Manufacturing lends itself to data collection. The problem is the sheer volume from onboard sensors to materials management and everything in between can be massive. But the benefits of getting it right are enormous, for individual companies, and entire industries. In this episode of Connected & Ready, host Gemma Milne is joined by Jon Sobel, co-founder and CEO of Sight Machine, a data analytics and insights firm focused on manufacturing, to talk about some of the hard lessons manufacturing companies have learned trying to realize the potential of digital data from the physical world, the importance of culture when it comes to getting the most from your data, and how individuals can help their companies take practical steps toward unlocking the value and business impact of their data. Microsoft Dynamics 365 Supply Chain Management helps businesses build agile, connected, and resilient supply chains to effectively meet changing customer demand and ensure business continuity. Using predictive insights powered by AI and IoT, Dynamics 365 helps streamline operations to maximize efficiency, product quality, and profitability. Request a live demo today: https://aka.ms/AA8l720 Thank you for listening to Connected & Ready! Do you have ideas of how we can improve the show? Want to recommend a guest for us to interview? We value your partnership and participation. Please drop us a note at firstname.lastname@example.org. We would love to hear from you.
Gemma Milne talks with Jon Sobel, co-founder and CEO of Sight Machine, about the world of big data in the sector that produces more data than any other: manufacturing. From the promise of Industry 4.0, to working with “dirty” data and the importance of a strong data foundation, the surprising lessons of working with digital data from the physical world, to understanding digital twins and control towers, Jon’s examples of real-world business impacts will help you understand why manufacturing data is so unique, and potentially valuable.
About Jon Sobel:
Jon Sobel is one of the founders, and CEO, of Sight Machine. Sight Machine has been helping manufacturers understand the data from their plants so they can unlock new opportunities, as well as create real business impact. Sight Machine creates a standardized data foundation to provide real-time analysis for manufacturing in every field.
Topics of discussion
Using predictive insights powered by AI and IoT, Microsoft Dynamics 365 Supply Chain Management streamlines operations to maximize efficiency, product quality, and profitability. Request a live demo today:
Follow us on social media
Gemma [00:00:05] Hello and welcome. You're listening to Connected and Ready an ongoing conversation about innovation, resilience, and our capacity to succeed brought to you by Microsoft. I'm Gemma Milne. I'm a technology journalist and author and I'm going to be exploring trends around how companies are adapting to a disrupted world and preparing for tomorrow. We're going to speak to the innovators who are bringing products, operations, and people together in new ways. In today's episode I'm chatting with Jon Sobel, co-founder and CEO of Sight Machine, to chat all things about the power and potential of big data in manufacturing. We explore the critical role of building a strong data foundation with reliable data. We talk about how to give people confidence and using data to unlock insights as well as opportunity. And we look at ultimately how making sense of plant data can empower organizational change. Along the way, as always, we cover tips and advice to make it all a reality. Before we start, I want to thank all of you listeners out there. If you have a topic or a person you'd love to hear on the show, please send us an email at email@example.com. We're so thankful for you all. Now on the episode. Jon, thank you so much for coming and joining us on the show today. Let's start with some introductions. What do you do and what have you been focusing on of late?
Jon [00:01:25] Delighted to be here with you today, Gemma. My name is Jon Sobel. I'm one of the founders of a company called Sight Machine. Sight Machine's about a decade old. We work with factory data. And I'm also the CEO of our company.
Gemma [00:01:38] We've been hearing a lot about big data over the last well, it seems like quite a long time now but the past couple of years. What was it that made you realize that there was a potential to use the technology around big data for industry, for manufacturing and factories?
Jon [00:01:52] So big data is a really interesting term if we think about the last twenty, twenty-five years in technology. As many of our listeners know, big data is often said to mean the three v's: volume, velocity, variety. If we go back to the history of big data, it began in the consumer Internet era when massive websites started to deal with a huge amount of data and then they started to want to calculate it very quickly. Those were really the first two chapters of big data. The founding team of our company is a bunch of Web heads. So we went through the first two chapters on the Internet side and we didn't think of it this way at the time but looking back, I now believe that the current chapter we're in with big data is really about variety. We've been working at large Internet companies and working with data problems. Five of us got together. Four of the five of us have grown up working in manufacturing and we all were a little bored with ADTECH and thinking about where would be the next area of opportunity. Physical world data is massive. It's valuable whether it's health care, manufacturing, transportation, energy. There's all these interesting fields with a ton of data that are very difficult to use because of the variety problem. So about 10 years ago, we got very interested in manufacturing and we were looking for areas of opportunity where the technology hadn’t been applied yet. And so that's where we started.
Gemma [00:03:17] And what was it that I guess you were setting out to do for the manufacturers? For the factories, of course, you'd seen that there was this gap. Right? But what was the I guess the maybe the end goal or the benefit for manufacturing and their factories that you were hoping for?
Jon [00:03:30] So all of us are from the Midwest. And our companies started a stone's throw from Detroit. And we were all feeling that we wanted to use technology to do something that might be part of helping to rebuild parts of the country and activities that at that point had not yet really participated in what technology was doing. Detroit went bankrupt a year or two after we started. I remember the first couple of plants we worked in were in Detroit. And there was this great feeling of this is really cool technology. We can also maybe one day be part of rebuilding things and doing what American software companies are very good at on a global scale.
Gemma [00:04:15] Yeah, really bringing a home to those communities that I guess really need it and hearing, I guess, the expertise in the hardware, shall we say, with that software. Maybe you can elaborate that with an example, maybe walk us through how a [unintelligible] manufacturing looks like in practice, how data is collected, made actionable, and the end results that can have.
Jon [00:04:34] You bet. We've learned a lot of lessons in this decade of doing this. You know, a lot of this was just trial and error and going to plants and trying to figure things out. We spent a couple of years just going to plants and asking them where is the pain? So one thing that might not be expected is there's already a ton of data out there. There's more data in manufacturing plants than any other source in the world. Twice as much data gets produced in plants each year as any other category of activity. So there's a massive amount of data already. At the same time that I'm describing our work beginning, there's really been a global focus on manufacturing in the last decade. Germany coined the phrase Industry 4.0, and so a lot of companies are focusing on digital transformation. The Industry 4.0 they asked the question you just asked, what's the journey? How do I do this? Pretty simple steps conceptually. First, you've got to have data. Check. Plenty of data in plants. Then you've got to get it out of where it is and into a place where you can work with it. Because if we do that in IT, it goes very quickly. Everybody assumed the same with manufacturing. They figured if we just get the data out, we're on our way. But it turns out that even when it's aggregated because of the variety problem, it's still quite difficult to work with. So the steps are get it out. There is a paradigm in data technology called stream processing, which basically means real time processing of data instead of batch processing. Most work with data historically has been batch processing. There's more and more real-time processing. To do anything really immediate in a factory to make a decision while it's operating you need real-time stream processing. So the steps are get the data, get connected to it, get it flowing. Then transform it, put it into shape to be worked on. Because of the variety problem, what companies found is if you just dump a bunch of data into a data lake and then try to work with it, it's too unwieldy. And that was a real wake up call, a real surprise. I think a lot of companies spent a lot of money and time thinking, hey, if I just get this in a lake, I'm good to go. That works with everything else, doesn't work with physical world data. So connect, transform and then analyze. And here's where all the stuff that everybody's talking about with AI and all the math tricks that can be applied to data now become very relevant. Once you get the data in shape, you can do all kinds of cool things. And so, to get very practical, believe it or not, most manufacturers lack visibility. They kind of know how much they're making, but these are such complicated operations, they can't really see into the guts of what's happening. So if they can just see how everything's operating and what's working, what isn't, that's a big move. The next step is to start getting kind of analytical, diagnostic, finding causes of problems, finding patterns. And then because of all the algorithms and techniques that are available for working with data now, you can start to get very sophisticated and optimize operations and set up a recipe every day for how you run your plant. You can work around how much energy you want to use. You can start to predict and do all kinds of interesting things that everybody talks about in this field.
Ad [00:07:44] Microsoft Dynamics 365 Supply Chain Management helps businesses build agile, connected, and resilient supply chains to effectively meet changing customer demand and ensure business continuity during times of disruption. Using predictive insights powered by AI and IoT, Dynamics 365 helps streamline operations to maximize efficiency, product quality, and profitability. Request a live demo today by following the link in the episode description.
Gemma [00:08:16] So you've mentioned this point about the realization of the data lake problem and a few other points here about variation in data and just being able to understand what's there and so on and so forth. But considering that you've had this experience over all these years, what would you say are the biggest challenges or hurdles that you've come across when trying to implement these sorts of data switches and big changes within the manufacturing sector? I mean, I'm thinking reliability of data, sheer just understanding of data or maybe is it something broader, like people not feeling confident and making sense of data and even just these new processes.
Jon [00:08:50] So exactly as your question suggests, there's technical challenges which have been fairly daunting for the industry and then there are organizational or cultural challenges as well. I'll give you examples above. We joke around here. We say manufacturing data, this is a very American metaphor, is the NFL of data. And what we mean by that is it is a tough, tough area to work in because you have problems like out of order data. So if you're processing data in real time, the data is actually coming from a whole bunch of different systems that are all on different rhythms. You might get quality data two days after production has been completed for something, but you've actually got to join that quality data that came in two days later with the original data. for any of this to make sense. And to give you a sense of the complexity of working with data in the manufacturing environment, typical piece of automation in an assembly line will have five or six hundred sensors on it. You pretty soon got tens of thousands of data points in each plant. And if you want to understand an enterprise, you've got hundreds of thousands or millions of data points, all on different clocks, all on different formats, some coming in late. Sometimes it's missing. It's often dirty, it's corrupted. So you've still got to make sense of all this while all of this is happening. That coordination problem around the data, people call it orchestration or harmonization is just a nightmare because none of these systems were built with the idea that they'd be analyzed. When we work with virtual world data, it's all coming from computers and networks and stuff that for decades has been meant to work together. So wrestling that data in a sense - big technical problem that a lot of people didn't see, we all just assumed would be like regular old IT data and it's not. The cultural problems I think are even harder. And we're spending more and more time thinking about those and trying to work with companies. They're hard, but they're fun. And there's so much potential to companies that are good at this out there tend to have really transparent, positive, solution-oriented cultures. Good manufacturers are all always solving problems. Here's an example of how this gets very real, very fast. Transparency. Almost every company says, oh, we want to know what the problems are, we'd love to see what's really going on. But in a lot of organizations, if you raise your hand and say, hey, you know, I thought our plant was doing really good, but we've got a problem here, you get punished. So suddenly when you come in and you get the data into shape and let it tell you how a plant is actually producing, you'll find all kinds of problems that nobody expected. And that can lead to great difficulty in scaling these programs because people in cultures that don't reward transparency have to protect themselves from the consequences of what the data say. It's a really interesting dynamic. And so you'll have sometimes you have management saying, oh, we really want to know what's going on. But if they punish people for the truth, then they're not going to find out what's really going on. No one's going to tell them. And so that's not a data problem. That's a culture problem.
Gemma [00:11:49] Yeah. I mean, let's dig into both of those. Let's start then with the data or technical problems that we spoke about before, as you were saying, about data being dirty and being able to just kind of make sense of the fact that this data exists but it wasn't really put there or built to be analyzed later on down the line. Let's talk a little bit of the idea of a data foundation. Why is it so important to have one in order to make sense of and use data? How would you go about creating one and then ensuring data is reliable?
Jon [00:12:14] So data foundation is a really cool term that people have started to use in the last year or two for this. And I think it's perfect. It's a great term because it is the essential, necessary kind of bottom layer for everything that we're used to doing with data. So if we want to do AI on a bunch of data, you can't just throw a bunch of raw data points into a VAT and apply AI. The data has to be standardized into units of information that are relatable. So when we use the term data foundation to describe manufacturing, we're talking about standardized information generated from all this terribly difficult data and generated into a couple of basic building blocks to stick with the foundation metaphor. The way our company does this is we represent work done in manufacturing facilities the same way regardless of what the facility’s making or regardless of what the assets are. So we literally have a row in a data table, a data schema that you can think of as representing each unit of work by a machine. Now, if you have a factory, let's say it makes cheese and I have a factory downstream that puts together gift baskets. And we don't know whether your factory is the problem in difficulties that we're having or mine and we're related. If I can actually compare everything happening in your factory and everything happening in my factory using the same standardized units of information, now we can start relating across sites and activities. So data foundation is a fancy word for the idea that information is standardized enough that you can go outside of one area or one activity and start making everything relatable. You know, the average car has thirty thousand parts. Let's think about this. You build a car, it's got thirty thousand parts. The car companies are trying to figure out all the time is the defect in the car the OEMs responsibility? Is it one of the suppliers, one of the hundred suppliers? When all of this information is put in standardized form, you can now look at that supply chain as a system and that's what everybody wants to do, is be able to really look across a bunch of different activities and standardized information is the key.
Gemma [00:14:32] So thinking about then, I guess, the specific types of data right, because we're thinking about how to bring everything together and create this beautiful machine, if you will, digital machine that works and makes sense and realizes the potential of all of these different inputs. But thinking about the types of data that manufacturers and plants actually create, are there specific types of data that you think are most critical? Ones that are going to have a big impact that maybe people are not thinking about to use but should?
Jon [00:15:02] One long-standing need and challenge in manufacturing has been the combining of quality data with data about the process. There are huge amounts of data about process, so automated processes just spin up a ton of process data and then there's often a quality function that does tests somewhere along the way often as well at the end of the process. And there's a bunch of quality data that's sitting there and this was a learning for us. We didn't realize how hard it was for companies to actually combine quality and process data because they're in different places. They're generated in different ways. They're on different time scales. And a lot of times when a part is made, all the data about production comes before there's any quality data. So the data is not married up. Quality data and process data putting them together is a home run almost always for manufacturers. There's a ton of Excel, CSV files, there's home-made databases. Our approach and Microsoft's approach, which is really impressive to see at scale, is to say to clients, let's get it all. It's all going to be useful. In the days of a relational database, we had to kind of know what we were modeling and go after data with the specific intent of building a model but we don't have to do that now. We want to get everything. So really valuable data's quality data, process data. There’s systems in manufacturing called MES - manufacturing execution systems. Those are really useful. There's a technology in manufacturing called historians. Companies often store data in historians. We like to get those. And really it goes on and on. Energy data. You know, it's not typically combined with production, but let's think about sustainability. Manufacturers literally have no idea which steps in the process waste energy and could use less. So if we can combine energy data from completely different sources with production data, now you can start to manage your use of energy in your process and get better. We'll take it from anywhere. And it doesn't matter whether it comes from inside the factory or not. If it's related to production, we want to know about it.
Gemma [00:17:09] Building on terms and ideas that are pretty buzzy and hypey that get talked about a lot, but mean many things. Let's talk about them. IoT, AI, machine learning, automation. You know, how do all these different ideas fit into the big picture of big data and making it usable. I mean do you need all of this? Or do you use some of them? How do you think about these things?
Jon [00:17:30] So I'll do my best to translate and distill these words. We struggle like everybody to get to a common vocabulary with our clients. There've definitely been chapters in this last decade in the application of technology to industry. The IoT chapter was the first one. So one way to think of IoT is as describing connectivity or infrastructure to get data flowing. It's an absolutely essential step in a digital transformation and it's kind of the base layer. Think of it as pipes. AIML [AI machine learning]. These are analytical techniques that we apply to data and there's a lot of focus on the algorithms. You need the pipes to get the data out and flowing. You need to get it someplace and then you need to do stuff to the data to make sense of it. And that's where the AIML comes in. And really, it's sophisticated math. And there's a whole range of analytics from descriptive statistics, inferential statistics, techniques like AIML. It's all analytical techniques. There's a lot of focus on automation and I think when people talk about automation and industrial transformation, there's actually two types of automation we talk about. One is physical world automation. There's a lot of equipment that's automating tasks. There's also software automation. We're Automating a lot of steps and software. So we think of IoT as being essential. It's about connectivity. We think of AI, AIML as examples of work you can do on data. And now there's a lot of really interesting, to kind of stick with this idea of buzzyness, now a lot of the best thinkers in AI are starting to talk about the focus really should be on AI models it should be on the quality of the data. It should be on things like data foundation. So you asked about data foundation. I think that's kind of the next category of buzzwords. These buzzwords play a function. They highlight what's important right now. Everybody kind of got the pipes. They got the tricks with the math. Now we got to get the data in shape so we can do the math tricks. And again, we're just trying to get to scale here. The more that you can put into repeatable processes in your software the more scale you can get.
Gemma [00:19:34] And what about digital twins and control towers you hear a lot about these terms as well. You know, how does, how does all this conversation around data kind of feed into this and the different systems organizations use to, again, really benefit the business at the end of the day.
Jon [00:19:47] Digital twin means a digital representation in math and software of a physical world activity or asset or process. And it can mean the simulation of a process so we can build a physics-based model of something. It can mean an empirical analysis of how something's working and it can mean a single asset or a whole system. So a digital twin is this idea of representing what's happening in the physical world, virtually. Control tower’s a great idea that's been around supply chain thinking for a long time. And it's the notion of being able to kind of at an airfield, sit in the tower and see all the planes coming in and landing, taking off know where everything is. Here's what's so cool about Microsoft's approach to this opportunity. For decades, supply chain has meant let's follow stuff on trucks and planes and boats as it moves from place to place. And let's follow the flow of materials between companies. We can do that. So it's measurable, but there's always been kind of these blind spots about what's actually happening in the plants. So to stick with the control tower metaphor, if you can see inside the planes as well. And know what's going on in the factories and combine that now you can take digital twins, what's happening in plants, combine it with your control tower functionality, and you can literally understand your whole supply chain, both the flow of goods and the making of all the materials and goods in the supply chain and get one view of the whole deal.
Gemma [00:21:16] Let's rewind back a little bit and touch on culture, because you brought that up earlier as one of the I guess the other side of the coin when it comes to challenges or hurdles, when it comes to, I guess, realizing the potential of all this stuff. How is the dynamics, or culture of an organization impact the ability to drive change thinking. About first, making sense of data, then using it to create business impact and then even using it to change how you work or how you make things. Tell us a little bit about that.
Jon [00:21:40] We believe a company's culture is the single most important factor in determining whether or not they're going to be successful in the future as a digital company. It's just so fundamental. One of the very best companies we worked with was actually the most primitive as far as its technology. But it had a culture that was deeply interested in truth, in data, in understanding what's really going on. So as we spoke about a few minutes ago, a culture that is interested in progress, is willing to take a little risk, and wants to get a little bit better every day. A company with that culture, regardless of where it's at technologically, is going to win. In manufacturing itself there are these great cultural traditions, ideas like continuous improvement. Lean Six Sigma is related to that as well. Really good manufacturing companies already have a tradition of working really hard to get a little bit better every day, and that fits beautifully with these ideas and digital transformation. They might be companies that make products with really high brand value. And so a lot of the value of the company is in marketing and branding and manufacturing is kind of an afterthought. On the other hand, if they're a company that really depends on making things well, odds are they got a pretty good manufacturing culture because they want to be a good manufacturer unless they have that. And so we really look to partner and work with companies that that have these values already. They tend to be great at adopting digital technology.
Gemma [00:23:16] So building on this point of culture is obviously all fine and well to have a culture of curiosity, wanting to change things, move things forward. But there's also a reality of actually having to understand these new tools and processes in order to put them in in a way that makes sense. You know, you can be excited that - you can be naive as well. Right? So you're going to need a level of expertise. So how critical and important is building partnerships to find solutions and improve outcomes? And how do you build this when perhaps the mindset of those in tech might be different from those manufacturers, and I'm thinking particularly some of the more traditional manufacturing companies?
Jon [00:23:58] That is a great question. There's a ton of justified skepticism in manufacturing companies. If you walk in the door and say, hi, I'm from a technology company, I'm here to help, you got a real hill to climb to earn trust and there's no getting around it. We spend a lot of time thinking about how to generate trust. You can't fake it and you use the word partnership. This kind of work really is a partnership. And Microsoft thinks a lot about these concepts around trust and outcome as a company. You know, when you screw up as a partner, the first thing to do is admit it. That's how you're going to generate trust. And so, as you said, these are very hard, difficult projects. And it has to be part of your culture that the way you work, you're going to generate trust or it'll never work with clients like manufacturers. I'll give you a couple of concrete examples. You know, you talk about culture within the client organization. Often we’ll go into a company and there's a metric in manufacturing called OEE, overall equipment effectiveness. What it really measures is how well are you making stuff compared to your kind of ideal maximum? So if you can make one hundred widgets a day, how many are you making at the right level of quality, really? And it's a little more complicated than just an output metric, but it's a widely used metric and it's very difficult to actually know. Companies will often say that they’re at 85, 90 percent OEE. But the reality is, once you look at the data together with them, they're often in the 50s or 60s. And there's always this moment where we go to these plants and they've been reporting up the chain that they're in the 90s and the data says, well, you know, we're in the 50s. Now that's cool. That means that you could almost double your production. There's all kinds of opportunity. But it's also a moment of potential embarrassment, frustration and almost always the client gets mad at the data, then they get mad at us. And you talked about earning trust. We show everything. We begin that conversation with we may have made a mistake, where are we wrong. And there's this very rocky period in many of our engagements where as the data is starting to come out and people are starting to see information in a new way, it's very uncomfortable and painful. But it represents a lot of opportunity. And the right cultures get excited and go, holy smokes, there's all this stuff I can fix. So it's opportunity in the eyes of some people.
Gemma [00:26:30] So spoke a lot about various different switches and changes, both in terms of the technology that's needing to be thought about and adopted as well as different mindset, it's different partnerships. So I would love to now just a couple of concrete examples and then also move on a little bit of advice here to kind of finish it off. But let's start with the examples. Once you're using all the data that's in a plant or a factory, you're developing insights and so on, what is it that you're able to do now or optimize or do really well that you couldn't do before? Give us a couple of examples of real business impact as a result of data.
Jon [00:27:09] Love to. So here's a recent example of a company we work with that is just so cool to see what they're doing. This is a company that supplies components to electric vehicles. So we're working with them in China, Japan, Germany, North America. All places with a bunch of auto plants. And first plant we worked at was in northern Mexico. Absolutely hungry, scrappy leadership at this plant. So much pride. And that's the thing, you know, good manufacturing environments they're so proud of the stuff they make. So this plant is in an intensely competitive environment. The automotive industry is brutal. There's tons of price competition. You're always kind of bidding against other people. They started to work with their data the way you and I have been talking about and the first thing they realized is they could improve how fast they make stuff and basically improve their productivity by about 10 percent. Just by making adjustments that had been essentially invisible for years before. Then they started to get very interested in quality issues because when they have to scrap a part, they have to throw it away. And they were basically taking 10 percent of what they were making and having to scrap it. So we worked with them to use machine learning to find very subtle, hidden causes of scrap. And the data helped them figure out how to improve that. We've already got a 20 to 30 percent improvement in that scrap. I'll bet we get it to 50 or 60 percent within a year. They are now way ahead of their competitors in having control over quality and being able to assure their customers that they've got good stuff. Here's what's really cool, the pride that people feel when they make that kind of improvement. And a lot of the folks that we work with who are brave, who raise their hand in this company. In manufacturing companies, unfortunately, if you take a risk and it doesn't work for you, sometimes you get punished, you get penalized. And I remember talking to this guy at the beginning. I said we were meeting with him, the chairman of his company, a 30 billion dollar company, the CEO of his division. And I looked at him and I said, what possessed you to raise your hand to do this? If this doesn't work... So he's going to have a great career at this company and beyond. The people on the plant are stoked. We're improving the sustainability very significantly and we're making them a better company, better competitor. Another example of where data is really making a difference in the industry is the glass industry. We were invited to go to a large glass plant a couple of years ago, and the idea was to study a furnace and to try to help them reduce the amount of energy that they use. A couple of months into working with this company, we realized that the real challenge for them, as with the whole industry, is around this notion of yield. Most glass plants around the world make glass of sufficient quality at a level about eighty five percent. Their first pass yield is eighty five percent. Which means 15 percent of the glass gets thrown back into the furnace and is remade. And what that really means is that at least 15 percent of the energy being used in the process is being completely wasted. Glass is one of those industries that is a huge contributor to global greenhouse gases. And if we can move the needle on yield and get it from 85 to 90 percent, we're going to have a significant impact on the energy just by making the process better and getting a higher yield. In this case, we were able to do a predictive analytic with this company that explained 80 percent of their defects. It takes actually three days for glass to go through the furnace and we were able to tell them two or three days before the glass came out, hey, the way the furnace is working, it's going to cause a problem. You need to adjust the furnace and prevent the problem and helped them achieve a significant gain in yield and a significant reduction in energy. There's a second chapter of this story that I'm pleased to share, and it's a great example of how these industries are really getting very serious about this. A couple of weeks ago, we started to work very closely with one of the companies that provides critical equipment in the glass making industry. They've been thinking about this problem for years and what they want to do is they want to put software on all the assets they provide to the world's glass making plants so that every company who's making glass can improve their yield. That's a big project. It's going to take a couple of years. But it's an example of the industry level transformation that's going on here. If we could get every glass manufacturer go from 85 percent to 90 or 95 percent yield, we'd have a 10 to 15 percent improvement in the sustainability of glass right away. And that's just the first step.
Gemma [00:31:54] I want to end on this question then, which I think follows on quite nicely from what you were just saying, particularly about people being brave and being able to take the plunge and make stuff happen. A lot of the time we focus on what can the companies do. But at the end of the day, it's individuals doing this within companies. Right? As people having to kind of put their hands up and say, why don't we give this a shot? So what advice do you have for people within companies? What can they do to accelerate business impact, both in terms of adopting these new systems and so on, but also perhaps some practical tips on maybe what's a step they can do right now with this understanding the technology, to trying to advocate on this behalf in the culture of the company itself?
Jon [00:32:40] Absolutely. So it's funny. Yesterday we had a conversation with a leader at one of the United States’ best manufacturers. This guy is smart, he's tough. He was so vulnerable. And this really impressed me. So here's suggestion number one. He said to us we were with him and another leader in his company and they were starting to build alignment around getting something like this going. And he said, I know the tech, but I don't know how to sell the idea of my company. It's not what I do. A lot of the people who are taking a chance, who are brave, who are trying to do this, they haven't had to lead change in their company before. So I think the first suggestion I would offer is build alliances, make friends, get a group of people around you who all agree on what you want to do and focus on the bringing together part the alignment part, getting other functions and other people excited. And there's no playbook for how to do that. It's an influencing skill. So suggestion number one is don't do it alone. The other one is insist on companies like ours explaining these concepts in one syllable words. Make sure that whatever it is you're advocating for really makes sense to you and all the tech speak and gobbledygook that gets thrown around, that you don't put up with it. Make sure that the ideas make sense. And three, partner with your technology partners. Don't view it as a vendor relationship. So many of these companies have processes and ways of working with partners that fail. You know, I tell companies that we work with I say this isn't a vending machine. You can't just put a quarter in and have a can of soda pop come out. We've got to do work together. That means I've got to be a good partner to you, but it means you've got to be a partner to us, too. And so I would say build alliances, insist on clarity and tech companies being able to explain what they do so that you understand it really and be a partner. Take responsibility. You don't just look at it like you're buying something. This is a partnership. This is work that we do together and take your half of the relationship seriously.
Gemma [00:34:45] Love it, Jon. That's a really lovely note to end on there. We've covered so much - everything from the more high level trends right down to the really specific advice, both about the technologies themselves, but also about how you really go about implementing. Because at the end of the day, a lot of the digital transformation and change when it comes to technology stuff, it really does come down to interpersonal things as well as about, you know, learning about the world of what's possible in science and tech. So, Jon, thank you so much for coming and sharing your insights and joining us on the show.
Jon [00:35:17] Gemma, it was a pleasure. Thanks for having us.
Gemma [00:35:22] That's it for this week. Thank you so much for tuning in. You can find out more about Jon's work and indeed some of the broader themes we discussed today in the show notes. If you enjoyed the episode, please do take a few moments to rate and review the podcast. It really helps other people discover the show. And don't forget to hit subscribe and tune in next time to continue our conversation about innovation, resilience, and our capacity to succeed.
Ad [00:35:50] Learn how Microsoft Dynamics 365 Supply Chain Management is helping businesses build agile and resilient supply chains. Request a live demo today by following the link in the episode description.