean, just to add, I think the digital color has been done for a few decades now, but I think the main thing is gaining confidence within the enterprise that we can trust the data on the materials and colors that can be accurately measured on a spectrophotometer. And once you get to that point, there is a percentage which is typically Somewhere around 80% of the product produced for a brand in a season can be done digitally. The other things are prints and patterns, which are typically done more looking at registration layout and the overall print. So those have to be done more visually or by looking at an image or taking a hyperspectral image, something like that. But I think the key is, is number one, you've got to do things accurately. You also have to prove by sampling and going through, say, a season or two, based on our experience, and to demonstrate to design teams and others that if we specify a digital color as a standard, and then the supply chain executes within our tolerances, then that enables us to get exactly what we specified. And that's our goal, is to enable the supply chain to give the specifier exactly what they asked for. And if we can do that digitally, then it provides speed, et cetera. So I think not everything yet can be digital, but I think a good percentage, and that's increasing all the time.

Can I add something from design side? I'm representing design today. The workflow for design for us begins at the very front when we're building the color palette. So for me, over the years, having seen the digital capabilities is so incredibly— I feel so incredibly hopeful. One of the things that I would say is a current challenge with virtual and physical is that when I'm starting off on a collection, building a collection with any team, really people need to be able to buy into the digital. And working on monitors and screens, that will affect the first decision-making of our workflow. So that is one of those challenges that I still am seeking some resolution, some guidance on how to navigate that well, ensuring that what we're seeing on our digital screen—

Accurate on-screen color.

—is accurate on-screen compared to the actual proto that we've received. Is that dye lot, is that symbiotic? Is it truly representative of what we're going to be emulating and bringing to bulk? So the entire workflow starts and falls back through supply, but it comes back into design because we are reflecting on all our hindsight business and everything like that. So to me, so far I've seen hybridization only because right now, as we know, there hasn't been 100% clearance on every sort of dye tactile textile that we do use. And there's complexity in some material. So until that is resolved, I think that is the reason why we have hybridization. But I'm looking forward to the future for sure on improvements in those areas.

[Speaker:CHRIS] Yeah, the only thing that I'll add is I think in order for this to be a success for retailers, brands, businesses, there has to be an introduction to the process, right? Breaking it down into almost a high-touch, a low-touch, and a no-touch model, right? What does work on materials to my brand requirements without issues? Can I pull that information from the suppliers out of the Natific system? Can I pull it out of the Color Digital system that I know it executes correctly? Okay, I might not have to worry about that. That's something that I can set aside, I can self-approve, and the businesses can move forward. Now I've opened myself up to the bandwidth, low touch and no touch, low touch and no touch. Now I've opened myself up to the high touch. I have to pay attention to, I have to take the time to look at these things that are the opportunities that I'm not sure about. And now I have the bandwidth to be able to do that. Okay. So the next question that was in, that was to us was, as AI plays a growing role in digital workflows, how critical is the quality and realism of the data we feed into these systems, and are we at risk of automating without relevance?

I'll speak to that one first. I think as we speak about the product quality that's going to be at retail, I think that's the most critical point here. And because of that, the quality of the data is hypercritical. So therefore, as we're assessing production lots globally, collecting all that data from a variety of suppliers that manufacture different components or multi-source items that are in the same color. Then what we want to be able to do is make sure that the data integrity is there. That involves accurate spectrophotometers, annual certification, supplier training, all those type things so that we can have confidence in the data coming in relative to the production that's ongoing so that when all of that lands in store, that we can have confidence that it's going to coordinate well, the metabolism between the different articles is going to be acceptable, and the consumer is going to be happy with the end product quality. So I think that's the main reason that the data integrity has to be high.

Yeah, I would also say it's, you know, It's, uh, if you start with something wrong, how can you expect to have something right in the end? And, uh, that's true for all systems, and it's also true for AI. So if you start developing on AI, I know we are all very excited at the moment, and I am too, because, I mean, what you can do is magic and it looks fantastic. But we also need to be absolutely clear, it's not based on realism. Right. And if we start to developing things in AI and then basically our suppliers end up trying to match what AI came up with, probably all the advantages that are coming with the AI are gone away in the production process. Right. So here we need to start with something that is valid. And, we all here on stage, we know that there are valid— there is valid data. Right. It just needs to be incorporated and it connects to so many problems. I mean, when we're talking about on-screen, then this is of course one of the main things. We are having a hybrid process here, right? We always now have something that goes into industrial production, that goes into a dye house. And these people, of course, I mean, they do spectral measurements since the '80s or even before. But on the other hand, we have this parallel approach in a digital world. And of course, we need to bring this together. And this is where all this kind of connects, be it on screen, be it, you know, on screen produced by AI. And so I think we need to make— we take this— we need to take this very serious because otherwise we'll have the same problem that we have with 3D. 3D is the same thing. We have all these beautiful productions in 3D. Do they really match reality? Sometimes they do, sometimes they don't, right? And sometimes in the same company, some do, some don't. And this is all according by not treating the data correctly and in a workflow that makes sense.

Yeah, I would also say, given all my history in working in design with various companies, it's really interesting because when we would work with 2D, And I would present the collection. There was never as much critique about color as there was when 3D started appearing. And then, you know, I would get asked, "Oh, is that accurate to the scale? Is that right? Is that the right color?" And more scrutiny would show up. And I think it's psychologically because as we're— as 3D and technology is getting better at being realistic, I think psychologically our eye is wanting to get that validation of proof and say, OK, is that— now we're tempted to say, is that really what that's going to look like? And that pushes an incredible challenge back onto us in terms of designs to say, OK, I need to explicate to you right now that we do need to understand how to read the digital images that we're seeing in terms of color. We need to understand what is our tolerance in accepting and understanding. Every presentation I have with any type of client, I always have the real color chips next to it because of the fact that I know that I will have to bridge that understanding in order to get the digital workflow going in terms of color down the line. And I think that's something that has become much more cogent as assets become much more realistic. And really understanding how to be very, very transparent, how to be very clear about what people think they are going to be receiving as a collection. Yeah, I agree with everything.

I'll add that. I'm going to coin a new term. I'm going to say, being here for the past 2 days, listening to a number of different presentations, I think what is really needed in the industry is AI squared, right? There's artificial intelligence and us, actual intelligence, right? So #AI squared, use it. Um, but no, I think that it's, it's broken apart into all these different pieces that we're talking about, right? If, if design isn't comfortable with what they're putting together upstream in concept, if that can't be delivered through execution and then ultimately delivered to the consumer, that then feeds back into the different retailers and brands that we partner with, then what's the point? If design spends time looking at colors that can never be executed, do they even have visibility to that fact? If the right data isn't in the system, if the right people, if the right metrics aren't being put in place to manage this metadata— I was looking at a number of different fantastic presentations. One of them, I believe, was a Raspberry group downstairs. I spoke with them for a little while. And we got into a conversation around, okay, what material information are you using to generate these images? What color data are you using to generate these images? Is it just a straight ASE file that's that color under a single illuminant? What if you have an inconstant color? What if the suppliers are submitting samples to that color and those samples are metameric? How to properly showcase that color across different screens. Is the screen that I'm looking at, does it have the wide enough gamut to display my colors? Just giving visibility to that information to team members, the actual intelligence side of that piece, to be able to make those high-level decisions, that's gonna, that's gonna be the win. So I'm gonna move away from the questions that we came up with ourselves. There's a number of different questions coming up down, down on the slide, but let me just ask the first one. Please explain the limitations and/or triumphs of measuring white samples digitally as compared to physical review. I'm gonna take that myself. I knew that. Yeah, you can do white digitally, matter-of-factly. You do not have to do it visually, period. Like, I have more to that, but you can matter-of-factly, right? And it's about just like with any color, it's about understanding the components. Typically whites are optically brightened whites. There are different classes and casts of optically brightened whites. There's red-casted optically brightened whites, there's neutral optically brightened whites, there's blue-casted optically brightened whites. They're not all the same, and they have different brightness values, they have different tint values. Understanding the components, right? The actual intelligence piece of the, of the AI, right? Partnering with companies like Natific who've developed white indices to be able to say, hey, I know exactly what colorants the suppliers are using. Is that in alignment with the colorants that were used to create my standard in the first place? If I ask for a last second increase in volume and the supplier pivots to a different mill, or I have to work with a different vendor, are they using the right optically brightened dye stuffs to be able to match the cast of my white? Or does my blue standard white now all of a sudden come in and it's tinted green, or it's tinted red, and now I have trouble with inventory? So the opportunity with whites is, yes, it can be done digitally, matter-of-factly, without question, But if it's not done right, the impact is huge because typically whites are core colors and the volumes are huge. So if you don't pay attention to those things, the pain can be severe. And that's why we instinctively want to do it visually, because we want to make sure it's right. But it can be done digitally, such is my opinion. I agree, of course. Yeah. All right, next question. Move on. All right. Brands still prefer— brands still prefer visual evaluation for color approval even when digital readings are available. What tolerance levels would encourage full digital adoption?

It's a question of what tolerance levels we are speaking about here. So first of all, I think it's— we are talking about change management here. We are talking about adoption of technology and you need to build trust into this technology. And I have no doubt at all that it works, right? I mean, we've seen I don't know, 500,000 of lab dips the last couple of years developing correctly, delivering on point, right?

97% was your presentation, right?

Yeah, something like this. So it's, you know, it just, it works. There's no question about it. Obviously still, and this is often if design is still involved in the process of getting the quality right, it's not adapted. I cannot really, you know, there are probably many reasons why. It's very tempting, of course, also for somebody who is really emotional in the topic, but I mean, we need to be clear about that at some point. At some point we are in an industrial production and then basically we need to also start to trust numbers. But that's a change topic. And I mean, the other thing to look at tolerances is of course what are the tolerances you accept as a brand? And that's of course totally individual, right? I mean, there are people who are fast and cheap and they will have lower tolerances and there are people who have maybe more time and more expensive products and they can have much higher tolerances, or it changes depending on which article you are manufacturing in one brand. But that's— I think there is no, there's no clear and easy answer to that. It's, it's change management. It's really, it needs to be changed in your heads, and it can only change if you develop the trust to it.

Yeah, I would agree. I think that in organizations, consistency, consistency Consistently, when I see brands implement standards, they go through the vetting process of going through their proof of concept and then can say, OK, here are some standards that we can stand behind. I think that is helpful to help build into the trust. And those standards would include deciding, based upon certain substrates, what would be the tolerance, or based upon certain regions, or based upon certain access to chemicals and dyes, what would be the tolerance? So I think that is something that every design company would want to have standardized, as well as understanding how to train everyone from the beginning, in internal and external suppliers, on what would be those brands' standards. That's, that's what I think would be key to success.

And I think there is not one tolerance level, right? There's also tolerance level on different materials. And you really need to look at what you maybe did in the past in order to learn what your tolerances are, because here we can pretty much say where your tolerances are. And I mean, I always see our tolerance is delta E1. I mean, this is the most quoted thing I hear. Yeah. So, but then when you look into what has happened the seasons before, it might not be really Delta E1 on certain types of things and it's okay, right? Maybe it doesn't need to be. Maybe it needs to be better or more loose. But it depends really on what we are talking about.

I think it's understanding, right? It's my opinion. It's management of expectations. Right, there are different tolerances in the industry for what we're looking at. There are color standard providers, ARChroma, CSI, ColorO, you know, they're developing their color standards to a Delta eCmc of a 0.5. That's a very, very tight tolerance. Typical design tolerances are 0.8 Delta eCmc, right? Production tolerances, we're dealing with volumes, that's a 1.0 Delta eCmc across a single illuminant, right? What tolerances am I paying attention to to encourage for, full digital adoption. I need to potentially look at reflectance curve matches. If you really want to match the color, the color is the curve, not necessarily individual color metric evaluations under a single illuminant. If you want to make sure to eliminate any types of metamerisms that would be in place, identify first if the color that you're dealing with is inconstant. Am I going to have a color that flares from light to light in the first place? I'm going to look at that standard as a designer in a design space, and then I, next to a window in Manhattan, and then I'm gonna walk downstairs and review it in a collection preview space with incandescent lighting. Is that color going to change? Does it have the potential to change? What colorants, colorant, not color, what colorants are my suppliers using to match that color across all my different components? So I would think if there's one tolerance that I would pay attention to, it's not necessarily traditional colorimetric tolerances. It's pushing the teams to pay attention to the reflectance curve to see how that color might react in any condition under all lights.

I'm just going to add one thing. And the initial comment in that question was that brands still prefer visual over digital. And that's not the case. We work with some of the largest brands in the world. And everyone is going toward digital color approvals. It speeds the development process, it reduces cost, saves time, is more efficient for everyone. So there are millions and millions of lab dips, first production, and production yardage approved annually around the world in over 40 countries. So you can have the confidence that a digital execution technology for color, it does work. So if you're interested in that, then you don't have to be afraid. There are ways, technologies, systems, et cetera, to help you to do that today.

OK. Next question. I think I know the answer to it immediately. What are the best technologies to aid in color accuracy? Maybe Natific and Color Digital DMIX? That'll be their answers. I don't know what—

Probably there are some more. But obviously, I mean, what I would say is, a good approach is if you start measuring colors with whatever instrument you have, right? It doesn't really matter. And, uh, over subjectivity, that's— yeah, I mean, that's our common ground here, I think. So basically, if you start to measure and be scientific about color, that's the best starting point, and then the rest will basically fall in place because it's all there. If it's Natific or if it's us or if it's whoever, right? It simply works as long as you find the right starting point.

Okay, can I jump in and say something a little bit radical? I think the best technologies are ourselves and our brains. Because when I am taking care of a collection, I will also think— and this is the exciting part for me— it's not just creativity, but strategically. There are certain programs where, like in womenswear, Right now, mixed media is hot. You've got a knit top with a woven sewn together. Mixed components are something where, okay, as design, we would take care to communicate and nurture that across the supply chain to ensure that, okay, when we are looking at what are the tolerances of two different substrates, we need to blend those factors together and say, okay, this is going to be dropping as one package. Or one capsule, or one— whatever those specifications are that we think would drive the experience for the consumer. Because ultimately, in the end, that's what matters. Because what was that statistic? Like 70%— 65% to 70% of a consumer's choice to pick up a good is based on color. Color is emotional. It's psychological. It's attractive. Or unattractive. And so understanding that when we're designing, yeah, it's great that we have these amazing tools to measure the data, but putting it in the correct construct to understand this is very important in the collection, this mixed media. When do I want to push the tolerance? I think that we are the technology that is going to be those deciding factors that makes judgment calls. And sometimes we have to refine what is our SOP. Sometimes we have to reevaluate what are our resources. Maybe resources are getting better. Maybe our facility got better chemicals. Things like that means that we are the technology making that judgment to help carry the systems through.

I think the whole tools are there for you. I mean, we're talking about tools here. We're not talking about something that does something magical by itself. You as a designer, you are driving this and you need to decide, of course, you know, where to go with this. But I think the important point is at one point we need to get away from the idea and we need to go to a value. And at some point in the process. Yeah. And then basically the rest can follow because, I mean, we just had a discussion before, you know, choosing a color in New York and then expecting somebody in Sri Lanka or in somewhere in China to know exactly and really exactly what you thought about this color will never work, right? It's trial and error. And this is what this industry is doing for years and years and years. And you, I'm quite sure you have all seen piles of lab dips of one color and not one was right. So that's where it's, where it's coming from.

Last 15 seconds. I just want to make a comment real quick. Probably can't answer the whole question, but what role does AI play in digital color approvals? How are you using it in this area? 10 seconds left. We don't have a lot of time. AI can play a phenomenal role. It can, it can do almost everything if the data is leveraged correctly. If you're paying attention to material information, if you're paying attention to the correct color reflectance information, if you're leveraging metadata from services like Natific and Color Digital, then AI can do just about anything that you want it to do.

And maybe it's not even needed right now in color approvals, right? I mean, we have all the technology in the world for this ready, sitting there working, and you don't need AI for it at the moment, right? There is, of course, there are algorithms and there is automation in this process. Yeah, maybe there will be AI on top at some point, but it's nowhere needed now.

Yeah, right now everything based on data and science is solid, right? But maybe the AI can just simply help supplement in terms of like, what is— what are consumers picking up? Like, what is the data that can support different— AI squared, use it— built, right? What are the most popular colors? Artificial and actual.

Sorry to interrupt you. I think we're done.

I think we're running out of time.

Thank you. Yeah, thank you.