Erik: Samuel, thanks for joining us on the podcast today.

Samuel: Thanks for having me.

Erik: Great. I'm really looking forward to this one. It's your first podcast, as I understand. But your colleague roped me in with the fact that you're a listener. That always touches my heart when somebody reaches out and says they actually listen to the podcast. So a pleasure to meet you here today.

Samuel: A pleasure to meet you, too. I've listened to many, many podcasts already before, so I'm very excited to be in one this time.

Erik: Yeah, fantastic. Well, let's get into you before we discuss Pozyx in detail. I thought your background is interesting. Because, basically, you were doing your PhD. You were researching, I think, this very interesting topic of cooperative localization. Then that led you directly into founding the company back in 2015. What was the spark that said, okay, this research I'm working on is interesting; I think there's a company here?

Samuel: Well, there were actually several sparks. I think one was that, in research already back in 2015, research community was about indoor localization. That problem is solved. So it was very difficult to publish papers about it, to find novel things. Because things like the newest technologies, they all existed at least on paper in research. So my research was already about collaborative positioning which is something way advanced. And today, we don't even yet see a lot about that. So there is a big gap between what research says is possible and what actually is possible in the industry. That was one big motivator.

Then what really sparked it was, at that time, a new chip came to market, a new ultra-wideband chip. That's one of the technologies that we use. That new chip came to market. And really, we used that advantage to bring something to market, using that chip, that could really drive down costs of ultra-wideband location systems. We started the company, and we started with a Kickstarter campaign, crowdfunding, where we created really a kit to make indoor positioning possible to be used by pretty much everyone. So we used to get that. You could hook up to an Arduino or a Raspberry Pi, your robot or your drone. And you could start doing indoor localization already instantly. That was the premise of how we started the company back in 2015.

Erik: Okay. Cool. Yeah, I was thinking about IoT solutions in two different domains. There's the one domain which is the very expensive complex, pushing the cutting edge of what's possible in a certain area. That's often used in, of course, military applications. It might be used in certain heavy industrial applications like for a chemical factory where they're trying to understand 1,200 sensors and what that data means. Then there's this other area where it's taking one of those very sophisticated technologies and then reaches some point of maturity where you say, okay, 20 years ago, this was being used on missiles by the American military. But today, we can put it on a cow. The price point is at the right stage where this is something that anybody can use.

I think that's always very interesting. That the biggest impacts are often about just the cost reduction and the simplification and standardization of the technology, to the point where you can find all these value in these use cases that all of a sudden emerged, right? It sounds like that's the innovation track that you guys found. This place where a technology that was at one point fairly expensive has reached a point where anybody can use it via an Arduino. And I suppose now the cost is even much lower than it was when you found it. UWB, is that still the core of your tech stack, or is that one of the multiple technologies that you're working with?

Samuel: It's still the core, I would say. We've also expanded to supporting Bluetooth positioning, which can diversify a little bit about what kind of accuracy you require for something — ultra-wideband when you want to position something very accurately, and then Bluetooth when you want to position something less accurately.

We also added a significant part in the software piece, a software solution that can also accept GPS location and even RFID using the unlock standard. We've moved from only just being a technology company with ultra-wideband to something where we can provide solutions. The solution requires more than just location data. It really requires an application on top of that to drive that solution. That's something that has been some of the journey that we've been through in providing those solutions and not merely XYZ location data.

Erik: This concept of cooperative localization that you were working on, I guess the traditional way is kind of a hub and spoke where you have some hubs and then you have the sensors out in the field. They're communicating with the hub. Is that moving beyond that to more of like a mesh, the sensors are communicating with each other to make more precise localization? What's that concept of cooperative localization?

Samuel: Well, exactly, that's the concept in collaborative localization. You may have some sources that you can use to figure out your own position. Then you use that to help maybe another tracker to find its position as well. And so, after a while, everybody can figure out its location. That's the concept of collaborative positioning. We use it to some extent. The problem is, we cannot use it everywhere because it is still — it consumes more battery power. You have to communicate with your peers, so it requires more communication.

One of the downfalls of that is, of course, also that if you don't have anyone to collaborate with, then basically you can't position. And in a lot of applications, that's really not acceptable that you have a circumstance where you can't position at all. That's why this part of the research — again, from a research point of view — extremely interesting. But from a more commercial point of view, we use it in the deployment of the system. But then, in the actual running of the system, we typically don't use that.

Erik: Got you. If we look at the full tech stack, I know you launched your platform recently. So you're providing the sensor solution. You're providing, I guess, the back end. How close do you get to the front end? Do you have a front-end app? Because we're going to be talking about agriculture later? So do you have something that a farmer would be using directly on their mobile phone? How close to the front end do you get here?

Samuel: We have two products in a sense. One is really the location system by itself. That's where it all started. That system just provides location data. Then we have this software application where it really goes all the way to the front end. But there, it is limited to certain verticals. That's more for manufacturing and distribution. These other kinds of applications that you just touched upon like in agriculture, we are partnering with another company that has a great deal of experience in the agriculture business, where we do not. Obviously, they provide this whole front-end solution for that one.

Erik: Yeah, that makes sense. It's always a tricky question. Because on the one hand, to some extent, every technology provider would love the end customer to be opening up their app and seeing their logo every morning. On the other hand, a farmer doesn't want 12 different apps to check 12 different data sources. He wants one platform that he can use to manage his farm, right? So there's always a bit of a tension there.

When you work with these partners, how does that typically look? Is it you're providing a data stream, and then you're basically — they charge the farmer, and you charge them per data point? Is it a monthly, just kind of a flat monthly fee? What does that tend to look like?

Samuel: Well, maybe I can go into this partnership a bit. For this one, we've partnered with a very large company. We haven't partnered with dozens of agriculture companies. We've partnered with one, a very large one, in the dairy industry. And so when we were still a relatively small company, they saw the potential of this tracking technology of ultra-wideband. They said, "Okay. This is the future of animal tracking. We want to have that, and we'll require a technology provider to help us in that." That's where we came in. We provide for them the technology. Meaning, the infrastructure, the location engine for that customer. We custom built the ear tracker. That goes into the ear of the cow. We've set up production for them, or we assisted with them in the production of that specific tracker for them. We've trained all of their employees to deploy the system at scale and to support and then to maintenance it. So we are then still provide support, but not directly to the farmer. Only tier three support. That's what we then still provide. There, we're the technology provider. In that scenario, we have a commercial agreement based on licenses. That goes up with the amount of, in this case, cows that are being tracked. That's how that collaboration works.

Erik: What are the use cases? I mean, I guess, to some extent, the use case is tracking the location of a cow. But then, it's always more precise than that in terms of why do you want to know the location of a cow how are you using this. So what are the different use cases that they're trying to address with your solution?

Samuel: Yeah, that's a good question. That was a question that I had as well when they reached out. They said, "Well, we're very interested in this tracking technology. We want to really do it at scale. The cost has to go really down. Can you do it?" Because what they wanted to do with it is, finding the cow is really more of nice to have. It's still an important use case, in a sense. But what really drives the value is that this location data can be used in all kinds of analytics and behavioral analytics. This company that we worked with, they have a whole team of machine learning and AI experts on animal behavior. They are using this data to get some valuable insights.

And to give you a bit more of an idea of what that is and why the accuracy of then the location was so important, we were tracking cows. The cows are inside the barn, and they go about their lovely life. Sometimes they're resting in the cubicles. Sometimes they're walking around, and sometimes they're feeding or getting milked. For example, feeding. What happens is, they put their head through a fence typically, and then there's a lot of hay or any other feeding thing that they could eat. By tracking the cow — more specifically, it's on the ear of the cow — you can really see that the cow is moving in his head or her head through that fence to start eating. Then you can start detecting that the cow is eating. And so now you know that, okay, the cow is eating. How many times per day is it eating? How long is it eating? Same for drinking, the same for milking. You can do the same for resting, for playing. Yes, the cows, they play. And so that's all the kinds of things that you can really see by now having this accurate location. And you can really define some kind of an activity based on a certain location. That's what they use to really get a full view of what the cow is doing the entire day. The cow has zero privacy, in a sense.

It's remarkable then how much information is in that data. They can now really see, is that cow getting sick? Is it sicker? Is it getting sick? Is it maybe starting to limp, or is it time for inseminating the cow? That's, for example, something that I didn't know, which was kind of stupid of me. But the cow needs to be inseminated before it can give milk. And so this process needs to happen a couple of times throughout the lifetime of the cow, because milk production goes down. Then at the right exact time, you want to inseminate the cow again to get the optimal or the maximum milk production out of it. So you have to time it right. But how do you know when to inseminate the cow? And so that's something that based on this data, based on location data, is something you can extremely accurately predict. There is the true value of the system.

Erik: Got you. Then there's also this analytics engine behind that. I guess that is something that needs to be customized to the use case. Do you take an active role in helping to customize that, or you help them collect the raw data and then they decide how they're going to analyze that in order to figure out these things, such as is a cow becoming sick?

Samuel: Yeah, so we have a full analytic suite as well where, depending on how far you want to go, you can tap into that. Because just processing raw data, its raw data. It's a lot of data in the case of these animals. But in other cases, as well, we're getting location updates every few seconds or every several times per second. So it's just way too much data to process. And so, we are already transforming that to data that you can use. For example, time budgets, or heat maps, or really even metrics and KPIs that are relevant for a certain use case. We can provide that. It's not required for our customers. If they want to do it themselves, it's perfectly fine. But we also provide these analytics as well.

Erik: Yeah, I think this is fascinating. Most people, I suppose when they hear this type of tracking technology, immediately think I can find where something is, right? But there are some use cases, maybe if you have a tooling use case and you want to know where's the tool, that's actually quite useful. But in a lot of it, it really is about — not where is it, but where has it been for how long, in proximity to what? Then in aggregate, what does that data tell me, I think, which is quite interesting, actually.

Let's talk quickly about the scalability here. So you mentioned that the customer, when they approached you, really wanted to work on bringing the cost down so that this was affordable for them to scale across the operation. How does the cost for a solution like this scale would you go to 10, 1,000, 100,000 units? Because I suppose there are some aspects of it that scale quite well, certainly the software aspects. Then there's the hardware ones, which scale to some degree but maybe not as much, right? You still have to pay for atoms. So how does that tend to scale as you ramp up deployment?

Samuel: Well, for ultra-wideband, I would say it's not yet on the level of things like Bluetooth, for example, when you really scale it. But it's getting close. So I can't share the exact numbers of that. I think that's more for our partner. But it's coming down to a point where the cost of the tracker is getting less relevant. It's still relevant, of course. But I think when we started, when I was doing my research, I think we were experimenting with an ultra-wideband system. There, it was $2,000 dollars for a tracker or for an anchor. As compared to that, I think we're almost two orders of magnitude smaller than that. So the price has come down tremendously.

Erik: Okay, so you're looking at tens of dollars for the hardware. How does that compare to Bluetooth?

Samuel: Well, for Bluetooth, I think it depends a bit where you're producing. I think in Europe, the typical prices are also around that range. When you're going more to China, you could really get below the $10 point or even significantly below that. That's, of course, what do you have to compete with? But production in Europe or if you buy it more from a reputable source, I think most Bluetooth trackers are also in the range of between $10-$20 for tracking.

Erik: Okay. Got you. Then your colleague also mentioned a second use case, which was a global e-commerce leader and that's a bit more of an industrial use case. Can you walk us through what was the problem that they had, and then what was the solution in the end?

Samuel: Yeah, so it's a similar setup as with any animal collaboration or the agriculture collaboration. They reached out to us and they said, "Okay. We mostly require data. We require an RTLS system, and we also require some help putting this up and deploying this throughout all of our distribution centers," which there are plenty. So that's already an operational challenge that we had to tackle. That's where really our experience in the animal or the agriculture business case really came out. Because there, we already had to go through this exercise of, okay, how can we deploy the system at scale? And how can we do that without requiring always an engineer maybe to be present? But really, just by having technicians to be able to do that but also to support it and to manage it overtime. So that's more from a practical point of view.

What their challenge was, was really automation. So there, it's more tracking the cart` in a distribution center. They are going outside and then coming in again. They have to go in a truck. They have to go through the dock. Sometimes these carts, they're positioned in the wrong dock. So they go to the wrong location. That's one small thing. But the next step really is that you want to automate that. We have seen a lot of videos already about these automated warehouses where you have all this kind of robots picking up things and placing them again. That's exactly what's going on here, except that you don't really know where these racks are to begin with. So you don't know where to pick them up definitely because these racks, they leave the facility. They come back in. They are just loaded off the truck, and they're just spread around. So you have no clue of which rack is and what the content of that rack is, and where it belongs to.

By having really accurate positioning there, now you can send an automated guided vehicle to that rack and then have it pick it up and then get it into the automated flow of the rest of your facility. In this case, we're talking really about tremendous amount of racks that are moving from one facility to another one. This was a use case that has a very large return on investment. And now we're helping them pretty much deploy the system.

Erik: Okay. Interesting. I can imagine the messiness of this process. When you look at these two scenarios, one is agriculture. You're working with an animal. The other is working with machines. I guess they both have different challenges, right? What are the challenges that you encounter in the deployment, whether it's, let's say, the animal or the asset itself and fit deploying the physical hardware? Or, if it's more the challenge of the system as a whole and making sure that the system is properly set up? Or maybe it could be also human challenges and making sure that the users understand how to properly use it, and maybe modify their processes around this? What are the different challenges that you've encountered in these two different use cases?

Samuel: So what we typically see is that during deployment, still, a lot of things can go wrong. Well, for a location system to work, obviously, in systems that require infrastructure, you have to know where is my infrastructure installed. And if you want to have a very accurate positioning, then obviously, you need to very accurately know the location of your infrastructure. If your infrastructure, if you think it's in the kitchen but in fact it is in your living room, then obviously the system will always think that you're in the kitchen. So it has to be correctly installed. And so that's where we have provided a lot of tools to automate that process. The system can automatically determine its own location, and it doesn't require anything from the user there.

When we first started, people would still have to write down, okay, I've installed this infrastructure point in this location. And so they're noting that down, and then they have to do that through the entire facility. This could take a lot of work. And we saw that there are still a lot of errors in there. These errors are sometimes difficult to see. We just see that, okay, the location accuracy is not very good in a certain area. Then you have to start digging into it. That's where we fully automated that process, where it's now much more foolproof. So if people would swap things or put them in the wrong location, the system would detect that and give you warnings about that, give you even warnings. Okay. Maybe you have to locate it a little bit different. You don't have to measure the coordinates anymore. It's an option, of course, because it's less accurate than when you really go in and laser it out. So if you really need the highest level of accuracy, that's still the preferred way to do it. But if you don't require that level, then you can just fully just install it wherever. Press a button, and it will automatically tell you, okay, this is where all these anchors are located. They don't have to do anything manual anymore.

Erik: There's another technology that I guess you're probably more familiar about than me. The RFID or these tags that can come down to sense in terms of the cost if it's really high scale. They're just kind of passive tags that something will read and say, okay, it's at this location. For this e-commerce solution, do you see circumstances where these — do you imagine a future, maybe in the near term or the medium term, where every package going through this center would have one of these on it that says the real time or maybe not the real-time location, but let's say the last time this past a certain gate? Do you see those being a substitute, or a complementary technology, or a non-viable technology in these scenarios?

Samuel: Well, I think because this technology with RFID, they already exist for a longer time. We constantly come in contact with companies that have experimented or tried with that. For a number of use cases, it just falls short. Because RFID, you still have a high percentage that you will not be able to scan it, and definitely in more challenging environments where you have more metal. So it could be that you're not reading it even though it passed through this kind of a gate. Also, because you have to install these gates, you have a more limited view of where things are. That limits a large number of the use cases.

So I would say, for example, in this case with the e-commerce, there, it's more the cart that holds a lot of goods. The cart is something where an active tracker makes a tremendous amount of sense. Now, on the individual packages that would be in the cart where the price would be paramount, their RFID could still make sense if you have use cases for that. I would say, in the future, there a lot of companies now working on, okay, how can we marry these two? So we would have the advantages of RFID, low cost and maybe just the sticker. Then having still an active tag with some sensor data. Maybe not very accurate location but at least some sense of location. So these are things that could replace RFID perhaps over time. But that's still some way to go, I would say,

Erik: Yeah, that makes sense. So if you look out into the future, what do you think that's exciting here? Are new technologies on the horizon? Do you see either cost or capability changing in some way that will enable new capabilities? What's exciting for you right now in this industry? Also, with your business in particular, are there new products that you have coming out in the next year or so?

Samuel: Yeah, so what I would say is that, from a technology point of view, things continue to improve. But I would say that the use cases are still the most important. That was also one of the main drivers for us to also now start focusing on more of a software application, specifically, in our case, manufacturing and distribution. Because it's impossible to create an application for every kind of vertical. So we focused on those ones. Because we still see that the examples that we gave, these are some of our biggest customers. These are innovators. They have a clear view of what they can do with it. They also put in the manpower themselves to get that to the level that they can deploy it. And because of that, it still limits a bit how broadly location technology in general is being used today. For us, we have seen so many applications already talking to so many customers. We are trying to bring that together in some kind of an application that can really convince and drive business, that more businesses can leverage that. That you don't require an entire software engineering or integration team to get started. But really, you just have an out-of-the-box solution almost to tackle certain use cases.

The challenging thing that always existed, I would say, with indoor location is that there's not so much a thing as one killer application. Sometimes there is. But frequently, it's more a combination of all small or medium-sized use cases that all have some kind of a benefit. And if you add all of those benefits together, then you have a tremendous amount of benefit. But you have to have something that brings that together. And if you don't have that, people will always just look and search for that one business case that makes sense to install, or invest in a location system and then will say, well, for one use case, it's sometimes difficult to get that return on investment. But if you can have, so all of these use cases or several use cases combined, that you can just pick and match whatever makes sense for you, all of a sudden, it's becoming just more easy to install it and to make that decision to go for it.

Erik: Yeah, that's a great point. I often think about this kind of value stack where you can say, okay, if we have this sensor data, the maintenance team can reduce their cost of maintenance. Because they know where the asset is, or maybe the state of the asset. The marketing team can more accurately allocate marketing budget because they know where assets are deployed, how they're being used to some extent. The customer can get more value because now we can provide them with better analytics and information around that. Maybe we can charge for that. So now we have a new revenue stream. So you start to build up this stack of use cases that provide value to different people in different ways.

I think one of the interesting challenges here is that sometimes the different use cases, they provide value to different organizations, right? It could be, of course, different internal teams. But also it could be, it provides value to your distributor, or one of your solution partners, or to the customer, or to completely a third party that just wants to know maybe Wall Street. If somebody is trading on cows and they want to know, is there a wave of illness in Europe that's going to drive up the price of milk. I need to start making a bet on this market. How do you view these cases where the data is valuable to a wider set of organizations? Then you have this challenge of how to share data or monetize data maybe between organizations? Do you see this as viable or just too many legal and where people in organizational challenges to actually realize the value in a lot of those cases?

Samuel: Yeah, I would say we're not there yet. Because now, I think you need to get some use cases on the floor to really drive that solution. Maybe next that you can think about, what can you do with the data? Typically, we're more in the industrial space. So in manufacturing or in distribution, they're also protective of their data. Often cases, they would still require yo to have even an on-premise system that doesn't touch the cloud. We actually offer both. We try to explain the benefits of the cloud, of course. But even that is a challenge. Let alone, sharing this data with other parties or other stakeholders. So I would say that is more difficult.

Erik: Yeah, my experience as well. It's still that promise of the future, right? We're still quite early in the IoT journey just as an industry as a whole. Samuel, I think we did a good background of the business here, a good overview. Anything that we haven't touched on that's important for folks to understand?

Samuel: Well, I think what we focus on now the most is some specific use case. One is on the forklift tracking and the whole fleet management, trying to optimize there. Another one is in manufacturing, really tracking orders that go to a process flow. The final one is containers and carts, tracking those to really improve the efficiency of those. We haven't really touched much about that, and that's fine. But I just want to be clear that I think we're trying to get or bring all of our experience that we've seen with so many of our customers together in these applications in manufacturing and distribution. For us, we feel that we're coming to a point where it's less about technology, and it's more about the solution that it brings. I think that's an important evolution.

Erik: Yeah, I agree. It's this space in the industry where everybody has been doing pilots and watching each other. Now we're getting to the point where it just makes sense. People realize the value, and it's just about finding the right use case, the right ROI, and then making those investments. That will be in the next 10 or 20 years, that people case by case making those investments.

Samuel, I really appreciate you taking time. I wish you and the company a bright future. I think you're certainly in the market at the right time. So I hope the growth continues.

Samuel: Alright. Thank you very much. I hope my first podcast was up to your quality standards.

Erik: Yeah, absolutely. Let's do something again in 24 months or so. I would love to see where the company is then.

Samuel: Yeah, that will be awesome. Thank you very much.