Join the Insider! Subscribe today to receive our weekly insights
Willem (00:00)
Welcome, you’re listening to the IT/OT Insider Podcast. I’m your host Willem and in this special series on industrial data ops, I’m joined by my co-author David and Martin Thunman from Crosser. If you want to hear more about what’s shaping the world of industrial data and AI, don’t forget to subscribe to this podcast and our blog. Welcome Martin.
Martin Thunman (00:21)
Thanks a lot, really happy to be here.
Willem (00:23)
and welcome David.
David (00:25)
Thank you Willem. And I’m also really happy to have Martin here with us. Martin co-founded Crosser in 2016. I’ve been following them for the last couple of years because they focus on integrating OT data from everywhere to everywhere. Is that more or less accurate, Martin?
Martin Thunman (00:43)
Yeah, no, I think that’s a good good starting point, at least that’s where we started. Then, you know, things have evolved lot since that, but definitely one of our key key areas.
David (00:55)
All right, so let’s kick off with maybe just, I would say, short introduction. How would you describe Crosser when you’re sitting around the table? What are the first things you tell your customers or whoever is interested in Crosser?
Martin Thunman (01:11)
Yeah, I think we like to think about Crosser as a kind of a next generation integration platform that is purpose built for industrial verticals. So we see ourselves kind of as a combination of industrial data ops, next generation iPaaS, but also stream and event processing platform all in one providing a number of capabilities that will
probably go in little bit more in detail in here. So I think that’s kind of how we like to see ourselves.
David (01:47)
And for those who don’t know what iPaaS is, maybe just a brief explanation.
Martin Thunman (01:53)
Yeah, so I think integration platform has been around for decades, Most companies, people are familiar with the traditional enterprise service bus and platforms like Microsoft BizTalk. That was kind of the first generation of integration platform, system to system, often very batch oriented platforms. iPaaS was the next generation that kind of moved things up to the cloud.
and on-prem is more in hybrid nature. But again, just integrating more application data and not just streaming industrial data, but also application. So ERP to CRM, MES to ERP, whatever scenarios you would have more on an enterprise and cloud and SaaS level. So it’s basically stand for Integration Platform as a Service and leading.
know, traditional legacy player in this field are players like MuleSoft and Boomi and Informatica.
David (02:52)
Yeah, yeah, OK, so I think maybe to, say, as a first discussion point, I’d like to refer to our industrial data platform capability mapping. So we published that a couple of months ago. We need to start making sense of all the offerings we see popping up on the internet. There are a lot of players in the data ecosystem, but they don’t focus.
not all of those players focus on, I would say, the same capabilities. So that’s why we made this first capability map. And maybe as quick side note to our listeners, so you can find that map on our blog, and we’ll also put a link to the map in the show notes. So maybe, first of all, given our map, given these seven capabilities, where do you position yourself? Where do you, I would say, maybe outperform the competition? What’s your take on this?
Martin Thunman (03:52)
Well, I think it’s a very, very interesting kind of first take that you’ve done with that mapping. I think Crosser, we kind of map a number of those steps, but Crosser is a data in motion platform. We don’t store any data. We typically say it’s a combination of intelligent edge layer and cloud. So we always…
are in the context in the ecosystem. I think one of the comments that you made in your article is about this is not one platform. This is an ecosystem that needs to work really well together. And we see ourself as an intelligent layer between any industrial data, any on-premises cloud and SaaS application. But in terms of capabilities, of course, it all starts off with
connectivity, very important part in industrial. But I think most companies have realized that just connectivity and syncing data and moving data from one place to another is just not enough, right? So, why do you have a platform that has connectivity? Why don’t you take the opportunity to do things with the data while you’re moving it? So this is where kind of real time processing comes into play.
David (05:04)
Yeah.
Martin Thunman (05:17)
Crosser, I think one of the part that we wanted to be really good at from the beginning was to not only move and transform data, but also do things with the data. I do deep analytics, real-time analytics on the data in motion and be able to build logic and workflows and automations around that data. And I think that is a part that I didn’t really see very well articulated in your
David (05:39)
Mm-hmm.
Martin Thunman (05:46)
capability mapping that think we believe is one of the differentiators with Crosser and other players in this field. So I think that’s first one observation I had in that. And as I said, we always work with data in motion, sorry, data at rest players. So that could be local historians, local databases. Some companies wants to have a combination of
David (05:47)
Mm-hmm.
Martin Thunman (06:16)
storing everything raw locally at the same time, sending transformed ready to use data to a cloud layer and WoodCrosser, we can facilitate both. So I think that’s, but we always rely on a third party storage layer. Same goes with visualization, which is of course very tightly integrated to historical data. Although more and more platform comes
David (06:41)
Yeah.
Martin Thunman (06:45)
with kind of real time dashboarding capabilities. So in that case, we can feed straight into dashboards that have that capabilities.
David (06:58)
So that means, if I can summarize that, there is the connectivity part, there is the transformation part, and that’s a very good point that we didn’t really touch on that right until now. So something to remember for our second release, because we also want to use these interviews as inputs towards a second release in a month, month and a half, two months maybe. So that’s very good input.
the analytics parts as well. So really, would say in the map, we’re sitting still on the left side. I would say from, if you take a bit more look to the right side, so you mentioned you integrate with additional tools, you integrate with trending tools or maybe with storage platforms and so on.
Maybe couple of examples where would say where the strength of cross relies or what the complexity would be to if cross wouldn’t be in the equation on how to get the data over there.
Martin Thunman (08:02)
Yeah, I think one of the challenges companies face is that they have so fractioned application layer. There are ERP systems that could be different from different vendors because different plants have different, they could be from different vendors. They’ve acquired a company and they have different ERP system, different sites. could be different versions of ERP system. So ERP integration is one very common one, but there’s any type of…
This could be QMMS systems, could be MES system, could be CRM systems, supply machine systems, and so forth. Crosser we’ve verified that we can integrate to over 800 different systems. And I think that is a little bit on one of our strengths that we, not only on the industrial data connectivity, but also on the IT and enterprise system layer, we have a very, very strong support.
David (08:49)
Yeah.
Mm-hmm.
Martin Thunman (08:59)
So that is one important part. And I think another area that you in your article, you’re talking about capabilities, which is super important. And because without capabilities, you don’t get anywhere. But I think another topic for kind of evolving your content around this to help the end user is to…
to go a bit deeper on what you call kind of supporting capabilities, which we find is often a really, really differentiating part for many of our customers. It’s about, and just for the listeners who haven’t really read and studied the article, the supporting functionality, that’s kind of the…
David (09:28)
Yeah.
Martin Thunman (09:49)
deployment, monitoring, the user management, all of these kind of boring enterprise features, but that makes all the difference when you’re rolling out to a multi-site in a multi-site world. And also, I think another part, which I believe could be of value for the listeners is kind of the architectures because
David (09:50)
Yeah.
you
Yeah.
Martin Thunman (10:16)
In this world where there’s often a mix of an edge layer and a cloud layer, the edge layer can be built in different ways. And the traditional disadvantages of on-premise software with having being a monolithic application, the ones you install it, almost obsolete because the new features available, IT doesn’t want to upgrade.
So you’re stuck with that type of scenario. So I think that is something that we identified when we started the company. We kind of asked the question, would it be possible to develop an architecture where you can have the same advantages that you have in the cloud, but deliver that in an on-premise or distributed environment? And this is not an easy task to do, but this is one of the advantages that
Willem (11:13)
Hmm.
Martin Thunman (11:13)
that we wanted to achieve. So going back to your question, having the ability to deploy and distribute and run a network of different use cases in different sites over the full lifecycle, we believe are key for customers.
David (11:34)
Yeah.
Willem (11:37)
Martin, you were touching on something that I’m not very familiar with, which is the analytics in motion part. Do you have maybe an example of what it means concretely? Why can it not wait for an hour, so to speak?
Martin Thunman (11:47)
Sure.
Yeah, sure. No. we have one could be anomaly detection. That’s a kind of very common use case here. So identifying anomalies and be able to take actions on those anomalies straight away. And that action could be, okay, here’s an anomaly. It’s outside the range of the values that we want it to be. So we should either
send this to an HMI and an alert to an operator to double check this, or we integrate directly and create an automation directly back to PLC and stop a machine. That is one scenario that the customer is using this for. Another one is if you’re running machine learning models distributed in the edge. It’s another capability that many of our customers are leveraging.
Then you want to often combine that in some type of closed loop automation. So that is another important value adding capability that you can have if you have the capability to identify, create conditions either through machine learning model, but being very dynamic or with fixed rules and be able to build automations with that. We have one customer that is
that’s reading real-time data from a historian, taking in thousands of data tags and are running individual conditions on each tag. And depending on the outcome of that real-time analytics, they then integrate back to their SAP system for dynamic work order creations. So that will then trigger a work order for an operator to go check it.
machine or change the setting in their process and so forth. So yeah, being able to act fast, think is super relevant. And with a growing number of sensors, also new type of sensors like video sensors or audio sensors, being able to create that real time intelligence and actions opens up new opportunities for customers.
Willem (14:18)
I’m also intrigued by the closed loop that you were mentioning with machine learning. Have you seen this evolved? I mean, you guys are working for a while on Crosso. Has it been taking off like only recently? Has it been a long path? And how does, let’s say, the OT world reacts to working with machine learning models?
Martin Thunman (14:39)
Yeah, now we’re touching a very interesting subject which has to do with the skill, different skills of different teams. And I think we see ourselves and that’s also one of the things that we wanted to do when we developed the platform. It’s like our crossroads are low code platform where you have drag and drop and you build visual workflows. And we looked at this and we saw, okay,
everything can be done in code. If you have developers, you can deploy everything in code. But we looked at it and said, hmm, in the modern industrial world, there are different capabilities that needs to coexist. And there needs to be collaboration between these different capabilities. So we’re looking. if you think about the flow in three steps. So you have the industrial connectivity on one side.
you have kind of the logic or the automation in the middle. And then you have often an enterprise connectivity in the end. Here you see all of a sudden there are three different competence sets that comes into play. In the beginning, the first phase, that’s where the OT expertise comes in.
They know their tag data, they know their assets, they know their machines, they know their industrial process. They can get access to the data, they can create transformation, they can contextualize the data with whatever metadata they want to add on top of that. But then you come into the more of the kind of analytics part, the middle part. And when you have machine learning,
Here we then introduce often either internal data scientist team or data analysts, or there are third party vendors or consulting firm that have helped developing machine learning models. These are most often if we forget LLMs for a moment, these are traditionally Python type of models. So here in the middle, you’re then different
different set of people. And then in the end, that’s the domain of the IT guy, right? The ERP system, SAP, no, no, no, everyone cannot touch, you know, SAP, or it could be their cloud instance, it’s Azure AWS, or, or Google Cloud, or whatever Snowflake or Databricks or, or RvEgo Connect, whatever it is, it is, you know, someone else. So the visual components of this being able to
That allows for teams to be able to, I build a first step, you build a second, and IT build a third, and we can get the working end-to-end flow together. That’s typically what we’ve seen evolving. answering your question, Vilum, it is definitely something we’re seeing picking up. After November 22, when everyone had the epiphany of the chat GPT,
This became a topic that previously had been very much driven bottom up. From now since then, in every boardroom there are conversations that are taking place now. like, how can we create advantages with AI? So now it starts trickling down from a top-down perspective, which opens up a more kind of a pressure to innovate for industrial companies.
Willem (18:04)
Mm-hmm.
David (18:11)
Yeah.
Yeah.
Willem (18:30)
Yeah, Martin, David also mentioned that there’s a use case you brought along. So I’m curious to hear a bit more about that one also.
Martin Thunman (18:42)
Yeah, think picking on use cases is always a little bit different because different customers implement different things. But I like to talk about use cases that goes beyond just a traditional kind of data transformation, shipping to the cloud type of, you know.
Two, three years ago, this was implementing a common data model. Now it’s UNS. It’s just like UNS in my world, the way I see UNS is it’s a common term to think about implementing a common data model. And then that can be implemented in different way with different technologies and so forth. But in the essence, the business value and the value that it brings to the company.
David (19:16)
Yeah.
Martin Thunman (19:36)
is to standardize their model with very clear hierarchies and very clear structure, et cetera. I think that is, for our customers, been kind of almost kind of the baseline, if you will. So I see that very much as a baseline, but a super important baseline, because without the baseline, you can’t move as fast, right, if you have to reinvent every use case.
David (20:06)
And maybe too, because this is also a topic which is really, really close to what I’m thinking about. one of these things is with, I would say with data management, as such is not really a new thing. We’ve been doing data management in the industry for many, many, many years. But now there is this…
Martin Thunman (20:06)
So that’s,
David (20:33)
There is this new indeed, this new vibe which originates in probably in the unified namespace discussion. Are we now doing things differently compared to a couple of years ago? Is there are there new tools available? Is there a new way of thinking available? Or is it just the fact that we are now putting some focus on data management?
Martin Thunman (20:57)
Yeah, no, think it’s a little bit of a combination, to be honest. I do think there are more and more players are focusing on this area, which drives innovation. And the innovation in my head is making it easier for the end user, which always is a good thing. So although this has been implementing, customers been using and thinking like this for a long time, just by the industry kind of coming together and say, okay, let’s call it UNS.
let’s try to get some momentum going. Let’s educate the market that this is a foundation that is required for you to be able to implement hundreds of use cases really, really fast. Because UNS in itself is not a use case. It’s the starting point for implementing use cases. But that foundation is super important.
David (21:46)
Nope.
Martin Thunman (21:56)
And yes, I think we’re driving, you know, the industry is innovating and pushing different vendors are kind of pushing the envelope here, try to get an edge and try to get a little bit better at implementing a kind of a UNS and this common data models. are evolving. I just don’t think it benefits the industry when the concepts
David (22:14)
And, and, yeah.
Martin Thunman (22:24)
and technology are mixed. And this is what we’re seeing from some parts of the industry and some part of our peers, that they’re kind of trying to kind of say that, well, this type of concept requires a certain technology, which we think is not benefiting the end users.
David (22:27)
Yeah. Yeah.
I couldn’t agree more with that statement. And also, if you do so, if you mix, then you don’t really take meaningful steps forward. You might pick a technology and you might implement the use case, but you’re not as an organization, you’re not really taking steps forward. You need to take a step back first and think about what does data management mean for me.
Where does my master data model actually live? All the type of who is going to maintain it, et cetera, et cetera. So there are so many really important questions to answer. And yes, you technology able to support you in those steps, obviously.
Willem (23:33)
I agree with you that of course that data management layer or ULS layer, however you want to call it, is an important foundation. But like you said, a lot of companies are working on that, they’re making steps, they’re making progress. What’s next once they did this, they have use cases working, what’s going to be the next big building block to really work on to make progress there?
Martin Thunman (24:01)
Of course, you know, I’m biased, but of course.
David (24:04)
You can be biased, you can
be biased, please be biased.
Willem (24:06)
That’s okay.
Martin Thunman (24:09)
But as I said initially, we believe the ability to act on the data and be able to build more data-driven automations, that is going to simplify the implementation of a lot of new type of use cases. So once you have that foundation ready and that can be done, then that’s also one of the
one of the reasons why it makes sense to implement this model already in the edge, because if you implement it in the edge, you can take a decision straight away on that data. And you can implement local automations and integrations that go also horizontally. So it doesn’t just go.
in, transform up to cloud, something triggering in the cloud, go back again, and then go horizontal on the shop floor. So I think it makes much more sense from all kinds of reasons to just get the data in, transform, analyze, and take action and create automations straight away on the shop floor in the same instant, in the same millisecond.
David (25:32)
Yeah. And maybe another question from my side. we have our articles on data stuff, which is really cool. But the IT OT insider is the IT OT insider because we also talk about organizational stuff and understanding each other. And is it one or the other? But I’m wondering where do you actually typically sell Crosser to? Is that to IT people or to OT people?
Martin Thunman (25:59)
Yeah, it’s a super good question. I think it’s, well, let me take one step back. I do believe organizations that are able to create a collaboration and preferably a joint organization with OT and IT are the one that will ultimately be able to be more successful. And with success, I mean, they will be able to implement faster and to a much lower total cost. So
So I think ideally we interact with IT that also have a, because IT often owns kind of the top layers, right? And there’s so much momentum right now to move everything upwards. But on the other hand, OT kind of owns the often the final ROI right on the shop floor.
So I think where there can be a combination, think those are the situation where projects move much faster is when there are both stakeholders. There’s both IT and OT leadership on this. sadly, would say we feel that Europe is a little bit behind North America here.
David (27:25)
Okay.
Martin Thunman (27:26)
And I think this is an observation from our side interacting with both on both sides of the Atlantic. European organizations often tend to start with one or two specific use cases and they want to solve individual problems. And then they try to find a problem or solution to that one or two initial use cases. Whereas we see in the U.S. where there is much more of a
David (27:41)
Mm-hmm.
Martin Thunman (27:55)
larger, deeper understanding that, well, data, if I can get a data layer here, an intelligent data layer between my cloud and my production assets, my shop floor, if I can get that, that will open up lot of opportunities for me. So they typically tend to start from a different perspective and which is much more of an architectural perspective than it is use case driven. So.
David (28:21)
Okay, that’s interesting.
Martin Thunman (28:24)
So that’s what we see. And the organizations that have the capabilities or ability to think much more of a long-term vision-based, they’re the ones that understand, well, maybe there will not be an ROI in use case one or two. But when we have implemented five, six, or 50 or 60 use cases, we will have an enormous ROI on this investment. And so that’s where some of the leadership in Europe
needs to step up and change your thinking, in my opinion.
Willem (28:59)
Where does that difference stem from you think? Is it a cultural thing? Is it?
Martin Thunman (29:04)
There are different studies around this, I think the difference between US enterprises and European enterprises is that the US companies have twice as large IT budget.
That’s about what they have. So it’s just a deeper understanding that IT and technology leads to more efficiency and leads to benefits and business. could be competitive benefits or speed, cost reductions, what have you, process improvements. There’s just an understanding about that. So I think that is where European companies, unfortunately,
still are a little bit lacking and they need to step up their game to be able to compete in the long run.
Willem (30:01)
It’s because they’re, again, it’s all speculating, but would you say that then those companies that do have this big vision, let’s say, bit more long-term view, they’re the ones who dare to take the leap and say, we believe that this digitalization will bring a lot of value and you cannot start it with one or two use cases hoping to get the right ROI and build slowly upon that.
Martin Thunman (30:26)
Yeah, no, think that’s definitely a different attitude, right? That they’re walking into the project with different touch. But all companies are not equal. There are leading companies in Europe that also do like this. So one of our public references is Tesa, which is a German company, of course, in adhesive tape. It’s a division within the Baierstaff Group.
And TESA implemented across all of their factory worldwide, including China, in two months’ time. They rolled up a large number of use cases. They had a clear vision of what they wanted to do and did that very, very successfully. So there are absolutely examples like that. there are many others as well. But as a group, there is a big difference.
Willem (31:20)
You were also mentioning that you believe that the companies that managed to bring IT and OT working together will be the ones that will be succeeding more in this market. What are you seeing are the catalysts to enable that corporation? Or on the other side, what are the blocking points that you see coming up across different customers?
to get that joint cooperation working.
Martin Thunman (31:53)
You know, are Warren Buffett, you know, I believe said at one point, if you show me a company’s incentive structure, will show you their future results. And it’s a little bit like that. And it’s often tied to how the budget is distributed. So companies that have distributed budgets where the sites have their own kind of
control over the budget. In those instances, it’s much more difficult for organizations to create a central joint unit and then make central decision and roll out. They need to then kind of, even if they centrally developed a concept and if they have decentralized budget, they need to convince the local plant manager to…
to kind of deploy because they don’t control the central budget. So how the budget is distributed and how these plant managers are incentivized, those are very, important things that are there.
Willem (33:06)
Follow the money.
David (33:08)
Follow the money. Well, that’s been super, super interesting. that’s a wrap, I guess, for this episode of the IT OT Insider podcast. So where we explored how to make industrial data work for us, a very big thank you to you, Martin, for sharing your insights and to you, our listeners.
Martin Thunman (33:09)
A little bit, yeah.
David (33:31)
for tuning in. If you enjoyed the conversation, don’t forget to subscribe at itotinsider.com and leave a rating because that really helps us. And I’ll see you next time for more insights on bridging IT and OT. And until then, take care. Bye bye.