Coty Inc. (NYSE: COTY) (Paris: COTY) has been undergoing a corporate-wide transformation, the “All-in to Win” program. Coty, with nearly $6 billion in revenues in its last fiscal year, is one of the world’s largest beauty companies. Coty’s headquarters is in Amsterdam. The company has a portfolio of brands across fragrance, color cosmetics, and skin and body care.
Coty announced the program during the height of COVID-related disruptions. The company’s goal was to boost its margins by significantly reducing fixed costs. Lower costs were to be achieved by simplifying its supply chain and achieving procurement savings. The goal was also to improve revenue management.
At the corporate level, despite a challenging market, the company generated over $700 million in savings from fiscal year 2021 to 2024. The company also achieved gross margin expansion of over 400 basis points and EBITDA margin expansion of 130 basis points. EBITDA is a helpful metric for assessing the profitability of a company at the operating level.
The supply chain, which does not include the procurement function, has contributed to this. The inventory has declined by 10%, service performance is flat despite a much more volatile market, and the company’s sustainability metrics have improved tremendously. There was an 82% reduction in their Scope 1 & 2 emissions. More than half of their manufacturing and distribution sites are now carbon neutral. And there has been a 16% reduction in water usage.
I interviewed Graeme Carter, the chief global supply chain officer at Coty, about the supply chain transformation that occurred as part of the ‘All-in-to-Win’ program. The interview was condensed for clarity.
Steve Banker: “Can you walk me through the supply chain transformation?”
Graeme Carter: “I joined Coty in April 2022, that’s now 1,350 days ago,” Mr. Carter said. “It’s no accident that I know the numbers. I’m about the numbers, as everyone in the supply chain should be.
“When I came to Coty, it was a good company making good progress, but it needed a supply chain transformation. Our service was good, our inventory was okay, our cash was okay, but our COGS – our cost of goods sold – was creeping up with inflation. And we had a fairly disparate organization.
“So, my role when I arrived was bringing the knowledge I learned – from the likes of P&G, Unilever, and Amazon – to a company in the beauty industry. With my team here, we developed a vision for July 2026, articulating what we wanted the future to look like using an Amazon approach – a working backwards document. We wrote the Herald Tribune headline or the Forbes headline, saying, ‘it’s now July 2026, and the Coty supply chain looks like the following.’ And then, ‘this is how we got there.’ And then for the last three and a half years, we’ve been working against that vision.”
That vision is based on responding to market needs with agility through simplification, standardization, and centralization.
“Where we can, (we need to) keep things simple. Don’t overcomplicate it. Don’t have a dozen processes. Have one process and document it. Make it simple, and then, if that works, make it standard across all of the sites, all of the markets, all of the countries, so that we can then improve upon that standard everywhere.”
There are now 350 people at their supply chain hub in Barcelona. Barcelona has manufacturing and distribution, but what makes it the hub is that this is where the global supply chain planning is done. This centralization enables simplification and standardization.
Simplification and standardization should not impede the ability to respond with agility. “The work in that hub (is done) very efficiently, using the scale of our company, with one team doing it brilliantly. But it recognizes that each market and each customer is slightly different. But there are archetypes. There are customers focused on cash. There are customers focused on service. There are customers in each country and in each region. We can recreate, in the center, a standard way of working with each of those archetypes.” This allows for both scale and the ability to respond specifically to market needs and customer needs.”
Banker: “So, Barcelona is where planning is done. What about the rest of the supply chain?”
Carter: “Manufacturing is where the majority of my people are employed. We have seven manufacturing sites around the world, primarily in North America and Europe.” According to their annual report, the company manufactures and packages about 81% of its products.”
Regional manufacturing primarily supports demand in its region, but if demand soars in one region, products can be manufactured elsewhere and then shipped to where they are needed. That rarely occurs.
The company also has 26 distribution centers across the world. Core inventory, raw materials, and WIP are located near manufacturing sites. Finished goods inventory is situated close to the consumers, “close to the retailers in the region where it’s going to be necessary.”
Barcelona is where the company’s supply chain AI initiative is centered and where its AI-based demand planning solution is employed.
Banker: “Presumably, regional manufacturing reduces your tariffs.” In their annual report, the company estimated tariff impacts at $70 million. But they believed they could mitigate $15-20 million of those costs.
Carter: “We had a short-term response to the tariff situation. We made sure we had inventory in the right place before the tariffs (went into effect). Then we’ve done some medium-term actions, changing the footprint of the factories so that I can produce locally for local markets. We’re looking at a number of longer-term strategies to better balance (demand and supply and to reduce global shipping). But that’s a significant investment. I’m holding off on that.”
Banker: “And to really get the tariffs down, you also have to source regionally. Have you increased your regional sourcing?
Carter: “Not as much as I’d like to.”
Banker: “I’ve interrupted your flow. Please continue.”
Carter: “We’re now on to what is next. This is AI.” The company has an AI-based demand planning tool.” They are using a solution from o9 Solutions. “And a factory maintenance tool, and AI is used in its safety program. “My friends in procurement use AI significantly as they do in marketing. But there’s so much in the pipeline. My intention is identifying the problems we have in Coty, and I don’t mean things that are broken, I mean things that could be improved.”
“I’m writing problem statements for those areas, then going to my friends in the digital team, saying, ‘this is a problem that could be fixed with a large language model, or a domain-specific language model, or agentic AI. What are you seeing out in the market that could fix these problems?”
“In parallel, I’m working with my team saying, ‘Look, I listen to podcasts every walk, every morning, on the walk to work, I’m reading this magazine. I’m talking to that consultant. Are you? Because this is the future.’”
“If you’re not investing in your own growth in the digital space, you’re going to be redundant in a few years’ time. I’ve literally used that term in a global supply chain town hall. I said ‘you won’t be needed here. I want talent in Coty (that is) excited and engaged and wants to be in the digital space.’” AI, in the long term, will increasingly do planning autonomously, and planners will be responsible for orchestrating the global network.
Banker: “I have a lot of questions here. Graeme. I’m an old supply chain guy. We have redefined AI to include optimization and machine learning. Both have been in supply chain solutions for over two decades.” So, from an AI perspective, what is special about the o9 solution?
Carter: o9’s “major competitors tended to have a solution that worked for you if you had a relatively limited number of SKUs and your drumbeat and direction was predictable and reliable. That’s not the beauty industry. We have a 10,953 active SKUs, last time I looked, across 71 markets in the world.”
And demand can spike if just “one influencer says ‘I like that shade of red.’ And suddenly our demand (for that SKU) quadruples that week.”
“So, we need a tool that has that ability to get right down into very (granular) details.” The tool needs to look back at which statistical modeling versions work best to predict what is currently happening in the market. Social media creates “a lot of noise in the market. I don’t know how to put that social listening, that collective sensing, into my online model, but we need to find ways, with Google search, with Tik Tok” of getting that social data into the forecasting models.
“We’ve got some data scientists on the team who are starting to understand – from what we predicted previously, to what actually happened – what signals that were not included that should have been. We’re exploring all of those and learning as we go. So that’s why o9 has worked for us. Some of the others would be a bit too rigid.”
Banker: “I just want to get a sense of the scalability of this. Are you forecasting at the SKU level? Have you segmented your audience by buyer behaviors? Are you doing daily, weekly, or monthly forecasts? Just walk me through that.”
Carter: “SKU level. Absolutely.”
“I was having a discussion with one of our commercial friends in New York. He was saying it’s difficult to forecast at the SKU level. I said, ‘well, great, but I don’t to make big investments and then be out of stock. So, you’ve got to tell me at an SKU level. If you want yellow instead of green, you have to tell me, because otherwise we’ll only have green.’ So, we’ve got to be at an SKU level. And the o9 model allows me to do that. At each SKU, by channel, by retailer.”
But a demand planning solution is not magic; it must be supplemented with human intelligence. “I keep reminding my team, look at what Walmart is doing and then what Target is doing.” If Coty runs a promotion for Walmart, Target will work to counter that promotion. “There’s a level of overlap there that we’re starting to understand how to take into account.”
“At the moment, we’re in a discussion about weekly versus monthly (forecasting). My current production program can adjust weekly, but I’d rather do it monthly because of inbound materials from relatively remote locations. But I’m playing with the demand forecast changing every week.” The team is being challenged to produce weekly forecasts of inventory dispositions.
But there is pushback. “Graeme, why are we doing all of the work to have a new forecast every week when we’re not going to change the production plan or the supply plan? That’s a lot of extra work that, at the moment, isn’t necessary. In two years’ time, when we’ve got the data science and the tools working really well, maybe a weekly cycle will make sense. Then I’ll talk to my manufacturing sites about better responsiveness. But at the moment, honestly, monthly is good enough.
Banker: “What about AI in manufacturing?”
Carter: “We have a tool where we’ve input all of the downtime on our packing lines, and it takes the operators through a logical flow of action, from check that setting, check this, check that, check the pressure here, take out this part of the equipment, put it back in. Is it fixed now?”
“It takes them through a very logical methodology that it’s been taught by our top mechanics. Now I could go in and work on our lipstick line and follow the procedures, and follow the pictures, and do that piece of work. This is based on a large language model.”
Coty has made progress with this tool. “In the last three months, it can now say, ‘look, skip those other five steps, because the last time we went through those five steps, it didn’t make any difference. Now we’ll take you straight to the step that fixed this problem the last time.’”
Banker: “What is the name of this solution?”
Carter: “Yoshu.”
Banker: “I’ve never heard of Yoshu.
Carter: “That’s one of the joys of the current world. I want agentic models, large language models, capable of taking the data, processing it, and then feeding it back. I just need a thin mesh connecting all of these little models back to my data lake, so when I get a better downtime maintenance program, I can just plug it in, and I don’t have to change the whole operating model. I can just add a new one.”
Banker: “Are you using Agentic AI?”
Carter: Not yet. But to get to an autonomous supply chain, agentic AI will be necessary. “My ‘intelligent enterprise” – as I describe it in Coty – the new supply line, is going to eventually be making decisions on behalf of the team. That’s the demand signal, therefore, that’s the supply signal, and it will do that automatically. I don’t think the technology is there yet, but I’m developing the skills and capability in my team to be ready for that, so that when it’s robust and it’s ready, my team can let the system run.”
“We’re looking at the disposition of inventory, and maybe the weekly AI modeling of my demand. I could start (using this to autonomously) set up those truckload movements. But I’m not sure of the value add of this versus just having a couple of people on the transport planning team with a decent transport management system making these decisions.”
Banker: “Are you having any black box problems with your AI solutions?”
Carter: “My team tells me o9 is completely a black box. They don’t actually understand how it got to its numbers. That terrifies me, being a traditional (practitioner), but I know that we’ve got 23 different forecasting methodologies in the o9 tool. 22 of them are good old-fashioned statistical modeling. And the 23rd is AI.” Only about15% of the forecasts use the AI model.” The AI model is where the black-box problem arises.
Banker: “Have we covered the AI topic?”
Carter: “I’m trying to be very careful with AI. I just think ‘digital.’ I had a digital transformation. I want the simplest possible solution. I don’t want all the bells and whistles. I don’t want it really complicated. I want it simple, standard, and central. I want it to work good. And then let’s move on to the next challenge, the next opportunity.”
Banker: “Are you looking for two-year paybacks? One-year paybacks? What would you shoot for?
Carter: “I’m looking for less than two.” With the transformation work they are doing, if the payback period is 3 years or longer, there is a risk that the process the tool supports will have changed before payback is achieved.
Banker: “Sustainability. What are the sustainability initiatives you have engaged in that were both Green and saved money?
Carter: “Four of my seven sites now have solar panels generating 13 to 25% of the site’s electricity needs.” This was not a two-year payback. It didn’t even pay back in five years. But it was the right thing to do for the planet. It’s the right thing to do for employee engagement, and frankly, the thing I wanted to do.”
Employees have greatly helped with their sustainability efforts. “One of my favorite examples comes from our Barcelona factory. My team came to me when I was visiting and said, ‘Look at this.’ It was inbound glass bottles with plastic-based packaging.” The boxes were on pallets with layer after layer of protective wrap.
“A third of the weight of the pallet was the packaging used to protect the product, because the bottles are very precious. And they said, ‘We don’t need this.’ ‘Well, great, but show me!’ And they did. They took away 70% of the packaging, and said, ‘Look, if it comes in like this, as we move it around the site, it doesn’t get damaged. The bottles are fine. Can you talk to the supplier and get rid of all of that packaging?’ So, our inbound packaging has now been reduced by 70% and we still have perfectly good quality of materials. That was driven by the team at the site.”
Banker: “As part of your transformation, do you use Lean?
Carter: “We use zero-based budgeting methodology and lean thinking across the sites. We have a site strategy and vision. And then a whole list of actions that we take to our CFO. ‘These are the five or six projects we want to do. Here’s the full cost justification, this one and this one payback in two years. This one pays back in four, but added up, it’s two- and a-bit years. Is that okay?’ Because sometimes those projects are necessary to deliver the overall site vision.” But in isolation, the cost justification does not work.
This bundling of projects is how the supply chain team at Coty has justified its AI and digital initiatives. The company intends to pilot AI solutions and engage in fast learning.
Banker: “These digital initiatives depend so much on clean data. Can you describe where you were and are in that process?”
Carter: “Where we were when I came in, was adequate master data in the world that we were operating. Fortunately, I have a smart team, and they were saying, ‘Look, garbage in, garbage out. Data needs to be fixed.’ I endorsed that. I put a team on it.”
Coty is now working to obtain more data from its suppliers and existing customers. If this data can be connected across the supply chain, they will have a digital thread that can be very useful, particularly for sustainability. For example, if they know where aluminum in some packaging is being smelted, “you can declare your taxes on sustainability.”
“We’re very nearly there. There are a few elements to deliver. So, the master data is now pretty good.” But this took “18 months of focus across R&D, the packaging team, and the procurement team. The company is “making sure the inbound data, particularly for the new initiatives, is robust so that it can be used downstream.”
Banker: “Any final thoughts?”
Carter: Mr. Carter is using a new metaphor to guide his thinking on AI and supply chain management. He uses the term ‘Supply Biome.” Artificial intelligence is more organic than inorganic, and the connections between suppliers – all the way through to the retailer – are becoming almost biological. “So, the economics are becoming biological. Think about Adam Smith, and the invisible hand. AI is becoming that invisible hand driving the supply biome to be more efficient and more effective.” Think about Darwin and the survival of the fittest. The fittest will be the ones who adopt AI most effectively and efficiently. Coty is going to win in that space.”
