Digitizing Supply Chains and Manufacturing

STEVE HENDERSHOT

The events of the past few years have exposed a major weak point with many companies: susceptibility to supply chain-related risks. As these challenges cascade, hitting sectors and regions across the world, organizations are innovating and working to build resiliency.

That’s the case with manufacturing. 5G-enabled robots can move across the factory floor, unencumbered by wires and cables. The efficiency potential is off the charts, and for organizations interested in a first-mover advantage, time is running short.

LEEFKE GROSJEAN

If you do not get involved in this now and get onto this wave in time, then you will not be among the first to make use of this enormous potential and opportunities that are in this area. It’s time to jump on it, and so I think we will see a very fast development over the coming years.

NARRATOR

The world is changing fast. And every day, project professionals are turning ideas into reality—delivering value to their organizations and society as a whole. On Projectified[r], we’ll help you stay on top of the trends and see what’s ahead for The Project Economy—and your career.

STEVE HENDERSHOT

This is Projectified[r]. I’m Steve Hendershot.

As I mentioned at the start of our episode, organizations across the globe have experienced dramatic upheaval over the last couple of years, as COVID-19, materials shortages and geopolitical conflict have combined to wreak havoc on the global supply chain.

Three-quarters of companies experienced external supply chain disruptions in 2021, according to a Hubs industry survey. In response, companies are looking to bolster their manufacturing operations with greater resilience and improved predictive capabilities, as well as more efficient factory operations.

That equates to an array of fascinating projects, and today we’ll meet a couple of project leaders working to define the industry’s future. We begin in Stockholm, where Leefke Grosjean is a senior researcher at Ericsson and project leader for the 5G-Smart project. Over the last three years, her team has worked to show how 5G can boost smart manufacturing, launching live-production pilots at three factories in Europe.

MUSICAL TRANSITION

STEVE HENDERSHOT

One of the interesting facets of this project is the combination of industry partners and the different real-world contexts where you’ve gotten to test this 5G technology. Give me an overview of the project—what you’re doing and what you’re aiming to achieve.

LEEFKE GROSJEAN

This project was planned at a time where 5G for industry was still in its very early phase. Everyone was talking about the potential for industries, but it was actually never really quite validated in real life. So in particular, not for advanced use cases. We wanted to go from theory to practice, from pen and paper to real trials, and also from bilateral engagements that we’ve done before to involving more and more partners and, eventually, the entire ecosystem.

We have one trial in Sweden inside an Ericsson factory where we, together with ABB, look into collaborating robots, as well as interaction between humans and robots. And then we have another site in Aachen in Germany where we look into workpiece and shop floor monitoring with the help of sensors. And then in the third trial in Reutlingen, also in Germany, we’ve been operating mobile robots inside a 24/7 operation or semiconductor factory of Bosch. The team working towards this is multidisciplinary, so we have 16 different partners involved, ranging from ICT partners such as Ericsson to operational technology suppliers [such] as ABB and Bosch, but also network operators, device vendors and academia.

STEVE HENDERSHOT

How did you pick these three pilots and test sites? Is it just that these were what the manufacturing partners wanted, or did you pick use cases that were sufficiently different from one another to provide very different data points?

LEEFKE GROSJEAN

It was very important for us that we actually go inside an operational factory, that we’ll actually see how does this all work on a real factory floor and not an artificial one and not a research setting or anything like that. Because in the semiconductor factory, for instance, it’s a very complicated shop floor. There are very high ceilings. There are very narrow corridors. It is kind of [a] difficult setting. What are the requirements, also from that aspect [what] was important for us to learn [was] what does it mean to go inside this factory? In this particular factory, it’s actually a cleanroom area, so there are very specific constraints on who can enter and what can enter. So that was very important for us.

In Reutlingen, we have these robots, but we also have TSN/Industry LAN use case there. But in Aachen, the focus is more on process monitoring, so that is more the aspect of monitoring using sensors in order to retrieve information from the use cases and what can we do with that. And then for the Kista trial site together with ABB, it’s a lot more about the collaboration of robots and also this aspect of human-robot interaction. So what does it look like for a factory worker on the shop floor? How will he or she interact with a robot, and how can we make that interaction better?

STEVE HENDERSHOT

This is especially interesting from a leadership standpoint because you’re leading a collaborative project that includes lots of organizations with different cultures and interests. Was this a unique experience for you?

LEEFKE GROSJEAN

I haven’t done this kind of thing before. It’s a very interesting project management challenge. And actually, having [the] COVID pandemic in the middle of it made it, from a project management aspect, even more interesting because every partner had a different take on how to handle the COVID crisis. So we really had to take in this information from all of the different teams, and we suddenly had team members who had to go on vacation, reduce your working time and all of these things. And I had to keep track of those 16 different partner organizations and restructure, reorganize, and still continue to strive and fulfill our promises.

But on the other hand, looking at COVID and what happened during that time, we all knew also during the time that we were working on this project, that this is frustrating and everything, but what we are working on in the project is absolutely relevant for future crises. So these remote-controlled robots or factories that can quickly change their production from one product to another one, this is what really makes the production continue during times of crisis. So the overall thinking we maintained in the project was that this is important, this is the future, and that this is what really brought us successfully through the crisis as well.

STEVE HENDERSHOT

What were some of the other collaboration challenges you had to work through?

LEEFKE GROSJEAN

One of the challenges in the very beginning of the project had been just to understand each other. We have been very enthusiastic and discussing and working together in the beginning, and then we noticed we didn’t quite understand each other. We didn’t understand what we meant when we said something like “end-to-end delay” or something like that. It wasn’t clear for everyone, although everyone thought that it would be clear.

End-to-end delay, for instance, coming from Ericsson, you would think of the end-to-end delay as something which is only focused on the radio, while somebody working with robots does think of this as the application end-to-end delay. And even though you might think you say the same thing, you don’t. It was really a very large difference in the beginning. We actually had to write a document together to find our common terminology.

STEVE HENDERSHOT

So now you’re three years in, coming up on the project’s close in May 2022. What are some of the key findings?

LEEFKE GROSJEAN

What we’ve seen is that if we really want to go to this 5G enhancement and not look at this as simply replacing a cable with wireless links, then we also need to take a deeper look at how this information is communicated. For instance, these robots, what information needs to be sent by them and in what format can it be best sent? Because if we then also work on that side and improve that part and not only go for the old way of sending it wired, then we can see how we can together come up with much more smarter ways of wireless data exchanges in the factory.

It’s also about the other enhancements that we’ve seen, how we can gain to higher-level goals going from these improvements and the use case execution that we’ve seen, and then actually measuring these effects on our efficiency or the waste reduction or the battery levels that we could reduce. So that is also something we could only see when we actually did it in the end.

STEVE HENDERSHOT

How many of these findings are transferable good practices coming out of the project? What can be widely used across factories versus a specific learning that’s really only applicable to the specific application?

LEEFKE GROSJEAN

The main idea behind this is to remove functionality from the device itself, and [move] then into the cloud. And this cloud control then allows you to do all of these things, as having a common map or optimizing paths. And this is, of course, very specific, what we’ve done in the project, but I think this is extremely transferable because in the factories already today, you have fleet management, and you try to do this in a nice way. But Bosch approached us and wanted to improve on that because they could see that, for instance, the robots that were encountering each other on the factory floor, they couldn’t move close to each other. They couldn’t really operate in the narrow corridors because they were not collaborating with each other. They didn’t know about the path of the other robots. Fleet management, for instance, is something that you would immediately be able to use also in other factories than just in this particular use case that we’ve been looking at.

STEVE HENDERSHOT

Looking ahead, what’s your sense of what’s possible when it comes to 5G and manufacturing? How can these particular use cases demonstrate to the sector at large what can be accomplished by embracing this technology?

LEEFKE GROSJEAN

I think the potential for 5G-supported manufacturing is amazing. Factory automation and process optimization, but also digitalization, these are areas that have already come to an impressive level in the area of manufacturing. But now, adding 5G on top of this, this really opens a new whole world of opportunities. With 5G, it is now possible to remove cables and transmit even critical information wirelessly, and this means that we can now, for instance, make even critical robots mobile because we have a reliable communication to them, or we can move around machines and equipment on the factory floor more easily while keeping this communication link on, and then make this vision of [a] fully flexible factory a reality.

But then only looking at this at removing the cables, that is only one single aspect. If we can see this more generally, then we can see—and that is something that we’ve seen in the project as well—that 5G contributes to so much more than just this mobility and flexibility. So what we expect, for instance, is that efficiency and productivity in factories will increase because we can have mobile robots deliver material in production chains in a more precise and coordinated way. We expect the quality of production to go up, for instance, because we can monitor production processes now with higher degrees of freedom when we can use wireless sensors. We can also expect to contribute to safety aspects. If we have shop floor monitoring, then we can also identify more easily dangerous situations or areas on the shop floor where we need to do something and then potentially automate those processes in the production that would typically put factory workers at risk.

MUSICAL TRANSITION

STEVE HENDERSHOT

Companies are also launching projects featuring tech, including AI and blockchain, to better plan supply and demand for their customers. Take demand sensing—using algorithms to make very specific predictions about what products will be in demand where and when, and then adjusting manufacturing plans accordingly. One such tool is in use at global retailer Ikea, where Peter Grimvall is supply chain design and planning manager, based in Basel, Switzerland.

MUSICAL TRANSITION

STEVE HENDERSHOT

The global supply chain has come under a lot of stress—and scrutiny—over the last year or two. How is the industry changing due either to these external pressures or to emerging technology?

PETER GRIMVALL

I think it’s a very interesting moment in what has happened recently, of course, with COVID and also other supply chain interruption. And that is that, of course, technology hasn’t been very targeted to improve perfection and automation. And something we also need to think about: risk management. How to manage an increased volatility together with the new technology that is available. Pre-COVID, I think most companies were targeted to refine and perfect what they had. And now many companies have to rethink a bit on how they plan their supply chains going forward.

STEVE HENDERSHOT

Amid this disruption, you helped create an AI tool called Demand Sensing to improve the accuracy of Ikea’s demand forecasting, helping to better estimate the products that Ikea will sell across its different locations, including e-commerce, over the course of a year. How do technological insights allow you to hedge against some of these evolving supply chain trends?

PETER GRIMVALL

For us, we were in an environmental change already in the retail area because the retail landscape is really changing. And in one way, you can say my job was fairly easy in the past. The customers came to our stores. They picked up the goods. They paid for them, and then they put them in the car and drove home, yeah? So it’s quite simple to plan, and also our stores were similar in size and in offer. Something that really started before COVID—but really got traction during COVID—is that we have to manage many different sections today. It’s not only the store. You can buy online. You can pick and collect. You can do the lockers that are actually very popular right now. You can choose home delivery services and assembly. So in one way, this planning complexity is just increasing exponentially, and then continuing with the same level of automation and requirements and accuracy.

It’s not only about how to react on things that happen. Demand Sensing is a great capability to pick up trends quicker, so not only like downturns and those interruptions, but where we really see the value is that Demand Sensing will help us to come back in a much faster way compared to if you wouldn’t have Demand Sensing. Because we can work with each individual article in each selling unit every day and by a high level of automation, yeah? So then we can quickly adapt on what happens, but also of course find the way back when of course our customer behavior changes because of those big interruptions.

STEVE HENDERSHOT

So now let’s get into the genesis of this Demand Sensing project. Ikea is both a manufacturer and a retailer, so you’ve already sort of had an advantage in terms of this sort of insight. And now with the Demand Sensing tool, you’re taking that to another level. What was the inspiration for the project, and when did you sense—pun intended—that the time was right to pursue this?

PETER GRIMVALL

You can say in Ikea, we like to keep things very simple, and we had actually a very entrepreneurial approach on this. We just asked one of our data scientists if he could beat our traditional demand plan in the short term for one of our most predictable and popular products, which is a plastic storage box called Samla in a random store in Germany. So that is actually how we started. We were really exploring what we could do with machine learning and with our data science team. After [the] first success, we of course wondered if we were lucky or not, and then the challenge increased with adding more products and then more stores and then [an] entire country.

But then of course, Ikea is, as you say, very much vertically integrated. We are, of course, different companies, but we work under the same trademark, and that means that we are very integrated from the point of sales independent if it’s online or in our stores, down to our own factories and our own manufacturing partners. So the signal that we can pick up in the demand in each of our selling units, that is actually forwarded to our production partners and our own in-house manufacturing.

STEVE HENDERSHOT

So you’ve rolled this tool out in a few countries. What has been the impact on the ground? Once you have these insights, how does that then influence the day to day in terms of how and when calls are being made about when to accelerate, slow down or manufacture different products?

PETER GRIMVALL

You can say we have a very global sourcing approach. We have a lot of regional supplies, and what we sell, for example, here in Europe is mainly produced in Europe, and what we sell in Asia is mainly produced in Asia. And with Demand Sensing, the initial benefit is actually in the specific selling unit—because we are still in a rollout, so we are rolling this out store by store, country by country. So the immediate impact is actually in the accuracy in that specific store when we implement it. And then as we have this more global sourcing set up actually in manufacturing, we will only see the big effect once all selling units are covered. Because the manufacturing unit of course delivers to many different stores. So that means that the big difference for manufacturing will not happen actually until we have rolled this out globally.

But it’s also the local planning in each store so they can plan with the right inventory for their coming two to three months. We are roll[ing] out in [the] U.S., so of course, if you have a local supplier already delivering to [the] U.S., their entire supply plan is then driven by Demand Sensing. And so we already see effects from the manufacturing perspective also, but until we see it on the global manufacturing scale, then of course everything has to be implemented.

STEVE HENDERSHOT

Does this also increase your durability? It feels like we’re encountering black swan-type challenges more and more often—can this tool help you endure some of those unexpected circumstances?

PETER GRIMVALL

We have seen the last two years is that those extraordinary events are becoming more frequent. We also have ones like hurricanes or port congestions, strikes. The weather is actually also putting a lot of stress on the supply chain. And that is also where we can see that working manually or fully manually to adapt to these changes, that is very difficult for a company of the size of Ikea and also the vertical integration that Ikea has. So being quicker to adapt and being able to have tools that actually can pick up trends quicker—not only when things happen, but actually they’re rolled back because in all those changes, our customers’ behavior also change and their preference and their needs. That’s why it’s so important also to pick up the new needs in a much faster way compared to before when we used more statistical forecasting methods.

STEVE HENDERSHOT

Looking back on the project development or implementation, what are the takeaways? What are the learnings or good practices going forward as you consider other similar applications and tools?

PETER GRIMVALL

It has been run as a project, which I think is important when you have many stakeholders that we actually don’t work with that often, and especially implementation in different countries. But we have also adapted an agile approach in this. And we’d have to try, and at the same time, we’ve not been afraid of actually rolling it back if it doesn’t work. That actually happened when we were rolling out in our first country in Norway. There was an odd peak in the summer sales that algorithms couldn’t really pick up. But the people working in Norway, they saw, “We work here. So we know that we will have this peak, but actually, the forecast is not picking up.” We rolled it back, fixed the problem and then went back online. And I think that has been our approach. Speed is very important, but then you also have to be prepared to take one step back if it goes too fast. In the end, when you dare to do mistakes, I think that is also the time when you actually learn a lot.

STEVE HENDERSHOT

What do applications like Demand Sensing mean for supply chain tech? And what do you expect the legacy or impact of this project to be on Ikea?

PETER GRIMVALL

First of all, I think it’s important to be able to do something at scale. What we have seen from many industry peers is there’s a lot of proof of concept, explorations and so on. But then there are fewer and fewer [initiatives that go to scale], especially for companies with a bit longer history than some that are purely digital and online from birth. But it is difficult to go to scale. And that’s also what we try to joke a bit [about] in the development organization. It’s only two things that are difficult in life. First is the proof of concept. And the second is actually to scale up, yeah? And this scale-up, I don’t think that should be underestimated at all. For Ikea, it also means a lot that [we] are able to scale up something with AI in our supply chain. And that I think will just motivate for more things to scale up within this team because there are so many more application areas.

MUSICAL TRANSITION

STEVE HENDERSHOT

It has become clear over the last couple of years that business as usual isn’t going to cut it. Fortunately, a run of smart projects around the world are demonstrating that the supply chain and manufacturing sectors are entering a new era—and business as usual isn’t part of the plan.

NARRATOR

Thanks for listening to Projectified[r]. If you like what you heard, please subscribe to the show. And leave a rating or review—we’d love your feedback. To hear more episodes of Projectified[r], visit Apple Podcasts, Google Play Music, Stitcher, Spotify or SoundCloud. Or head to PMI.org/podcast.