Digital twins: Flying high, flexing fast

After discussing the fundamentals of digital twins and the opportunities they offer, we’re expanding our exploration by taking a look at how digital methods help transform product development. Join McKinsey partner Kimberly Borden and senior adviser Dr. Will Roper in a conversation about the value of digital technologies in production applications. Their discussion, which examines the use of digital approaches to design and development in Formula One racing and in aerospace and defense, shows how a software development style can have a big impact on the production of hardware.

This episode, recorded at Farnborough International Airshow and introduced by Christian Johnson, the managing editor of McKinsey’s Operations Practice, explores the digital imperative in the design and development of products. It also considers the need for a pragmatic delivery approach to ensure that big ideas can be executed. The conversation has been edited for clarity.

Christian Johnson: Your company’s future success demands agile, flexible, and resilient operations. I’m your host, Christian Johnson, and you’re listening to McKinsey Talks Operations, a podcast where the world’s C-suite leaders and McKinsey experts cut through the noise and uncover how to create a new operational reality.

Our previous episode looked at one of the most exciting examples of a digital transformation of industry, the digital twin, covering what digital twins are, where they can add value, and, most important, how to capture it. One of the guests in that discussion, McKinsey partner Kimberly Borden, is helping us push the discussion even further in an interview recorded at the 2022 Farnborough International Airshow. She’s speaking with Will Roper, a former US Air Force assistant secretary for acquisition, technology, and logistics and now a senior adviser to McKinsey, about the digital imper­ative in research and development—an application where digital twins are already generating substantial value. The conversation focuses on how he saw the potential of digital transformation in Formula One racing and on the need to dream big, but with solid implementation plans always in mind. Let’s listen in.

Kimberly Borden: Today, we’re joined by Will Roper, who is a former assistant secretary of the Air Force and led a lot of the digital transformation and acquisition for the Air Force. Thank you very much, Will, for joining us today.

Will Roper: Thanks, Kim. I’ll talk about this subject anywhere, anytime, but it’s especially good to do so around a lot of airplanes and space systems—an especially appropriate place to talk about digital transformation.

Kimberly Borden: Terrific. Given your Air Force experience in really transforming the way they thought about and worked on airplanes and the acquisition process, can you tell us a little bit about what made you realize that the change was needed and exactly how you went about implementing that change?

Will Roper: Good question, Kim. I did not go into the Air Force and Space Force acquisition job thinking “digital transformation” and that digital engineering was going to be the platform I’m going to push. It was actually when we awarded contracts for the T-7 program—the replacement for our current trainer—that I realized something was different about digital. The cost came in significantly lower than usual, and new airplanes were built a lot faster than we had seen in the past. So I got under the hood and tried to learn what’s different about digital and realized that the digital-engineering techniques that have been perfected in other industries—the automotive industry and Formula One—have now started crossing over into defense. I thought, “Well, this is going to be exceptionally potent. It should change how we do business, but I need to get smarter about it.” I tried to determine who’s leading the world on this.

By far, it was Formula One racing. I was very fortunate that just down the road here, the McLaren racing team was gracious enough to open up how they compete in races every week. I realized that what they do on a weekly basis looks like the way we do software—the same agility used for software was being used for hardware. That’s a powerful concept I’d love to talk about today with you. That’s where I got excited about it and realized that this approach can change how we compete against other militaries. It can change the cadence in bringing new technologies into programs and really create more opportunities for companies in defense acquisition. So I pivoted to digital transformation, digital engineering, as a key thing I wanted to leave behind. I wrote a lot about this because I was so passionate about it.

Kimberly Borden: That’s amazing. It’s one thing to have a vision; it’s another thing to get the entire Air Force as well as prime contractors on board. What were some of the things that allowed you to change many organizations and even the Air Force itself?

Will Roper: Well, the first way I tried failed. You see the successes after you hit the failures. But I ran around talking about digital engineering and model-based system engineering, and I heard that we’re already doing that well. We use computers for everything. Look at our computers, look at the CAD,1 look at the testing scripts that we run before we go to physical testing. I realized this is going to require a very different messaging approach, a very different branding approach.

I’ll give a lot of credit to my wife. She let me take our vacation and write a document called There Is No Spoon, a Matrix, sci-fi–inspired report that presented our digital-transformation strategy for the Air Force and Space Force. I wrote it that way very intentionally for a couple of reasons. One, I didn’t want it to look like anything that had ever been written in the government before. Most govern­ment documents, they’re good bedtime reading. They’re not written well; they’re often not written by the people who put their names on them.

I would always ask myself, when I was a younger person in government, why a document is important enough for me to read if it’s not important enough for the principal to write it. So the 36 or so pages were all written by me. I tried to write this in a different way. I really liked the idea of using The Matrix, a 22-year-old movie. The concept is that you can create a digital reality that in almost every respect can substitute for physical reality. We’re not there yet, but for engineering purposes we are. We should adopt this approach.

The concept is that you can create a digital reality that in almost every respect can substitute for physical reality. We’re not there yet, but for engineering purposes we are.

I also wanted to make the report readable. People would be able to digest it, and it would not have a lot of tech jargon in it. We could really use it to get people inspired to find the digital-transformation opportunities in their programs. A handful did a great job of doing that. These included the ground-based strategic deterrent, Sentinel, now the replacement for the Minuteman III; Next Generation Air Dominance, the B-52 commercial engine-replacement program; and the A-10 wing-replacement program. There were a handful of programs that did this. The more programs do it, the more teams will be inspired to follow them. It feels like you’re doing a continuous thing, but it’s actually fiercely discrete.

Kimberly Borden: It sounds like you used the vision and the strategy to inspire others. Were there learnings along the way as you advised companies how to do this—lessons that you think other companies should follow as they take this on?

Will Roper: That’s a great question, Kim. I’ve learned a lot since joining McKinsey. I left the government knowing what I knew and being able to do this in a handful of programs. That was great. So I joined up with McKinsey as a senior adviser, and the most common thing I’m asked to help on is digital transformation. There is a common script, when the door closes, for the first time with a client—they all say, “How far behind are we?” There is a general perception, especially in aerospace and defense, that everyone is behind, but that’s actually not the case. That’s the good news. Now, Formula One is way ahead, and I’m sure we’ll talk a little bit about what they do and why it’s amazing. But defense is actually doing a good job of leading the charge, if they continue to follow up. Usually, the first thing I tell companies is “You’re not as far behind as you think; you’re actually in a place to drive this industry.”

But then the second thing is that the C-suite level has to be involved. You can’t put this in your change management engine and have a digital transfor­mation come out as a by-product. This has to be driven through discrete events that are often painful organizationally, and potentially expensive, especially the modeling and simulation. It takes an investment approach. I think most of the clients we’ve engaged with have understood that and are doing a good job. It’s just not easy. If this was easy, it would have happened already. You will be blazing new trails for your company if you’ve got the right team and sufficient data. Others have done this, and you can too.

Kimberly Borden: You mentioned the team. In the Air Force and in companies, how do you tell them to build the capabilities they’ll need? Many of these digital capabilities are very different than the capabilities that exist today.

Will Roper: There’s no easy answer for this. If you and I start a software company today, we can deploy code on the same day we start because there is an amazing tech stack for developing code, which has so much automation for both developing and deploying. This continuous-integration, continuous-delivery approach to software is so potent and powerful. Now that containerization allows software to go out to devices with different kinds of computing onboard, it’s even more powerful than it was when it first burst out, about eight years ago. Software has an amazing tool set to do what is now being applied to hardware. Most companies will have to start by creating a development platform to enable digitization. I think that’s where a lot of them will lose a little enthusiasm when they learn that they are going to have to invest in this over time, though they really wanted a quick win.

The good news is that you get quick wins that have significant value before your full tool set is done. Companies have been really excited about being able to do things more in parallel—things that have typically been done serially in aerospace. When you can do things in parallel, there are more cycle times. Cycle times are a great metric that you typically want to track in a digital transformation. You have more time for engineering, not integration. Integration starts becoming easier when you bake it into the tools themselves. You can see these windfalls early on, such as certifying something digitally, far in advance of your doing something like Formula One. That’s the good news; with the proper strategy, you can get the return on investment quickly enough to keep enthusiasm and keep momentum. But most companies have needed advice to get there.

I’ve definitely seen most companies lean toward thinking that they have many digital-transformation ideas and a road map, but none of them actually crystallize and deliver something. My approach in the Air Force was that it’s great we have head-in-the-clouds ideas; we should always be dreaming big. But if there is no project that makes a lightning bolt go from that cloud to energize the ground—boots-on-the-ground implementation details—nothing will stick and there’ll be no template to scale. What I’ve been doing since joining the McKinsey team is helping companies to see where, across their portfolios, they can create these lightning bolts, where there is recurring value after they’re done. These are not science experiments that prove this is possible. Formula One proves this is possible. The goal is to now “productize” it for clients.

Subscribe to the McKinsey Talks Operations podcast

Kimberly Borden: We see, a lot of times, that companies struggle, even though they’re trying to measure the value, thinking about the metrics, and how to drive value all the way to the bottom line. To your point, there will be a great idea and they’ll start it. But actually capturing the value is very difficult. What did you do in the Air Force to ensure that happened? Was it the metrics?

Will Roper: The metrics are the thing that all organizations are going to rely on because when you do a digital transformation correctly, you’re in effect turning hardware into software. That’s what Formula One has done right. Their cars, which have a lot of hardware, are software. A digital transfor­mation is going to let hardware “eat the world” too via software.2 There’s a ton of methodology in agile techniques for tracking areas like how are you managing your backlog, how’s your cycle time, how are you doing deficiency retirement. All of that applies; you just have to put it on a timescale. Then there’s this peculiar thing: sometimes, physical testing and prototyping will still be needed if you’re pushing a boundary, because hardware compiles in the real world. We’re watching airplanes fly out here. Their computer is the universe, and we’re trying to substitute a metaverse for it.

If you don’t have the data to anchor that metaverse, you’re still going to have to prototype and test. But you should never do something in the physical world that you could do digitally. That’s what Formula One is great at. In the Air Force, we were good, early on, at knowing what we could fully shift to digitization. I would argue that without digitization, there is no way that the Sentinel program—the replacement for the Minuteman III—could succeed, because you can only test intercontinental ballistic missiles occa­sionally, four times a year. That’s not enough data to validate the physical testing, so digitization is absolutely required. It was necessity that made us clear-eyed about what substitutes for reality and what doesn’t.

If you don’t have the data to anchor that metaverse, you’re still going to have to prototype and test. But you should never do something in the physical world that you could do digitally.

Kimberly Borden: You mentioned data a couple of times. Many companies say that’s a challenge for them. What great ways have you seen people tackling that today, even though we’re probably not where we need to be eventually?

Will Roper: No one has enough data. That’s just a universal problem and certainly was true for us in the government: in the Air Force and Space Force. If you don’t have a data road map—a no-kidding data strategy that people are accountable for and that is resourced and that people are properly empowered to execute—there’s a lot you can do, but you ultimately can’t go the distance with digital transformation. In the Air Force, I approached everything by saying, “Let’s do one lightning bolt at a time until people can replicate these without the leadership having to direct them anymore.” The most common bit of advice on data has been that you need a team that does data. They’re thinking about how to unlock, collect, and change the data calculus.

I was with a McKinsey client, and they were saying, “We wouldn’t be able to understand what happened at this part of our process life cycle.” I told them you can change that. You can change the design of your system, so the thing that you say is not knowable now becomes knowable. The process becomes an input, from a data view of that program—it would not be an output from a hardware-based view, unless you made it someone’s day-to-day job. You’re going to be lucky if you happen to have a program manager who thinks like a data scientist and a program manager at the same time.

You can change the design of your system, so the thing that you say is not knowable now becomes knowable. The process becomes an input, from a data view of that program.

Kimberly Borden: Yes, it’s a tricky thing. What about leveraging ecosystems and other partnerships to harness some of those capabilities or data sources? Have you seen that work well?

Will Roper: It’s worked well. Especially in space, there’s a lot of commercially available data that you can begin with to start training algorithms, so you shouldn’t recreate what exists. I think the hardest challenge—it was certainly a very hard challenge for me in the government—is that when you hit models and simulations, aerospace and defense is especially model rich. I don’t know how many models and simulations we have in the Air Force, but it must be in the high thousands. All of those models are built by people for people. There’s a team of experts who know how to use them, and if you’re not an expert, you can’t use that model appropriately. You may get an answer, but it’s not a valid one. So if you ask a question—like can we fly an F-22 into harm’s way under specific conditions—the answer is simple: it’s either yes or no, with a certain probability.

But to answer that, you would have to take the question and start decomposing it into the right subset of the problem for particular models. People did that decomposition, people ran the models, people took the outputs, integrated them together, and then ultimately brought an answer back to a decision-maker on paper. The first thing you have to do if you want to do a digital transformation for hardware is to start directly connecting those models. That’s not simple. You would think, “Oh, how hard could it be?” Well, models replicate reality, and no one can do that perfectly. They’re making lots of estimations. Unless you can understand why those estimations happen to be congruent with each other, you can get model components that seem sensible but would be dangerous to base a real-world decision on. The hard work for us in the Air Force was ensuring that as we moved between different models, we weren’t picking up error sources we didn’t understand.

Kimberly Borden: How did you get buy-in to those models? That is another big challenge—getting people to trust the virtual reality that you’ve created.

Will Roper: It is. It makes us feel good and feel more confident, when we’ve designed something, to see it physically built and tested. That’s just human nature. It’s also fun—let’s not take that away. It’s fun to build things and watch them work. There’s a tendency to think that these things are real because you can go touch them and that the digital thing is not. But if you think about this from a certification standpoint, physical things are not the same unit to unit; there’s variation. The test conditions are not the same day to day; the measurement conditions are not perfectly accurate or precise. What we call ”real” is actually just an estimation. It’s electronic signals interpreted not by our brains, as Morpheus said in The Matrix; it’s just interpreted by computers. A model and simulation is just an estimation that has different error sources. The question, really, is can we minimize error to an acceptable rate that would work as if it were data on physical things?

The way we got the Air Force to agree that we could do this was by bringing in certifiers early. In a company, if you spring something on your quality or safety person at the 11th hour, you’re probably not going to get a high five. If you bring these people in early and have them start thinking about the different error statistics, from physical data to virtual data, you’re likely to find a path where you can do some early substitutes for physical tests. That’s what starts building the confidence that “Yes, we can now replicate many aspects of reality sufficiently that we don’t have to go build the physical thing and test it anymore.” That’s just recreating old data. We should have a fierce desire to end nonrecurring engineering. Engineering should be recurring or reusable; that’s what really happens once you get digitally transformed. Engineering should always be recurring. Occasionally, if you’re doing something that pushes new physics, that’s the time to physically build and test. Like Formula One, you stream a bunch of data off that system, and it goes right back to your digital model. Now your modeling ecosystem is even better.

Kimberly Borden: What you’re talking about is a pretty fundamental and massive cultural shift in a company—not just the capabilities that we were talking about earlier but also the way people function on a day-to-day basis. What did you do in the Air Force to drive that cultural shift?

Will Roper: I had a big advantage, which was luck. Luck is always your best way to get ahead. When I went to the Air Force, I knew you can’t focus on too many things. I’ve watched government leaders do that, and you can’t get anything at depth. If you don’t get depth, it doesn’t scale. If it doesn’t scale, you have no impact. It all ends up being talking points, so I wanted to be very, very focused. The thing I focused on first was software because I had been working with Eric Schmidt3 on the Defense Innovation Board, learned a lot about how the private sector developed software, and realized it’s not really a product anymore; it’s a process. They’ve “operationalized” it, to borrow a term, and we needed to operationalize it. I focused very, very heavily on that.

It was hard at first. But once we had that first success, the Air Force started looking like Google, and others replicated the process because it was not a one-and-done experiment; it was a scalable process. That meant bringing in contracting professionals, business managers, so it was not an experiment; it was a new way of doing business. When it was time to bring hardware to the dance, we had already had that change. That’s what I was passionate about. I was like, “Oh, this is basically agile for hardware. We can think the same way, we can act the same way.” That transferred a lot of the things that we had already been learning from software. But I definitely saw, in some parts of the Air Force—and I see it in companies—that you talk the talk, but there’s a product-centric tunnel vision that just comes back in.

The first thing that goes through your transformed process when you’re trying to do quick cycles, continuous integration, and continuous delivery, but in hardware, is that a voice says, “Too risky; it’s going to take too long, too much cost.” What do you do when that happens? You slow things down. That’s especially true in aerospace, where systems are complex and you have a lot of subsystems. If you have ten kids and you must leave at a certain time, the likelihood that one of those kids will be late and make the whole group late is very high. Well, that’s why defense programs are always late. There are a lot of kids that have to get in the van, and one of them is late. You can always be assured that if you say the program is likely to be delivered late, math itself is on your side, statistically.

Is that a wrong way of thinking? No, it’s exactly the right thinking in pre-digital work. If you need all of those subsystems on the platform to fight the war and if integration is hard, that’s the assumption that underlies this work. Since integration is hard—too risky—we have to design for all of the integration at once so that when we go into production, we don’t have any quality problem that goes from design to production. But in a digital transformation—and I have watched companies do this—there is no reason to wait for the kid who’s still getting ready. You can come pick them up later, right? I have watched companies design for something that didn’t exist yet. They go ahead with their platform, drop in the missing thing when it’s available, and go into operation the same day they dropped it in. When I see that, I think, “Wow, that’s process thinking.”

One of the amazing things about Formula One, by the way, is that over the course of a racing season, 85 percent of their cars improve. By the time they get to the final race, their first car would be the slowest car on the track. Formula One isn’t producing race cars to win. There is no race car that can win in Formula One today. It is a race car–building process that wins. They have operationalized a process and applied it to racing. That mindset is very hard to ingrain, right now, in government and in aerospace and defense. That was probably the biggest impediment—diverting back to product thinking.

Kimberly Borden: Very true. Are there other companies that you have seen do this well, in addition to the examples you’ve given? Have you seen other companies or even subindustries, outside the military, that are making advancements?

Will Roper: I have seen a lot of them, thanks to working with you, Kim; you’ve introduced me to a lot of them. I’ve seen some really great progress in agriculture, which I had no idea was digitally transformed. Agriculture pretty much looks like science fiction now, with the very precise use of artificial intelligence across almost everything that happens. I had no idea that when you’re harvesting something, you’ve got AI that’s looking at the height of what you’re harvesting and adjusting cut patterns and how much herbicide is sprayed, and even recycling the organic waste as fertilizer in real time. All that is happening because the physical farming apparatus is connected to its digital cloud. This is really living the Formula One model today for a very important problem—producing food. I’ve seen it in medical devices and in pharma as well.

I’ve also seen it start to emerge in the automotive industry with Industry 4.0 manufacturing, which is nearly completely automated. It’s coming. The reason I hold up Formula One is that it’s the only industry that has been completely transformed and shows us how fast the pace of competition may be in future. In aerospace and defense, how would things be if we had to deal with 85 percent obsolescence every year? Formula One’s obsolescence rate isn’t because supply chains ran out. Their obsolescence is the result of becoming outdated from a competitive standpoint. If we had to deal with that in the Air Force, our process would break, because we buy products. That’s the transformation that is still left.

The thing that surprised me after I left the government is that there are not more examples of successful digital transformations. The statistics show that most companies put a digital transformation in the top three priorities on their list, but very few of them think they’re seeing value. It’s that slippery slope between reality and near-reality. If it’s near-reality, you can’t create a digital substitute for it. If it’s reality, you can. It’s very easy to miss reality in your digital ecosystem and to create something that’s really just a better simulation. But that’s not the same as what we’re looking for—a substitute for physical testing.

It’s very easy to miss reality in your digital ecosystem and to create something that’s really just a better simulation. But that’s not the same as what we’re looking for—a substitute for physical testing.

Kimberly Borden: Where would you like to see product development or R&D in five years? Where do you think it’ll get to?

Will Roper: Because most industries are not races and have to produce things in quantity, they’re never going to go to the extreme that Formula One is at today. What I do want, especially in aerospace and defense, is to see the gap between R&D and production shrink. With software-enabled manu­facturing and with a high degree of automation coming into a lot of industries, I don’t think there’s going to be as big a case for mass production. This goes far beyond 3-D printing. Manufacturing is becoming a software-enabled enterprise. In fact, you can create a miniature metaverse for your manufacturing alone, and companies do. I think that’s going to be the trend. We can reduce the gap between R&D and production. You may not have quite the economies of scale you used to have, but automation and advanced manufacturing let you speed up cycle times to bring more R&D into your products. That is going to increase innovation.

My prediction is that the companies that harness this first will really shake up the markets they’re in. It is such a competitive advantage. They’ll have to do the pivot—think of themselves less as companies that make products and more as companies that manage a superior product-making process. It’s really the process that gives you agility. You can adapt to anything, just like Formula One adapts to changes in the racing season and to the weather.

Process gives you adaptability. I don’t know what the future is going to be like. But if I’m confident I’ve got the most agile, adaptable process, I can feel that I will lead my industry because I can “outturn,” to use an Air Force term, my opponent. And if I can outturn my opponent, I get to pull the trigger first.

Kimberly Borden: That’s great. Any other reflections or advice for companies as they try to take on this fairly daunting task?

Will Roper: A few. One, do it. It’s fun. It’s not as hard as people think, but it is hard, so go in knowing that this is not a cakewalk. Two, get help. I got help from people who had done this before. That’s how I was able to guide the Air Force. And end nonrecurring engineering—you shouldn’t have to relearn what someone else has learned. Get help, so that you get it right the first time. The most important thing is don’t wait. It’s truly amazing how fast innovation is happening via digital transformation. It is going to get worse if you don’t do this and better if you do, because the digital transformation is enabling AI to go from the Internet of Things into the physical world.

Software didn’t really eat the world and the internet. It ate the “dataverse.” It ate the places where data is produced; it’s naturally amenable to that. It has not crossed into the physical world, because it has needed a representation of the physical world as software. That’s this metaverse concept, a metaverse for engineers. We’ve seen examples, such as other racing examples: teams have used AI to train racers, and not just to perform well but, eventually, to perform better than any human team has ever done before, because these teams had such a frequent, rich source of digital data and, probably most interestingly, defied the old logic for that sport. If that has happened in competitive racing—and we certainly have seen it in chess and Go4—what’s going to happen when AI gets into other industries? What kind of old logic for captaining industry is it going to defy?

That’s why I think this will be an industrial revolution that turns hardware into software and lets AI have access to physical things that it can master. Once that happens, I expect the pace of innovation to increase even more. You don’t want to be behind that power curve. It’s already fast and it’s going to get faster.

Kimberly Borden: That’s great advice. Thank you so much, Will.

Christian Johnson: Great discussion. I was particularly interested to hear the story behind Dr. Roper’s famous paper, There Is No Spoon, and the need for digital transformations to be driven by senior leadership. They have to be top-down efforts underpinned by timely investments and capability building.

You’ve been listening to McKinsey Talks Operations with me, Christian Johnson. We’ll be covering more aspects of digital-twin transformation in coming episodes. Don’t forget to subscribe on Apple Podcast, Spotify, or wherever you hear them.

Explore a career with us