Your digital transformation is only as good as your data: tips on avoiding ‘shiny object syndrome.’

Guests: Allen Turner, Advanced Analytics Team Lead, International Paper
Co-host: Mariana Sandin, Industry Principal, AVEVA.


We speak with Allen Turner of International Paper (IP) about the unique challenges of digital transformation in a legacy industry like pulp and paper. Mr. Turner, as part of International Paper’s global technology team, is working on an initiative called “mill of the future,” which focuses on cost savings and optimization across International Paper’s mills. Mr. Turner lays out the importance of establishing foundational data quality—or “cleaning up the messy basement” of old data tags and stale data tags—in order to ensure a successful digital transformation project.

Subscribe to our biweekly newsletter. Visit our website. Follow us on LinkedIn.

Transcript

0:08

Welcome to Radio PI, everybody. I'm Nick D'Orazio from OSIsoft, which is now part of AVEVA. Today, our topic is shiny object syndrome and how to avoid its pitfalls. Or to put it another way, your digital twin is only as good as the data behind it. Our guest today has a lot of experience rooting out those pitfalls that come with projects like digital twins and digital transformations. So let me introduce Alan Turner. He's the Advanced Analytics Team Lead at International Paper. Welcome, Alan. Thank you very much happy to be here. And also our co-host today is Mariana Sandin. She is the Industry Principal for forest and paper products here at AVEVA. Hello, Mariana. Hey, Nick. Hey, Alan. How are you? Good. Great. Thank you so much for joining us, both of you. So International Paper, it's one of the world's leading producers of fiber-based packaging and pulp and paper with 48,000 employees, more than 25,000 customers in 150 country countries. So what do you do at IP? First of all, the best way to look at this is if we focus on the manufacturing side of International Paper. So that's where I sit. So this is the facilities, the technology group, the infrastructure that supports taking woodchips to paper or recycled to paper. And so I'm a supporting role in that operation. I'm not located at a mill site. But I am part of a global technology team that helps support the mills in terms of cost savings and optimizing. My specific role is in an initiative called mill of the future. So, this team that we've formed in the Atlanta tech square area is focused solely on mill of the future initiatives. And this office that I'm broadcasting from today is our innovation center here. And it's called the advanced analytics center. And so that is my role is to ultimately build this team, this office, and support our mills for cost savings measures associated with data and analytics. So, Alan, where did you see the opportunities to fix things were? What was the problem that you decided to address? I don't know, I don't know if it was a problem, per se. But when you're coming into a middle of future initiative, and you start looking at what we want to do, you try to figure out what's going to set you up for success. And so, if we really quickly, and I'm sure this is the same in every manufacturing company, these days, if you go and you look at the raw data level, and you realize the kind of projects you're going to attach to that data, it's pretty quick and easy to see that they don't line up. So the data and the structures to develop the data may have been around for decades. And when it was put in, and it was not put in in such a way that it would proceed I'm gonna attach artificial intelligence. They might not have been able to spell AI, you know, in the 1980s, for example. And so, there's a disconnect. So if, if we want to do those initiatives that require this data, we needed to focus on this data. And I guess, one of the early ways I envisioned it to, you know, my team and the people that I was interviewing with, in terms of IP talking to them and figuring out where things existed is, I think we could kind of close our eyes and hold our hands out and we could bump into opportunity. That's how rich the opportunity is to improve our foundational data is. If I understand, I mean, it's been measured. It's in the billions. Is it fair to say it's in the billions the opportunity for doing things digitally? And is that an IP number? I wouldn't say that it's specific an IP number. I know that in terms of our mill of the future, when we look at productivity, when we look at cost savings, when we look at even environmental and safety. If you look at all of those ways that coming up with a digital strategy effects, it is definitely a seven-figure number, probably over a billion, and we're probably not alone. I would think other industries would see similar numbers. It's, um, it's actually so big, it doesn't matter how much math you put into it. It's pretty significant. Okay, so we hear the challenge, trying to do advanced things with old systems. What was your response to that? What do you see is the solution? There's no real magic solution, unfortunately. And I guess the key challenge is to convince the stakeholders that want to accomplish higher technology that we have to go back and fix the foundation, or clean up the messy basement is what I always like to say. You know, many cases are archives. Our data historians are a mess from old tags, stale tags, data that no one really knows why this PI Point or something was established. There's so much of that there. We have to go and either put a blanket over top and start afresh, or we have to clean it up. So that that would my answer to that question. Okay. All right. And I just want to add one thing to that. So, we need to remember that in the industrial world, we come, we come to a point where you had a plant that was probably acquired from another company, right, that had different standards, different kinds of technology, or that has been growing or adding capacity at different points in time in history. Right. So there are some paper mills out there have been running for the last 120 years. And so back then it was really hard to envision what was going to be possible today and all these shiny objects. I like how you said, Nick, digital transformations in plural, because sometimes it feels like that, that there is not just one but several that you can go after. And that's that's very hard to plan. It was very hard to plan. Right. So when we look at the physical assets, I think that the messiness or theon the structure of the data of the operational data that we see today is a reflection of that as well. So true. And those people who were familiar with the PI System, they know the data compression is an important part of data integrity, that can affect this as well. Right? That's what you noticed, right? I mean, absolutely, it's, it's things we wouldn't ordinarily really think about on a normal basis, you set it and forget it, you know, back when you set the tag, you set it, but now, we're trying to attach things to the data that might like to see more of it. So less compression. And at the time, maybe compression was set up, because it was costly for data storage, where now it's extremely cheap and effective. And so you know, convincing stakeholders that okay, to put in a digital twin, perhaps we need to run a tool or spend some time checking, compression or tag naming conventions. And I really like what you were saying. There's the fact that it's almost like, you've bought a house, and you're doing something but when it was built, it's not to anywhere near today's building codes. And so that's, that's kind of where it is, you built the house that was perfectly fine with your data, structure, set, your naming convention, set, everybody at the mill knew, and then all of a sudden, you're either acquired or change, change companies, divest some part of your mill, and now you have a different building code. And you have to match those up somehow. And that's many cases in this day and age with the number of mill that have changed ownerships in the past decade alone. It's, it's a real challenge, but one we have to get through, you know, you can't be successful until we get everything on the same basis, really. So data remediation, I guess, is what you'd call that. Okay. And that in the cleanup that you're going to do there? I mean, specifically, for folks who do have the PI System, there's one tool that you use for that, right? Can you describe that? Well, there's many tools. But probably the biggest piece, and this was, again, when we started out when you mentioned that question earlier about what did I see initially. When I came in, and a colleague of mine, Rick Smith, many people probably know him from PI World and his contributions there. But he showed me this tool called Asset Framework. And you know, I'm coming in from the operational site, an engineer not really that familiar with all the nuts and bolts of the OSIsoft platform. And it took about five minutes for me to realize this, this is exactly if I was going to build something as an engineer. And I've got, you know, scattered tools on the floor. I'm going to structure them in some way that I can find them, rename them, and so forth. And Asset Framework just jumped out as, you know, a perfect fit for the kind of mess we're trying to clean up. And so for sure, I mean, if we call the PI database or archive our heartbeat, Asset Framework is going to be right around that and in terms of giving it rigidity. Okay, if I understand, sometimes, you know, a shiny object project with a lot of impetus behind it. I mean, that can actually be a detriment sometimes. Can you describe that? Absolutely. So, think about attaching, you know, artificial intelligence to a platform where you know the data is is poor, missing, an error. I think, and you would see this immediately. Say you went for a walk around and said and walked up to an operator and a paper machine area, and you said, You know what, we're going to implement artificial intelligence to help you run the mill. And, and they, maybe they snicker, maybe they don't, but they start talking about some of the data integrity issues that they have. And so the challenge is, yes, we want to, you know, push the limits, we want to try new things, we want to come up with cost savings, that has no capital costs. So that's the lovely thing about AI and digital twins. I mean, they're basically man hours and software, but we're attaching it to something that's not prepared for it. And so if you've got a really, really eager, senior leadership team or something that wants to try something, and then we try something, but we're not on a platform that will support it, then there's a potential that we will fail. And then it becomes what happens then. Does the group understand the reason? Or is it automatically said, well, AI will never work in pulp and paper. And so that's really the challenge, especially in a legacy industry, like pulp and paper where, you know, mills may not be set up for the mill of the future. You know, they have a lot of possibly a lot of work to get there. And so the challenge is twofold. One is, how do you prepare yourself potential failure? But then you really need to be specific as to where you try certain things. Because of the data, the foundational data quality in certain mills. You know, it's very interesting what what you are saying because twofold: so, one, is how well prepared the organization is to adopt this new technology, especially AI, and, two, how good is that data that is going to feed this new technology in? Because of my role, I work with other customers around the world. And I do see that there are some they're really eager to get it done. Get it right. But they're just putting in automation. And like, well, you'll get there, right? It's good that you have that you'll get there. You're not quite there yet. And then we have customers, seasoned customers in data technology, like International Paper, that the organization is it's changing. It's a cultural change. And but then there, there is some groundwork that needs to be done for these things to work. So yeah, it's it's not only technology, definitely a cultural change needs to happen. And that's why you had to be very selective where you try proof of concept projects. You can't just pick one and go with it. You have to be really have a lot of forethought. Okay, there are some things that you were able to do to kind of get more buy-in this time around. Can you describe that process? Well, you know, I, I read a lot. So one of the books that I read a couple years ago, was really, really eye-opening in one part they were talking about a Lockheed had the Skunk Works division set up, right? And so you know, Kelly Johnson, leading that at this monitor, said, show me. Right? So instead of arguing, well, if something was worth doing or not. Okay, Show me. Prove it to me. You know? And so I had that same kind of mantra. So let's, let's do something. Let's show people what could be done. And in another case, in a number of cases, that's what we've done with this team. Thats, thankfully, strategically focused on these type of initiatives. So we've got a group that can build some prototypes, and show a subject matter expert or a mill engineer, what could be done. And, you know, kind of like Wizard of Oz. Don't look behind the curtain because there's duct tape and tinkertoys and and string. But on the outside, it shows them what could be done. And I think this is going to happen once you can show someone something tangible that they could use and see how useful it is, then you start to gain and engagement from the people that might be avoiding this whole transformation in the first place. So what were you able to show them this time? You know, initially, that got so much attention. Probably the best example is, really, just over a year ago. Our team was traveling with a group of subject matter experts to audit a mill. And before we left, we talked, talked about, you know, keep your eyes and ears open for pain points. Look for things that these subject matter experts are doing routinely. That may not be adding value, that is just grunt work, because that's a source of automation. And so we just happened to be along at one of those audits, and kind of saddled up to one of the subject matter experts working on an area process report. And so this report was done every 30 days. And my guess is a lot of people listening to this will have, this'll probably really touched our hearts a little bit. But so this subject matter expert every month was responsible for issuing an enterprise audit report, by email, several pages long, Tables, and word. And it was taking one area of the mill and evaluating each one of our mills. And so this was all done manually, pretty much for the most part. So multiple PI Servers, they would have to connect to, 40 plus spreadsheets for each mill, each line, each species of pulp, connecting to those multiple PI Servers, that was maybe 16 PI Servers. And so all of this was done on, say, the first day of a month for the previous 30. And then after doing all of that sorting, cleaning, filtering, and doing tables in Excel, there's a copy and paste into a Word document that's turned in to a PDF that's copied to an email. That is distributed. And so, to me, this was ripe for automation. And so that that's really the business case that we selected. Okay, so there's two things specifically that I want to talk about. Two initiatives that I wanted to talk about that you would put in the head a lot of success. One was process health diagnostics, the other was advanced pattern recognition. Let's start about the first one. Can you just describe that? What was your process health diagnostics initiative? Yeah, I mean, it fits perfectly with the example I just gave. So, you know, if we really want to give some kind of evaluation of how a process in one of our mills is behaving, you need to audit, right? And subject matter experts are paid to do this. And so if we could take this routine of manual work that we just outlined earlier with the spreadsheets and the PI Server, and word document, and copy into an email, and so forth, if we were able to utilize the Asset Framework. So, if we go back to what we said earlier, and we've got this flat database of PI tags at every one of our mills, and each mill has a different name, and possibly a different unit. If we use Asset Framework to add that structure, where we can get it in this the right naming convention, make sure all the mills have the same units, do the calculations that the subject matter expert was doing on our laptop, for example, in Asset Framework in the same way that she approves, and write that data into PI from there. That allows us one to standardize, to remove the data transfer into Excel, which was taking quite a bit of time. And then the whole data conditioning can be done an Asset Framework or possibly some attached. So, with that one, if you really look at just these foundational tools that we own, as an Enterprise Agreement. We have, we own these tools. We can use these tools. And so, if we're thinking about that we're using the PI database. We're using Asset Framework for our structure. And then we're using the PI Integrator for Business Analytics. And from there, we can utilize those calculations, that data, either in our own report or send it to a third party, such as AI or a digital twin. And so, for me, that was a perfect case. And we just kind of wrapped up that prototype, that show me stage, where we said, Okay, we're going to, we're going to show you how we can make this monthly area report in Power BI that updates automatically every week. So now, the mills, instead of waiting for the end of the month, they can now go in and see kind of how their performance is doing ahead of time, which is where we want to be. We don't want to wait 30 days to make a change. Okay. And there's another initiative that is often advanced pattern recognition and third parties that are grabbing your data, your vibration data, and whatnot. Can you explain that? Well, I mean, we can talk even general, I think, you know. What I keep having to remind myself and others is that if you look five to 10 years from now, where we are going as an industry, we are going to have to, whether we like it or not, more wireless sensors added to our facilities than not and so we have to be prepared for. What are we going to do with that data and, you know, vendors in that realm, a lot of them like to sell the sensors, but the the game is for the service. So, taking that data from the sensor, promoting that it's wireless, and maybe even has a cell connection. And we're going to take that data in the cloud and give you insight back. And that's a great short-term concept. Long-term, as an owner of the data for the asset, I would like to have that data, the raw data, even not the condition data. And so the challenge is, how do you get that data into PI, because it's a totally different route for the data. And so and that's what we've seen a number of initiatives. One is the one you mentioned, where we have, say, a vibration signal. And then it gets conditioned and gets manipulated, analyzed, and put back in some way, shape, or form. And we have to require, as a user, if we really want to get big data, we have to require that be put back into our data lake or data warehouse, which we call PI. And so that's really a challenge. And one, you know, people may underestimate how you get that data back, how you keep the time-series nature to it. That it's the right time stamp from when it was taken and conditioned and put back. How do you One of the issues we're dealing with now is, how do you deal with bad quality? You know, how do we set the attribute in the PI tag that it's good or not? Some vendors do it. Others may not. And so it's just opened our eyes that this is an area where we're going to have to develop some standards as a company on how we deal with this, because it's, it's only become, it's only going to become more popular to have these type of services and sensors. And the end result of all this is, there's some things that you can bring to the attention of system, these subject matter experts, You know, you're using some AI software. But but it, it's not going to do you any good unless you can bring that, you know, close the loop and bring it back into PI so that people can see these recommendations that are coming from these advanced systems. Is that, is that the end result of all this? I guess the way I think of it is, at this point, we can't, I don't think we can handle the potential loss of value by saying that data adds no value, or what another way to put it. At some point, you could have said years ago that the cost of storing that data outweighs its usefulness. At this point, the cost is so minimal, and you don't know the future potential of a data point. I'm not sure you can really adequately decide to throw away data. And so, for sure, every bit of data we can get, we want to keep. And so we have to get it back in some way, shape, or form. If it leaves, goes to a cloud, we either need it before it goes or need it when it comes back. And because it's time-sensitive data that I'm talking about in the manufacturing world, it's got to be the proper timestamp. Because all of these AI algorithms and statistical formulas, they'll take all of that data and decide and tell you which one's important or not. We as humans, I think will have a challenge determining that. But you don't know what insight you get until you combine SAP data and operating data and prophecy data. And the weather channel's data. And you know, the local sports, you don't know what you're going to get until you combine them. But if you decide upfront, you don't need it, then you'll never know. You're mentioning something that I think it's very important: how the subject matter experts today are able to, you know, get all these sources of information. And because of their experience, they are able to take some action, some informed decisions, because they've been in the industry 25-35 years, maybe more. And in the fact that we're using today, digital technologies to create data models that can give us some of those recommendations. What I've seen, it's because this expertise is starting to go out from the mills due to retirement. Sometimes there is not the same pool of talent to come into the mill. That's the other side of the coin too. So do you think It's not only that it is possible technologies. It's possible today. But do you think that's also another factor? Why we are going this route? I think it's a challenge from a change management perspective. I think we'd be crazy if we didn't think, Okay, this technology I know it works, because it's just algorithms and computer software and stuff, it, it's not emotional. You know, if we give it the data, it is going to give an answer. Its the chain, the people side that it's going to be, in many cases, whether something is successful or not. So I mean, perfect example, are our mills, and probably every industry is the same. But our mills are filled with 30-year International Paper employees. And they don't need data analytics. They don't need advanced sensors. You know, I don't know how many meetings I've been in, where we've tried to communicate an initiative of what we're going to do here from Atlanta. I'll come up to after meeting, you know, this is really, really good stuff. But we tried this in the 80s. It didn't work. And to be quite honest, I don't really need all that. You know, and so, but that's precisely the challenge. You have to say, in a nice way, You know, I get it, you know, you've been a long employee. You're great. You're an A-plus contributor. But it's not about you, you know, this whole initiative is not about you, the 30-year employee. It's about the person behind you, coming in that young, because you're gonna retire. So and that's truly one of the biggest insights from doing digital transformation is setting up the younger generations because of that, that experience gap. Right. And so it has to be done because we want ultimately, I guess, the tagline here in Atlanta is, you know, increasing the speed of transforming data to insight, right. So the younger generations have this advantage that they've got things at their fingertips that can take massive, massive amounts of data, and get to insight pretty quickly. And so, yeah, I could have used a 30-year experienced person to tell me which pump to work on. But what this software package, I can get them down to five pumps and have a series of 20,000 to go work on. And that's a much better odds to get a newly hired engineer down to five. And then they can use their training and talent to pick which one of the five is the right one. And so it's an interesting challenge we face and a really good time for computing power to be as cheap as it is. Okay, it's fascinating story, because what you're talking about is basically get the foundations right. And let's focus on the shiny objects. And but it's a good it's a great lesson, I think, for everybody. Anything. Yeah. Yeah, I would I mean, I think this was in my presentation from last year. But I love Michael Jordan's quote, where he says, you know, get the fundamentals down, and the level of everything else you do will rise. And that's exactly what you're saying. I mean, if we can, and maybe it's AI and digital twin. That gives us the motivation. It gives us the relational currency to work on these foundational tools that every person in IT knows we should have been doing 10 years ago. But now we can say, Okay, we're going to do AI, but let's work on data compression and data quality and naming conventions. And I think we're seeing that. We're able to get some of these projects that people in IT have said, We were trying to do that 10 years ago. But now it has value. And now it has a return because we want to try these other things. And so that's what I would leave with. It doesn't sound very exciting to say I'm working on compression, or PI tag naming, but I the results of doing it will pay off for sure. So I can't say it any simpler than that. Right. And we don't have the numbers of how much it actually pays off from, from other stories from other customers and the industry out there. There was one presentation last year as well that they said by having the right naming convention, we were able to decrease in 50% the amount of time that we actually spend building a new plant, and this is for a specialty cellulose product. And you know, and with stories like that It's when it becomes relevant or becomes tangible, how the foundational work is so important. Well, thank you so much. Thank you so much for taking the time. Hey, as we as we normally

like to do at these: 29:53

I got a quick set of lightning round questions for you. First, one what's the What's the most advanced thing, either packaging or paper or whatever, in your opinion, that IP creates? Most advanced thing? Yeah, I mean, for example, when I used to work at DuPont, we used to make this lubricating oil. I t would go on to like satellites and stuff. It was $5,000 for five-gallon drum. So, it was like, Wow, that's pretty cool stuff. Right? What about IP, anything comes in I think it's advanced, in my opinion. But, you know, if you're into the tech world, this is not gonna be that advanced. But when we were putting together this office, I just toured our design studio for container board. And Memphis is a consumer product-facing group. And I was just blown away that this product we're making in some of our mills, how much design goes into making boxes, and so forth. And so, I walked through there and got to see the CAD and a 3D design and laser cut boards just to make like an avocado box that you might see in a Costco or a Sam's. There's a tremendous amount of engineering in there. So the next time you walk by, and you see those boxes that have the stacking capability, like pallets and handles, and they're rugged, and you know, you could probably pull your car up on top of those. That is pretty advanced in my mind, when you think about it. It just comes from wood fiber. Cool. Now, I forgot to ask you, are you an engineer by degree and training? I am chemical engineer with a pulp and paper degree actually. And you've had management roles? What would your advice be to people who are in engineering who want to go into management? Hmm, I would think I would just say, the biggest challenge, or the thing to keep in mind is that you might have to do less than think more. And so, a lot of engineers have something they're really, really, really good at. And so, once you get into that leadership role, you may have to do a lot of soul-searching to get rid of that thing that you're really, really good at. Because when we're stressed, we tend to focus on the things we're good at, and not the things we should be doing. Okay. Well, since you are an engineer, do you have a memento like a broken piece of gear or fried motherboard or something that you keep in your office just for just for old times sake? I've got the old compact version of the PalmPilot. I bet that thing would actually boot up. I'm trying to think I would think that's probably it. I like to keep my old Nokia phones, because I remember that, and actually, probably my most treasured thing was when I got the BlackBerry and could type emails on that little keyboard. So, I keep those, and you know, one day, maybe they'll make it. You know, I don't know if you're familiar with Cracker Barrel country stores, but I kind of joke with people about, you know, things that are old and, I mean, one day, there'll be a Blackberry hanging from the old country store. There's probably one already. Okay, so are you are you downtown, or are you were at a production site? Where are you when it's not during a pandemic? Yes, I'm broadcasting from really our brand-new innovation center in tech square in Atlanta. So, we're right on the east side of Georgia Tech's campus. Okay, so I'm, I'm not at a manufacturing facility. But one of the nice things about being here is within a two to three-to-four-hour drive, we can be at a number of our facilities. Okay, so I got to ask you this, because this is something nobody outside of operations has an appreciation for. In all the operating areas you've been in, what's the coolest view you've ever seen? Because I remember I, I was at a sodium plant, once, and it looked like something at a Dante's Inferno. It was incredible. What about you? I would say probably the most breathtaking view I've ever had, if you could say that with pulp and paper. This is an interesting discussion. But in in Brazil, I went to at the time, it was the world's largest single-line production facility. And so, I went to the top of the continuous digester in that facility. And the continuous digesters are normally one of the taller vessels when you look at a pulp mill. And the interesting thing about that mill in Brazil, is that when you go to the top of the digester, not only is it alone, so you're on a very tall, it almost looks like a rocket when you when you drive up to the mill, so you're at the very top, but in Brazil, when you look around because these facilities are normally enclosed with clone trees, you pretty much can see all of the mills feedstock in a circle around the mill. And you could see the eucalyptus trees and various stages of, of growth. And to me that was that was pretty cool. Interesting to see. That's very cool. That's very cool. Okay, cool. So one last question. So you're invited to speak to a school to schoolkids about career day. What do you tell them that you do for a living? First of all, with, with the generation that we are now they want to know about the environment, I think. And so, the first thing I would tell even when I'm recruiting younger engineers from Georgia Tech, or NC State or other campuses, you know, the fact that this industry has and always will be a renewable industry. We just did a lousy job of marketing it back in the 70s, 80s, and 90s. But we're a renewable product very balanced in terms of our chemical use. We drive a lot of energy from our plants. So it is we are sending green power to the grid. And you know, it's a renewable product, everything we make can be recycled. We're one of the biggest recycle consumers and producers in the world. So it just, to me, when you tell them that, and you speak to someone who likes to solve problems, which is an engineer by trade. It just seems to fit really well. That's cool. Okay, good. Well, thank you so much, Alan, for presenting today. We've been talking to Alan Turner, Advanced Analytics Team Lead at International Paper. Thank you so much, Alan. My pleasure. It was great to be here. And Mariana, thank you so much for joining us. Thank you, Nick. Thanks, Alan. Great talking to you. So thank you again, everybody, for joining us, and we will see you in a couple of weeks. Bye bye.