A CDAO’s Perspective on Industrial AI Platforms

For years, manufacturers have known that data holds the key to better, more efficient operations. They’ve invested in sensors, systems, and analytics solutions, all designed to capture what’s happening on the factory floor. But turning that data into real, actionable insights has proven harder than many expected.

Francesco Marzoni has spent his career at the center of that challenge. With more than two decades leading global data and analytics strategies at companies like IKEA, Nestlé, Bayer, and Procter & Gamble, he’s seen how the role of data, and of data leaders, has shifted from solving isolated problems to driving decision-making companywide.

“I’ve always believed the honest objective of a CDAO should be to eliminate his or her role—to get to a point where data is so well integrated across the business that it no longer needs separate stewardship,” Marzoni says. “Of course, we’re not quite there yet.”

In this Q&A, Marzoni shares his perspective on where manufacturing analytics is headed, why contextualization matters more than ever, and why full control over data and operations is quickly becoming table stakes for manufacturers.

The impact is extremely tangible. In manufacturing, when you introduce a new analytical solution or data-intensive capability, you can very quickly see a clear “before and after” and calculate the ROI. In other corporate domains, there are often so many variables that isolating the effect of an analytical solution can take months, if not years. But in manufacturing, the cause and effect are more direct, making the value of analytics easier to prove.

The first, now more than ever, is minimizing waste. With more and more supply chain disruptions, companies can’t afford internal disruptions on top of external ones. There’s no excuse not to leverage every available data point to predict and prevent product quality issues. Data is the number-one ally in helping manufacturers control what they can.

Second, analytics drives agility by democratizing the performance of your best people. A strong analytics system allows every operator on the floor to perform like the best operator in the company. In this way, knowledge and experience are embedded into the process itself.

And third, I would say supply chain optimization and sustainability. Manufacturing analytics provides a quantitative, granular view of your processes. Minimizing waste isn’t just about avoiding supply chain disruptions. It’s also about contributing to your company’s sustainability agenda and raising your ambition around your environmental footprint.

What’s changed most is the level of ambition. Ten years ago, no one argued the importance of data-driven decision-making in manufacturing, but the focus was more on experimenting with digitalization or improving specific processes. Today, expectations are much higher. Companies understand that soon they’ll be able to control almost every step of the manufacturing process. What once felt like science fiction—using data to continuously optimize quality or prevent deviations in real time—is becoming the baseline.

That shift is driven, in part, by the growing number of external threats, from raw material shortages to transport issues. Companies know they can’t control everything, but whatever they can control, they must. And manufacturing analytics is central to that.

For a long time, data has been treated as a means to an end, managed opportunistically and addressed only when needed. That approach is no longer sustainable. With the abundance of data we collect today and the growing number of use cases it supports, the challenge becomes more and more strategic.

You need to treat data in an “always-on” way—integrated, maintained, and ready to activate in real time when the business needs it. If you wait to clean or harmonize your data until the moment you need it, it’s already too late to realize the value you’ve identified.

It’s also important to manage your data in a way that’s disconnected from the specific purpose it was originally collected or analyzed for. A good manufacturing analytics solution should integrate data from all your IoT sources while giving you the flexibility to tap into that data later for new analyses or challenges. This way, you’re not starting from scratch every time a new need arises; you can instead activate those data assets quickly and with agility.

Finally, context matters now more than ever. A few years ago, it was common to have centralized data teams treating every dataset the same way. Today, whether data is centralized or embedded within functions, you need specialized domain experts, because the volume and velocity of data requires deep contextual knowledge to ensure the right data quality controls are in place. Where you might have relied on one partner for marketing, sales, and supply chain analytics 10 or 15 years ago, those areas are becoming highly specialized.

AI and analytics solutions today rely on more data points than ever. Companies already collect data in abundance, but what AI needs to reach its full potential is the expert knowledge that lives in the minds of the great workers of the great companies around the world.

For me, contextualization is the process of codifying that expert knowledge and embedding it into your analytics or AI engines, using it to enrich the data that comes from processes and machines. And it’s not just a one-off exercise. Context must be added and updated over and over and over again, incorporating the knowledge of different subject matter experts and process owners at every step of the way.

Ultimately, manufacturing analytics is not a question of choosing between technology or people. It’s about empowering people with the right tools so they can perform at 10x what they do today, while continuously feeding the right knowledge and context back into the system.

One major advantage is the ability to react in real time or even prevent key deviations in your manufacturing process. Many solutions have tried to solve this, but now, with advancements in data frequency, data volume, and model performance, it’s possible to act or prevent issues at the right level of ambition and down to the most granular level.

Another big shift is the ability to go both deep and broad across your manufacturing landscape. Historically, there was always a trade-off—you could have a solution that went deep on one production line or process, or one that gave you visibility across multiple lines or sites, but not both. Until a few years ago, it wasn’t unusual to see completely different systems, KPIs, and ways of tracking performance, even within the same organization, or sometimes even line by line, depending on the equipment vendor. With Industrial AI Platforms, you can go deep down to the lowest level of granularity of a given production line and process, while standardizing where needed to create visibility across all your lines, sites, or geographies.

Ultimately, it comes down to control. Avoiding disruptions in your manufacturing process means avoiding disruptions to stock availability, which is an important competitive driver. The same goes for preventing compliance surprises or major product deviations, which protects your brand reputation and equity. If you can prove you’re in control, down to the last detail of how you remove waste from your production line, that becomes a key driver for both your reputation and your sustainability efforts. More and more, this ability to fully control your manufacturing is directly linked to both top-line growth and bottom-line results.

We may not be fully there yet, but we’re close. Very soon, any company without an analytics solution that lets them monitor 100% of their manufacturing processes will be seen as outdated—the equivalent of a company 20 years ago still relying on paper instead of email. The stakes are simply too high. We’re nearing a point where this level of control is the minimum requirement to be seen as a professional, competitive player in your field.

Be clear on where you want to build and where you want to partner. Historically, data teams have had a tendency to build everything from scratch. You have internal data scientists and ML engineers who naturally want to build, but that instinct isn’t always scalable.

We’re shifting to a paradigm where it’s important to adopt external solutions, adapt them as needed, and assemble them alongside your own internal capabilities. Partnering is especially important in a space as vertical and context-heavy as manufacturing analytics. If every company tries to reinvent the wheel—integrating data, building models, running simulations, preventing deviations—even the biggest data teams risk becoming bottlenecks.

It’s also important to recognize the value specialized third parties bring to the industry as a whole. These players offer a broad view across the industry, helping optimize processes and allowing companies to evolve and mature together. The more you partner, the more the industry grows—and ultimately, that benefits your consumers.