Join the Insider! Subscribe today to receive our weekly insights
David (00:01)
Welcome, you are listening to the IT/OT Insider Podcast. I’m your host David. Subscribe today to get the latest insights in Industry 4 .0 and Smart Manufacturing. Today I welcome Sophie van Nevel. Sophie is the Global IT Lead for Strategy and Governance and also Data and Analytics at Bekaert. Bekaert is a global player in Steelwire Transformation and Coating Technologies. Before joining Bekaert, she worked at several Belgian banks.
In this talk, we again try to bridge the IT OT data gap. Sophie, thank you for joining me.
Can you tell me a bit about yourself and also maybe about Bekaert? What does a day at the office look like for you?
Sophie Van Nevel (00:48)
Yeah, thank you, David. Yeah, I joined Bekaert almost three years ago now, and I’m leading our strategy and governance team, but also our data analytics teams at Bekaert. And in the past years, in fact, the team has grown quite a lot on the level of data and AI because we had originally only a data platforms team. But in the meantime, we put also quite a lot of focus on data governance and on the AI part.
And specifically what is Bekaert doing? Like you mentioned already in the introduction, Bekaert is a steel wire production company. We do manufacturing of steel wire for all types of applications. You cannot name it. We’re in Champagne, where you open a bottle of Champagne, but also in big bridges, in floating offshore wind, but also now in…
Energy transition, we’re very active in. So we’re really active in a lot of different sectors and industries with our products.
David (01:56)
And is that a global player? I think you have production facilities all over the world.
Sophie Van Nevel (02:01)
Yes, that’s correct. So Bekaert was originally founded in Belgium, but in the meantime, it’s indeed a worldwide company with indeed production sites all over the world.
David (02:13)
I do love my bottle of champagne, so every time I open one, I will now remember the name Bekaert.
Sophie Van Nevel (02:19)
David (02:23)
So you are producing steel wire, that’s a very specific process. Before we dig into IT versus OT or IoT plus OT, can you elaborate a little bit more on what does it take to produce steel wire and what is the role of sensor IoT data in that production process?
Sophie Van Nevel (02:47)
Yeah, so we produce steel wire and in fact it’s making sure that the thickness of your steel wire is becoming smaller. We add coatings on top or we bundle back together certain wires to make sure that we have new material properties. And important in there is that we can measure certain thickness, certain parameters that we want to measure. And that’s why sensor data quality is important in
that whole production process.
David (03:19)
These are probably gigantic coils with kilometers and kilometers and kilometers of wires on it.
Sophie Van Nevel (03:28)
Yeah, it’s a lot of spools that we have, so it’s our spools with wire on top of that and we make sure that we deliver a lot of different spools to our customers. Yes.
David (03:41)
So you are responsible, your group is responsible to build digital products. What is a digital product? What are the types of applications users at Baycard are? I would say what data are they consuming? What applications are they using, et cetera?
Sophie Van Nevel (03:59)
Yes, so in fact, my group is responsible for what we call data analytics and we do that in close collaboration with our teams at Business side, which are focusing on what we call intelligent processes. And intelligent processes is in fact a program that we have within Bekaert to really optimize our production processes by the use of technology and digital and data elements.
This is a specific one where indeed we focus on how to use, for example, specific dashboards to optimize energy consumption, for example, in the plants, but also AI models where we can indeed drive extra insights out, where we can optimize, for example, thickness of coatings, like I was explaining.
And also what we definitely further aim at doing is if we have sensors in our products that we can also really provide that as a service to our customers.
David (05:00)
wow. So you’re providing also, so you’re not only using data for your internal processes, you’re also sharing certain data sets with your customers.
Sophie Van Nevel (05:12)
That is in fact the goal that we have. We have a number of innovative cases where indeed we want to start sharing data with our customers through the sensors.
David (05:22)
One of the things which is extremely important when you work with data both internally and externally is data governance. It’s a term which is really common in the IT world. I always make the comparison that without data governance, you would never be able to run an ERP system, for example, because everybody would just be creating line items and views and maybe you have…
Sophie Van Nevel (05:32)
you.
David (05:52)
100 different ways of sending an invoice or you name it. But if we apply data governance to the operations world, to the sensor world, then suddenly, we start looking to each other and we go like, what does that mean, data governance? So could you elaborate on what does data governance from your point of view mean?
Sophie Van Nevel (05:57)
.
Yeah, it’s a very relevant question because data governance sometimes can seem a bit theoretic, I would say. But what we really try to do is we have a data governance practice being set up, like you mentioned, for the more transactional data. But sensor data is really a different kind of animal. And what we…
really are doing at this point in time is we apply a number of practices that you have in data governance. So being it really defining your data governance roles. So we have data stewards, custodians in place, and also data owners in place, specifically linked to what I was saying on intelligent processes and a number of those processes. And besides that, we also have a certain training program.
for those people to really understand what is their role in the end -to -end data delivery value chain. And that is an important thing, is it’s for us data governance is not kind of something that you put on top of people, but it’s really part of your normal delivery life cycle. And the more you focus also on that governance, the better your results will be being it now if you produce a certain report or an AI model.
or even like I was saying more on the service side, we really believe that by using those practices, you can really drive forward the quality and value that you get from your end product.
David (07:50)
I love to dig a bit deeper into the role of a data steward, a domain steward, a data custodian. Could you just, for example, let’s start with a data steward, for example. What is the, I would say, the short role description, what should he or she do? And also really important, are they part of the, I would say, the operations organization? Are they part of a separate governance organization?
Sophie Van Nevel (07:52)
I’m sorry.
.
David (08:18)
Do they have like more of a hybrid role? I’d love to learn more about their day.
Sophie Van Nevel (08:24)
Yeah, it’s an excellent question. Well, in fact, today for us, for example, if you now take a data steward, they are part of the intermediate team, I would say, between business and IT. So it’s in fact more a business role. And they really look at what are the types of data that we indeed want to measure, what is the mission behind, what are the different sources that we have.
And then the liaison factor with the data custodian is typically a data engineer or even data scientists that they are collaborating together with. But it’s important that indeed you look at the full view because data engineers can look more from a technical perspective to the data and how it is linked in your databases, for example, whilst your data steward is more looking at the overall definition. And then the…
the data owner is really looking at the end result and really looking at the quality of my overall data and where is my data used as well in other domains.
David (09:31)
Okay, that’s interesting. And I can imagine that appointing people to become a data steward, you make them responsible for, I would say, the data around a certain business process or a certain location or whatever. How did that take place? Was that a bottom -up initiative, a top -down initiative? How do you, I would say, overcome people who say like, yeah, but I have so much work to do. This is…
Sophie Van Nevel (10:00)
Thanks.
David (10:01)
This is on top of my normal day -to -day job.
Sophie Van Nevel (10:06)
Yeah, well, that’s exactly the point because you can appoint any data steward and say, today you’re data steward, for example, and you have already a day -to -day job, that will not work. And what we have done, for example, now in that specific case on intelligent processes is that we really tied it up to the entry zones that we wanted to have, being it a light dashboard or an AI model that we wanted to focus on.
mainly that part. And we really, like I said, we tied it up to the end -to -end value delivery chain. And by then saying, okay, these are elements that you need to take into account as a normal step in your way of working. That is the most important. And that is in fact, a part of your role as data stewards. But the most important is that you indeed perform those activities, that you see the value for the end result that you want to get.
And I think in that case we were very successful because also our data owner, who was a product owner for Intelligent Process, was a big believer of the end -to -end value of data. And it’s only in that way that it can work, not just giving people some role on top of their normal day -to -day work.
David (11:27)
Yeah, let’s make this concrete. So you already mentioned energy. You’re working around energy management. I assume that’s to, I would say to achieve certain sustainability targets or certain operational targets. Can you explain that case a bit more into detail?
Sophie Van Nevel (11:31)
Mm -hmm.
.
So in fact we have a program running which we call You Know What, and that is in fact measuring the energy consumption in our plants. We have indeed targets from a sustainability perspective to drive down the energy consumption in our plants. And that’s why we in specific plans started installing in the past two years energy metering.
And based on what comes out of those energy meters, we can indeed see based on different machines and energy consumption that we have in the plants. And our engineers are indeed then comparing different machines and the energy consumption of those machines. And one specific case that we were looking at is the fact that we saw differences between same processes that we are running, but different machines that we’re using.
And that was for us one of the reasons to start looking at how can we use platforms, platform lifetimes here where we’re using, where we’re collaborating together with on driving forward data quality. And what is now really the reason behind the fact that there is differences on those different machines. Is it linked to drift? Is it linked to certain temperature changes?
and is a data that was not measured during a period of time. These are all elements that we took into account.
David (13:24)
It’s always interesting to see that once you, I would say, move from using data only for your day -to -day operations, so what is happening right now, and you move towards reporting, dashboarding, and maybe even a step further towards predictive modeling, that suddenly a gigantic amount of your data appears to be not usable at first sight. So are you…
Sophie Van Nevel (13:35)
.
.
David (13:54)
Now more on the reporting side or are you already taking steps into modeling into more predictive energy consumption?
Sophie Van Nevel (14:03)
Yeah, both. So I think we started indeed first with reporting, because when the meters were installed, first step was just seeing what comes out of the meters and reporting. But based on those reports, we saw indeed that the accuracy of what was measured was not that high. And that’s the reason why we said, OK, we really need to now have a look at what is causing this. But also, how can we do?
predictions on top of those machine data. And that is indeed where we now really started looking at how can we use prediction models on top of our machine data and on top of our energy data.
David (14:48)
Yeah. Well, this is the IT/OT insider. And so one of the things we often talk about is that, yeah, from silos to convergence, yeah, journey, let’s put it this way. How does that look like at Bekaert? What is the, I would say, what is the role of the IT department? When do you step in?
Sophie Van Nevel (14:53)
Hmm.
.
David (15:17)
How does business IT collaboration look like? Is there some kind of a convergence going on or is it more, I would say, teams who collaborate together but they stay in their own organizations? I’d love to understand how working at Bekaert maps to our collaboration patterns we published.
Sophie Van Nevel (15:43)
Yeah, okay. Yeah, valid question. I think what we mainly, specifically in the cases that we have currently running, we have indeed still different teams. So we have still an IT team, we have indeed a business team, I would say, engineering as well, in the loop. And what we mainly focus on is bringing the right people together to solve a specific case. So it’s in fact a bit cross.
teams or cross team collaboration that we do, because we really focus on driving the end to end value. And that is for us the most important. Whether then you come, everybody comes with their different background, I would say, or the different specificities, but you still work together as a team in order to really drive forward the value. And in this case, it was now on how can we optimize on energy consumption.
David (16:39)
Was that, hey, you’re working at a breakout for three years. Well, was that a big change for you coming from the financial world? How is that different? I would love to, I would say, I always think of this big difference between being in an operational environment versus more an office environment or how did it look like for you?
Sophie Van Nevel (16:41)
Okay.
.
Good.
Yeah, well, I typically came from a more service oriented organization, because in banking it’s a more service oriented organization, while in fact, what we have a big cars definitely with our plans becomes much more concrete and you have other elements playing into that whole equation, like the whole OT IT element, you don’t have that at a bank.
But I think what is important and what I also see is that some definitely on the data side, the same issues and the same challenges come back whatever sector that you’re in. The only thing is that I see, for example, at Bekaert that by taking actions really on your data, you can even see more of the impact because it makes it really very concrete. You see a direct impact.
in your plans or in the way you produce, whilst in a service business it’s a bit less tangible, I would say.
David (18:12)
That’s interesting. I always liked the really tangible things we produce in our manufacturing plants, but even in infrastructure plants as well. Because if you’re moving water, for example, or all these things, you can see them, you can touch them. In the chemical industry, I always prefer not to touch them. Probably also something which is…
Sophie Van Nevel (18:17)
. Yep.
Yeah.
David (18:40)
closely linked to that change you made is also the way you deal with change management. So somehow, Bekaert is around for many years, but you’re sitting on these existing assets, existing production lines. Some are maybe decades old, others might be more recent.
Sophie Van Nevel (18:47)
.
.
David (19:06)
But in any case, you’re sitting on, I would say, technology, which is not purely digital, or where maybe digital is even an afterthought. So that means that on the one hand, we somehow need to convert that technology from more old school to digital native. But even more importantly, I would think, are the people who are operating those assets. How do you…
Sophie Van Nevel (19:16)
.
in.
David (19:36)
How do you make them part of your data journey? So is there also something you can share about these experiences?
Sophie Van Nevel (19:45)
Yeah, for sure. Maybe to come back to your first point, in 2030, Bekaert will be 150 years old. And then indeed, how do we take people along in that change management journey? So we have started now, around three years ago, a full data analytics journey. And we have, as part of that journey, we did not focus on all data.
So we really said, okay, we will focus on those critical data linked to a number of core elements of a digital transformation. And what we do in that approach is that every time that we focus on delivering a certain digital product, like for example, in intelligent processes, like I said, but also in sustainability, sort of commercial dashboards, for example, we always focus on taking those people along in the journey to focus on data literacy, for example.
who will benefit the most in the end -to -end delivery of that product. So it’s really again making sure that you embed that change management as part of what you deliver. Because people can really see the direct impact of their actions and working with data and working with AI, for example, that we really take them along into that journey.
And for example, if you talk for example on our plants or if we have the case on energy metering, but we have another case for example on our lubricants that we use as a part of our production process. First time that we showed those insights, operators were saying, yeah, but we know those insights. I don’t need a tool to tell me these insights. I’ve been around for many years and I already know this.
David (21:26)
Yeah. Yeah.
Sophie Van Nevel (21:34)
But then as soon as you start iterating and then if you start seeing, okay, but did you have a look at this or can we speed up the analysis on top of that or can our data scientists speed up a number of those analysis that suddenly you start seeing that they said, I have it. Maybe we’ve never noticed these are this correlation. And then it starts becoming interesting. And then they start believing as well that they can, that this can play an important role.
in optimizing the way that they work. And it’s by doing these type of things, making it very practical and really showing the difference that we try to bring people up to speed and focus also on that change management.
David (22:19)
I can relate to that. I actually forgot whether I shared this previously on the podcast or not, but one of my first jobs as a data scientist 15 years ago. At that point in time, we were already, I would say, using AI models because AI is not something from two years ago. The principles around machine learning are already, I would say, 20 years or more old.
But so we applied an AI model to predict a certain event from, so to predict whether an event would happen or not in a plant. So we had our own name for this model, but the operators nicknamed it the Oracle, because they looked at the model and they didn’t really understand, they didn’t understood what the model was doing.
Sophie Van Nevel (23:11)
.
David (23:11)
Obviously, because we were using a machine learning based approach where we combined several parameters and the model would only say, okay, not okay. So that was actually the only outcome of the model. So, and because the operators, for them, it was an Oracle who just said something. And that was my first, it was a very interesting, I would say learning process to understand how you can apply
Sophie Van Nevel (23:26)
.
.
David (23:41)
these algorithms into the physical world. But it was also a good way for me to understand that just modeling stuff and then throwing it over the fence and then hoping that they would trust it, that didn’t work. So that was also for me personally a really good exercise, so to say. And tightly linked to this is also a bit the difference between
piloting something and then scaling it to a service, you mentioned the word yourselves, a product. Is there a way you deal with, I would say, the scaling question? How do you scale from pilot to product? How do you also define or say like, okay, from this point, we need to scale up.
Or at this point, we just need to stop wasting more money. It didn’t bring what we expected it to bring. So I would say a bit more of a fail fast approach.
Sophie Van Nevel (24:54)
Yeah, we have that effect. So our goal is, of course, always what we do is to further be able to scale it because we don’t want to stick only to that pilot phase. But we have indeed a clear governance in place where we indeed look based on the pilot that we do, what is the value that comes out.
Do we see that applicable, for example, in other plants as well? What are specificities for specific plants we need to take into account if we want to further scale it? Because not every plant is the same or the type of product that we produce is not always the same. And we have indeed a set of criteria defines specifically in the context of manufacturing to determine whether we can scale a certain digital product, I would say. Yeah.
And now specifically, if you look, for example, we have that case on this is our lubricant baths that we use in our production process. This is something that we kind of need for the scale because we have that in, we started with two plants and testing that out in two plants. And this is now something that we can purchase scale across different plants with the learnings that we have done. But we have in at the same time, also a number of cases where we said, okay, maybe it’s not worth.
further scaling or we even stopped a certain initiative because there was no real value in, but that is also fine. It’s part of the whole agile process and failing fast and see, okay, where is the most value coming out.
David (26:29)
Yeah, I fully agree. I think stopping initiative is equally important as scaling other initiatives. And I have to say, especially in manufacturing and I would say, typically also within the more traditional way of how we engineer plants. So we think about something, we engineer it, we build it, and then we operate it. And I see a lot of people who are used to this kind of thinking, they also apply
Sophie Van Nevel (26:34)
Yes.
David (26:58)
this way of thinking to digital projects. And this is where I believe that bringing these insights from IT and combining them with, I would say the experience from the business from operations, these two together that creates firework, I would say.
Sophie Van Nevel (27:17)
No, then it’s a…
David (27:19)
One of the topics you already mentioned that in making scaling possible is the importance of data quality. Now, when we work with data, data is something intangible. The production line, when you are pulling steel wire and you’re making it thinner, that’s something we can see. The data which comes out of those processes is rather intangible. But then,
Sophie Van Nevel (27:29)
roof.
David (27:50)
I would say quantifying and working around data quality is even more intangible, I dare to say. So how did you start that data quality journey? I assume you can’t just do a big bang approach and say like, now we’re going to fix everything.
Sophie Van Nevel (28:08)
Indeed, no, if that was possible, you would have done it but that’s not what we started doing. Indeed. No, I think specifically then like for sensor data quality, we really started from a number of pain points that we saw. And we really said, okay, we will focus on a few cases where we see both that there is a high business value to it. Like for example, the one on energy has of course a big impact on
not only efficiency in the plants, but also the whole sustainability footprint. So it was a very nice case from that perspective to tackle. And we saw or we experienced that there were issues in the data quality. So that was a very visible one to get started with. The second one indeed was the one on lubricants, where we also were based on interactions with our plant engineers, that there were indeed a number of parameters.
still not working out. And there was also something where we said, okay, maybe there’s an interesting case to start looking at data quality. And it’s really by looking at those practical cases where we see, okay, there is something off and tying it to an end to end strategic value driver, because that’s as important as the first part, that
If we focus on these cases, then we really believe that we can drive value forward. And these are also now the ones that we’re looking at next steps to take where there is indeed a link to strategic value and where we see, okay, there are things that we can further improve or even drive value forwards.
David (29:50)
Maybe to wrap up, technology is changing fast, especially in the AI world. What do you think we will see happening in the two, three, four, five years to come? What do you hope to achieve in the coming years at Bekaert?
Sophie Van Nevel (30:14)
These are two questions in one. What I hope to achieve within Bekaert is that we can definitely still drive forward a lot of value also based on AI and data because AI is now indeed a big hype or even things that are already above the hype or beyond the hype. But I really hope that we can drive forward a lot of business value through the use of AI or combination AI.
information, digital technologies in essence. And I also believe that data quality will even become much more important because we will use everybody wants AI systems these days and there should always be somewhere an AI component or an AI source in it. But I really believe that if we then see the results of using those AI models, data quality and data foundations.
will remain extra important or will even become more important. And then you talk typically about data governance, which we already talked about, being aware of data quality practices. And that’s the reason why I think we already invested the past years into those foundations and we need to keep investing in those. And not only saying, yeah, we focus on AI and we have nice AI cases.
But those foundations are as important as the nice cases that we can build on top of them. So that is really what I think that is going to happen. And why I think it’s important that we still focus a lot on those foundations as well.
David (31:51)
focus on the foundation and if the foundation is okay and is scalable, then the value will follow.
Sophie Van Nevel (31:52)
Yeah.
That’s what I think too. Yeah.
David (32:05)
Sophie, thank you very much for joining me and to our listeners. Thank you for tuning in and until we meet again. Thank you very much.