The Operators: The Art of Being Least Wrong, with Patricia Coan
After twenty-five years running supply chains from L'Oréal to Pura, Patricia Coan has learned to get comfortable with something most operators spend their careers trying to avoid: being wrong. We talk about planned stockouts, the hidden cost of excess, why speed can matter more than precision, and what happens when AI compresses the distance between signal and decision.
By Kelly Breakstone Roth, Co-Founder & CEO of Prysmic · September 2026 · 9 min read
Patricia Coan has run supply chains for twenty-five years, from the global machine of L'Oréal and Jones Apparel to the beautiful chaos of founder-led brands, most recently as COO of the pioneering smart fragrance company Pura. She is also one of the best writers working in operations today, and if you want to understand supply chain, and how to navigate the shift into the AI era, she is the person you should be following. In this conversation she takes us from her days as a demand planner merging spreadsheets at L'Oréal to the AI tools that finally erased that work, and makes the case that runs through all of her writing: "The solution wasn't creating a better forecast. The solution was recognizing that the forecast was no longer relevant."
An operator is formed
I have to start with the writing. Your essays have become part of my morning reading, and I know I'm not alone. What got you started?
I've always been a bit of a writer. I went to a high school that had a big emphasis on writing, all kinds of it, creative writing and poetry alongside more technical, thesis-style documents. Sometimes you don't know the little gifts you get early in your life.
More recently I started thinking about the different stories and experiences I've collected. I've worked at big multinationals like L'Oréal and Jones Apparel, and I've worked at smaller hypergrowth, founder-led companies, and the differences between those kinds of organizations are so interesting. And the beautiful thing that's happened is that so many people from my past lives have been reaching out. A boss I had four jobs ago commented on one of my posts, then texted me, and we ended up talking for an hour. That's what keeps me writing: seeing that the stories actually make people stop and think.
You trained as an industrial engineer and you've run supply chains at both extremes, some of the biggest companies in the world and some of the scrappiest. Walk us through the path, and what that range taught you.
Industrial engineering, my degree, is essentially a supply chain major. It's all about optimizing processes, systems, and people. I originally thought I might spend my career designing factory layouts. I love a factory. Show me a filling line, an automated line, robotics, and I want to see all of it.
I started at a tech company in Silicon Valley, in MRO purchasing, then vendor quality, then program management at a subcontract manufacturer. After business school I did a stint in consulting, and even there most of my engagements were operational. It's a crazy pace, and you learn great skills very quickly, skills I took with me when I went back into the industry.
And L'Oréal was a dream. Ever since I was a little girl I wanted to work in the beauty industry, and it was as glamorous as I had imagined. My experience at L'Oréal was formative, because you see the entire product development life cycle in one place. L'Oréal has the scale to own almost all of it, the R&D, the manufacturing, so it's nearly fully vertically integrated, and with that scale comes deep specialization: everyone is an expert in one silo of a very big machine. Then I moved to the other end. At a small brand like Tarte, owning factories would make no sense, so everything is outsourced and everyone does a bit of everything. Neither model is wrong. The playbook that's right at one scale would sink a company at the other, and learning to tell the difference is what the range gives you. And I learned that I really loved the smaller companies: a different pace, where you get to do a lot of different things every day. That's the sweet spot I've landed in.
The complexity tax
So let's start with the rules that follow you everywhere. What are the truths you carry from company to company?
There are a couple that, from an operations perspective, are always true.
The first is that the SKU economics always win. You interviewed the CEO of Underoutfit, Daniel Malek, and I loved that interview, because he was dropping truth bombs left, right, and center. One of them was the size range. If you miscut it, you're basically hosed: you're missing the sales of a customer you paid to drive to your site, and you're sitting on all the other sizes no one is buying. That same problem exists in cosmetics. I am so fascinated by Rihanna's business, Fenty Beauty. I love it, it's wonderfully accessible, all the shade ranges, all the body sizes. But when some of her lines run 48 or 60 shades, how is she doing that? Because it's the 80/20 rule: 20 percent of those SKUs are driving all the volume, and if you try to cover everything, the minimum order quantities and the line changeovers make the rest economically inefficient.
Complexity is a tax that rarely appears as a line item on the P&L. Not every SKU deserves to exist simply because someone can sell it.
The second truth is the tension around the first one. Founders are very attached to all of the babies they've launched. At L'Oréal you don't have the founder mentality, but you have the marketing person who nurtured a launch through a three-year development cycle, and then maybe it isn't that great. How do you decide to let it go? It feels sort of ruthless. An operations person is analytical, nerdy, very organized, so we're just, "Hey, the numbers say this." But there are nuances, like serving underserved communities, that are real reasons to keep carrying something. The categories change. That tension never does.
The forecast is always wrong
You've written a line we think about a lot: "The most valuable skill in supply chain isn't forecasting. It's decision velocity." Where did that conviction come from?
Early in my career I had a team member, a very smart guy, who would come to me with his demand plan and he could not make a decision. And the thing about a demand plan is that the forecast is always wrong. You're simply trying to be the least wrong, month in and month out. So I would tell him: just pick. What is the difference between 500 and 450? That's the same number. Pick a number, and let's move on. We can't stay in analysis paralysis. I'm a big advocate of data-based decision-making, but you have to pick a strategy and then be prepared to respond, because there are so many things you're not going to know.
Competitive advantage comes from recognizing when reality has changed and acting before everyone else.
We had a product in Target that we thought we had an aggressive forecast for, and we could not keep it in stock. It oversells, and you say, "Okay, I'm going to increase the forecast 30 percent." That feels like a lot, and it's still not enough. And at retail it's not just the rate of sale: they're out of stock in 500 doors, and every one of those doors needs its presentation stock on the shelf. I have many, many instances of that decision not being fast enough. You place the buy, and, ooh, I should have ordered 20,000 more units. If you decide faster, you recover much faster.
So what is actually slowing the decision down?
A lot of it is who is the authorized decision maker. We try to have dollar thresholds, and above a certain amount it escalates. But part of it is just getting everyone in a room, having data that tells a story, and aligning on the decision. That process can take over a week. Everyone's calendar is full. You don't have all the data. The team puts it together, and I look at it and say, "No, this is not good enough." And the decision is never in isolation: if we divert resources here, what happens to the other campaign we had planned to support?
And the preparation itself used to be the job. You're pulling data from all these different places: download the Target sales from their portal, pull the inventory from the warehouse system, get the other sales from another tool, merge it all in Excel, and make sure every VLOOKUP is pointing to the right place. I did this work myself as a L'Oréal demand planner. Just preparing the data could take a day or two, before you could even start the analysis. With these tools, you're wiping all of that out. Here is the information. Now I can just look at it.
And oftentimes the signal was in the data all along. You as a human just couldn't quite see that this is aberrant behavior against last week and the same period last month. Then you see it again, and it's gaining speed. Without robust systems, you can't tell there's a sales trend happening for three weeks. And that means you've lost a lot of time.
Sometimes you should stock out
Here's an asymmetry we keep seeing. When the bestseller sells out, everyone feels it: the CEO is calling, the buyer is upset, the whole company knows your name. Too much inventory hurts just as much, but nobody calls a meeting about it. Doesn't that quietly train every planner to overbuy?
If you have a hero SKU that goes out of stock, whatever the reason, poor forecasting, or an influencer you weren't expecting talked about it on her TikTok, the consumers have found it and they love it, and you are now in the limelight. Everybody wants to know how did this happen. We're missing sales. Target's upset, and the salesperson is having to go talk to the buyer.
You have to have a certain temperament to take the heat. People always say about me, "Oh, you're so calm." And I am calm, but the duck's feet are just going, going, going underneath.
Okay, what rabbit can we pull out of the hat? You come to the table with options, we could divert this stock, these are the choices we have, and then you run a postmortem on how we got here, so you don't end up in that place again.
But the excess side is sneakier, and it's often the worse deal. That inventory ties up your working capital, and depending on your business it's really hard to get out of.
The way the math on days of supply works, as demand starts to drop, your days of supply doesn't increase linearly. It increases exponentially. The same 300 units was 60 days of supply, then it's 90, then suddenly it's 100.
A luxury brand doesn't want to liquidate through a secondary channel, because it's brand eroding, so you run a controlled flash sale, and there's a limit to what you can move that way. Beauty products expire. And if you have to throw it in the trash, you get nothing. Oftentimes the cost of having too much inventory is actually more than the cost of the air freight to chase the winner.
And this, to me, is the harder conversation: people have to get comfortable with periodic, almost planned out-of-stocks on C SKUs.
That is how you really get efficiencies. You run your C SKUs leaner, because that demand is patchy and usually not very profitable, and you prioritize your strongest SKUs, so those are never out. This is what a rigorous planning process is for: you take strategic risks on the items whose volatility indexes toward the upside, and calculated risks on the slow movers. A lot of people don't want to have that conversation, because salespeople want every arsenal in their toolkit at all times. They never want to commit to anything. But operations doesn't create value simply because it's efficient. It creates value when it improves the economics of the business. That's the math you have to be willing to stand up and defend.
No playbook
You lived the moment that tested all of this at once, and wrote about it: spring 2025, when the tariffs hit.
Like many people who have studied economics, I initially dismissed the growing discussion around sweeping tariffs. Surely policies of that magnitude wouldn't materialize. I was wrong. Practically overnight, we found ourselves navigating tariffs approaching 150 percent on products manufactured in China. For a subscription business that relied on giving away hardware to acquire long-term customers, the math no longer worked. There was no precedent. No playbook.
What followed wasn't a supply chain exercise. It was a leadership exercise. My team and I shifted from executing plans to building optionality: extended payment terms with manufacturing partners to preserve cash, accelerated production with suppliers outside of China, inventory held at origin while we monitored policy developments and timed shipments, Foreign-Trade Zones to delay duty payments until the inventory was needed. None of those decisions happened because we had perfect information. They happened because we had built strong supplier relationships, invested in contingency planning before the crisis, and empowered the team to adapt quickly as conditions changed.
Supply chain resilience isn't built during a crisis. It's revealed by one.
Build versus buy
AI has been inside almost every answer you've given, so let's get practical about the tooling itself. When a team decides to bring these capabilities in, the first fork in the road is build versus buy, and you lived that debate inside a company full of engineers. Where do you land?
Buy versus build was a constant conversation at Pura, because we had so many technology people, and tech people love to build something. They want it exactly to their requirements, and they don't want to modify an off-the-shelf tool.
And I just think it depends. Where I land is: build when the thing is truly core to how you do business, when it's so specific to your model that it doesn't really exist off the shelf. Buy when mature tools already exist, because those tools carry years of accumulated understanding that your team would have to relearn from scratch.
Give us the example of a build that was right.
At Pura, the business is a smart fragrance device, an IoT device that plugs into the wall, connects to an app, and diffuses from a huge catalog of fragrances, and the heart of it is the subscription. If we had low stock, we always wanted to prioritize the subscription business, because we did not want subscribers to have a negative experience. The cost of acquiring a new customer versus keeping the ones you have is significant.
The thing is, the subscriber count changes all the time. It's summertime, I no longer want my pine scent, I would like a citrus scent. The holidays are coming, I'd like something cinnamon. One month a fragrance has 50,000 subscribers, and within that month it becomes 41,000. One that was 9,000 is suddenly 18,000. A B SKU becomes an A SKU overnight, because all these people just subscribed to it.
So the team built an AI tool that watches all of those movements and signals: there's a big subscriber change, we need to protect inventory. It would remove that item from the website for one-time purchasers, so that if you weren't a subscriber, you couldn't even see that we sell it. The subscribers could. That made a big difference in fill rates and out-of-stocks for the people who mattered most. There was no off-the-shelf tool that understood our business that way. That build was incredible.
And the other side: the build you fought?
When I wanted to bring in demand planning software, an established tool, the team was intent on building that themselves too, and I was extremely against it. Could they build a great tool? They probably could. But that team does not in any way understand demand planning. They understand the basics, not the intricacies, like why you have to forecast at the retailer SKU level. We had SKU A in D2C, SKU A in another channel, and SKU A at Target, but in Target it's a six-pack. For the manufacturer it's all the same SKU, just a different configuration, and if your systems can't recognize that, you cannot plan it. These tools already exist, built by teams who have spent years understanding what matters in demand planning, and the newest generation of them is AI-native. Why are we going to do six months of development to relearn what a partner has already perfected?
What is our core competency? Is it this? Because if it isn't, why are we the ones building it?
So when you do buy, how do you run the process, and what worries you when you're choosing a partner?
We always run a standard RFP process, and the heart of it is scenarios: real pain points from our own operation that every provider has to walk through, showing us how their tool would handle them. Then it's the criteria that decide it: how does it scale, and how does it integrate, with the warehouse management system, with the enterprise system the whole company is about to move to.
And with AI solutions specifically, there are a lot of cloud-based tools that look great, but then you wonder: where is all my information going? What's the data integrity of it? If that system gets hacked, is everyone going to be able to see my margins, my rates of sale, the different deals I have with different retailers? Every retailer has a different average selling price. It's all negotiated. The moment that becomes public, that's a problem. The bar for anyone sitting on my operational data is enterprise-grade security. Full stop.
A better first job
How is AI changing who you hire, and what should someone entering supply chain right now expect?
On my last team I had people close to retirement and people who were literally in college yesterday. At every age, you want people with a disposition to learning and embracing these tools. Sometimes you have team members who are set in doing things a certain way and resistant to adopting technology, and that is highly suboptimal. The one thing we can count on is that things will change.
For a young person starting out, some of these tools are basically analyst-level tools. So I don't necessarily need my college entry-level analyst, because the person who was doing that work is now these tools. Which means the person coming in almost needs to come in a step higher: more decision-oriented, more strategic, instead of spending the day on the robotic work of pulling data, inserting formulas, and sorting it all. That work has already been done. What I need from you is to tell me the decision, the direction we should be going in.
I don't necessarily need my college entry-level analyst, because the person who was doing that work is now these tools. The person coming in needs to come in a step higher.
And honestly, that work is more interesting, so that should be exciting. My own first job out of college, in MRO purchasing, was a lot of requisition processing, and I kept thinking, why did I go to college? This is paperwork. The new version of the job is more challenging and more engaging. But you have to bring critical thinking, and you have to be ready to actually decide.
One caution for the company side: make sure the tools can do everything you need them to do before you wholesale eliminate people. Some companies have moved too quickly, and the functionality is not all there. AI should absolutely eliminate repetitive work. But eliminating repetitive work doesn't mean eliminating people. It creates capacity. And leadership determines whether that capacity becomes cost savings or competitive advantage.
