In this Q&A moderated by Jon Sobel, CEO of Sight Machine, manufacturing data leaders discuss the challenges and opportunities of digitally transforming the chemicals industry. Matt Smith, Ph.D., is Senior Vice President of Digital Transformation at Sight Machine, where he advises global manufacturers in chemicals, oil and gas, and pharmaceuticals on digital transformation. Hussain Mahmood is General Manager, Customer Success, US Manufacturing Sector, at Microsoft. His team of more than 300 architects, engineers, and customer success managers is responsible for a $3.5 billion / year customer portfolio.
This is an edited version of a December 2021 webinar hosted by Sight Machine. The full session is available here.
An Industry in Transition: The Digital Imperative
Jon Sobel: Hussain, you spend a lot of time as a leader at Microsoft interacting with customers. What’s top of mind for your customers in the chemical industry as they continue to invest in Industry 4.0 and digital transformation? What kinds of things are coming up in your conversations?
…organizations that leverage data to drive their transformation have 54 percent higher revenue and profit growth.
Hussain Mahmood: First and foremost, top of mind for most chemical company executives is what does digital disruption in my value chain mean for the future of my business. The second thing is how do I accelerate my Industry 4.0 transformation. The need for operational resilience and flexibility is universal across chemicals. The demand for chemicals is slowing. There’s a lot of pressure to reduce emissions. The market is becoming more volatile, which means you need to have more flexible production methods. And on top of it all, you have the pandemic, which just pushes the need for Industry 4.0.
A few years ago, people asked, “Hey, do I really need to do this now?” But today it’s all about, “How do I accelerate digital transformation?” Now it’s, “How should I think about my roadmap? What use cases should I prioritize? Where is the greatest ROI? What technologies do I leverage?” These are the questions that are coming up. At the same time, everyone is concerned about the growing skills gap, specifically the shortage of digital skills. Every chemical company I talk to is trying to figure out how to attract talent.
Matt Smith: Yes, no longer do chemical plants have 20-year veteran operators who can wander into the plant and tap a pump in exactly the right place to get it going again. It’s a more transient workforce. There’s a loss of skills, a loss of knowledge in a lot of facilities. People want to be able to replicate that with digital data, to help them train employees and fill that skills gap. The other thing we’re seeing in the chemicals industry is an increasing emphasis on sustainability. The question now is, “How do I use my renewable and nonrenewable resources and not jeopardize my future operations?” And, “How do I communicate with the public about my carbon emissions?” That’s becoming more politically driven, regulatory driven, and is also a huge focus for investors. All of these things are going to shape how the chemical industry moves forward in the next decades.
…the level of impact you can drive through Industry 4.0 is directly correlated to how robust and real-time your data foundation is. This is where a company like Sight Machine comes in, which is why Microsoft works with Sight Machine.
Jon Sobel: Hussain, what are some of the things that Microsoft is doing to help chemical companies turn the above challenges into opportunities?
Hussain Mahmood: A lot of companies have pilots in progress but some 70 percent of them are stuck in the pilot stage. They’re struggling to figure out how to scale across the enterprise. If I look at the companies that have been successful, my observation is they do a few things well. One, they build their Industry 4.0 business case over long horizons – typically three to five years. They understand how to operationalize a digital industrial platform. There’s tremendous economies of scale and scope.
Secondly, the leading companies tend to have a list of use cases that build on each other, versus one-off use cases. They start small but have a roadmap with scale in mind. Thirdly, companies achieving digital transformation have senior leaders who can connect dots across business units and articulate the value of Industry 4.0. If you want to scale, you need executive sponsorship. They also treat data as a strategic asset. Research shows that organizations that leverage data to drive their transformation have 54 percent higher revenue and profit growth. I see this anecdotally in our customers. Some home in on their business goals and work backwards from them. You have to think end-to-end if you want to be successful – single choices constrain you.
Leading companies are also adept at building strategic partnerships. You cannot solve these complex data challenges in manufacturing alone. You need help from the ecosystem. And that might include cloud providers like Microsoft. A lot of companies try to do this themselves and get overwhelmed. They spend a ton of money and fail. It is our responsibility to make this ecosystem work for them.
Also, a high-quality data foundation is the cornerstone of any Industry 4.0 program. In fact, I’ll go as far as to say the level of impact you can drive through Industry 4.0 is directly correlated to how robust and real-time your data foundation is. This is where a company like Sight Machine comes in, which is why Microsoft works with Sight Machine.
Use Data Foundation to Unlock and Scale Your Productivity
Jon Sobel: Hussain, Microsoft pioneered the use of the term data foundation. Can you define data foundation and explain why it’s an essential element of digital transformation?
A data foundation is taking that manufacturing data, wrangling and managing it, and making it ready to be analyzed logically, so that you can model the reality of your plant floor.
Hussain Mahmood: Manufacturing data doesn’t lend itself to be analyzed with traditional data tools. We talk about the three Vs of data: volume, velocity, and variety. The variety is different because the diversity of systems on a plant floor dwarfs the amount of diversity in a traditional IT data center. So the variety of data is immense. And then the velocity, right? You have data that streams millions of points per minute. A data foundation is taking that manufacturing data, wrangling and managing it, and making it ready to be analyzed logically, so that you can model the reality of your plant floor.
Once you have a data foundation, you can analyze a machine, a process, a specific line, or a batch. Building this analytic data capability is how I describe a data foundation.
Jon Sobel: Hussain, Can you talk about streaming real-time data and why it’s essential in manufacturing?
Hussain Mahmood: Most plants work with data that is 99.99 percent batch analysis. But we need streaming data – real-time manufacturing data. To have any hope of preventing problems, you need to know what’s happening in your plant in real time.
Batch analysis gives you descriptive insights into what you did yesterday or last week. But if you want to go from descriptive statistics to predictive statistics – predicting what’s happening at this second and in the future – you need real-time streaming data.
Matt Smith: If it’s not real time data, it’s not that helpful. All over the world, people gather in morning meetings. They look at reports for the last 24 hours that are already post real-time situations. Plants have production issues now. So we have to flip the mindset from, “What happened yesterday?” to “What is happening now?” That is crucial. And you can only do that if you have a platform that gives you real-time streaming analytics.
It has to scale, too. Because next week there’s going to be a new problem. And then the following month, there’s gonna be another problem. And every time you create a new point solution, or you spend effort on a problem, you need a new data set.
So we have to move on from what happened yesterday, and move away from troubleshooting. We need a data foundation that is contextualized, where the data is relatable. It’s exceptionally empowering to any organization to put that in perspective.
One of the things we touched upon is the supply chain. If one looks at investor reports lately, certain materials are out of stock or there are issues with plant production, which often ties back to plant reliability. When you have these mega facilities, if one of them goes down, it affects the global operation. And so people are starting to think, “How can I use my data tactically to keep my plant functioning reliably?” If you have a data foundation, you can take a holistic approach, where you say: “Well, let’s not look at that in isolation. Let’s look at it as part of our whole ecosystem.”
Jon Sobel: How does having a real-time data foundation give companies a lens through which to see their operations?
Matt Smith: One of the things our clients get the most value from is having their data in one place. They don’t have missing data; they have clean data. And it adds a fourth V to their operations: Veracity. How accurate is the data? If data isn’t truthful, it shouldn’t be used. With the sustainability movement, people more often are saying, “I want to know what my energy usage is on each pump, or on each piece of rotating equipment.” Or, “I want to know what my efficiency rate is on each boiler.” We can certainly help with that. But often it’s not about the magnitude of energy you’re using; what’s essential is the energy per unit of production. It needs to be framed like this: How do I create this energy or water consumption per unit of production? How do I minimize the amount of materials that go into every piece I make? And the way to do that is to run as hard as you can, reliably and safely. But that requires a new way of thinking, such as: “Hey, I need to be able to move faster and I need to know things as they are happening.”
Jon Sobel: Matt, you talk about the benefits of developing real-time data foundations inside chemical companies. But how do you get companies and leaders to trust in the data?
Matt Smith: It comes down to change management. It is the question of, are you ready to accept data and do the change management? At Sight Machine, we do hackathons and data-mining events with our customers. We start to interrogate customer data, build up confidence. We identify one or two use cases where we can succeed, so that people in companies are like, “Yeah, I see that. I see the return on investment.” It’s about building that coordination with the client. It’s not just, “Hey, here’s your software, goodbye.” Trust has to be built on continuous interaction, in a way that makes for a true partnership.
The Digital Journey: Leadership, Governance and Technology
Jon Sobel: Hussain, at Microsoft, you are thinking about digital transformation at the industry level. You’re working with companies on ambitious projects. How do you generate trust in data foundations and data strategy?
A successful transformation needs champions at the IT level and at the plant level.
Hussain Mahmood: You develop trust in data over time and you build that trust by asking questions of data incrementally. So in the context of manufacturing, if you’re starting to build a data foundation, it’s about honing in on a use case. Maybe the case can be reducing downtime, where you contextualize the data to solve that specific problem. See what the data tells you and if it matches your hypothesis. If it does, you are off to a good start. Also, don’t start with a complex question. Start with a question where you already have some insight and you want to confirm it.
Once you believe the data is good, it is easy to expand. You build a data foundation by starting to ask more questions of that same data.It’s almost like an Excel table, right? You’ve got rows. You can keep adding as many columns of data as you want. And the more columns you have, the more questions you can ask of that data. That’s how you build trust. You ask a prioritized list of questions, and if you get the right answers, you know that the data is good for solving problems. And then you scale to more use cases, and soon you realize that, “Hey, this data is actually holding up pretty well.”
Jon Sobel: In terms of a company’s organization, who must be involved for a digital transformation to be effective? What teams or leaders must participate?
Hussain Mahmood: It must be a combination of people on the operations and the IT side. From the operations side, the chief operating officer, the VP of manufacturing, VP of quality, and the VP of manufacturing. These people are almost always involved. And then the CIO, or the chief data officer if the company has one, or the CTO. None of these leaders can do it themselves. If it’s just the CIO involved, your chances of success are minimal. If it’s just the VP of operations involved, again you have minimal chance of success. You need some combination of people from both sides of the aisle. If it is a large-scale digital transformation, it is going to require CEO sponsorship. In companies that have scaled the data effort across the enterprise, the CEO was part of the discussion.
Matt Smith: There needs to be executive sponsorship, and the company has to be committed to the digital journey. It needs to have that culture of support: “Yes, we’re going to do this.” A successful transformation needs champions at the IT level and at the plant level.
Jon Sobel: Last question: How long does digital transformation take? How long realistically before companies can start to see results?
Matt Smith: So it’s often surprising to people that results can be weeks to months. Anyone who has gone through a large-scale data implementation may expect a payback to be three years or five years. Once you create that data foundation, though, you can open it to your whole organization. Once your organization is confident about data sharing, everyone can look at that data and start to do the mining. Often just having that data visibility starts to transform your organization in ways that are unexpected and rapid.
Jon Sobel: Hussain, do you have good news to report about how long digital transformation takes?
Hussain Mahmood: Yes. Companies overestimate how quickly they can scale but underestimate how quickly they can get something meaningfully accomplished. That’s the key observation.
Companies overestimate how quickly they can scale but underestimate how quickly they can get something meaningfully accomplished. That’s the key observation.
I can give you an example of one of our customers, a major paint company. They have a thousand employees working on color prediction, making sure they manufacture just the right color using machine learning and AI. It took them two years to launch a new color. But we helped them automate it and reduced the time to just weeks. And it took less than two months to build a solution that helped them do a granular level of color matching.
The key takeaway is this: It’s probably going to take you longer than you think to achieve scale, but to achieve something meaningful will take you less time than you think. In a matter of weeks – not months – you can solve one problem and begin your journey to digital transformation.