May 1, 2019
| 3 min read

Podcast #56: A Digital Twin for Business Process – A Conversation with Babur Ozden

Babur Ozden is Founder and CEO of Maana, a software based Knowledge Platform. Our conversation explored the origins of the company, which focuses on capturing best practices and tacit knowledge of business processes to provide employees with domain-specific software assistance created through extensive analysis of historical data. He discusses the challenges of deriving actionable knowledge from silo-ed data, and the principles behind the creation of the Maana Computational Knowledge Graph that powers the company’s solutions. He also shares why and how the company turned its focused to the energy industry and provided examples of successful use cases of the platform, shared insights around competition, risks and vision for the future.   

 

orange-line.png

We'll notify you bi-weekly about new podcast episodes, upcoming guests, and news. You can subscribe to the podcast and if you'd like to be considered to appear on the podcast contact us.

 

View Transcript

Good day everyone, and welcome to another episode of Momenta Podcasts, and today our guest is a Disruptor, we have Babur Ozden who’s a Founder and CEO of Maana, which we’ll get into a bit of what they do, but the reason that I had reached out to him in the first place was that we had seen Maana appear a couple of times in our Digital Leadership Survey, when we had asked our respondents what were some interesting companies that were doing work that we should pay attention to; and when you see something pop up a couple of times you definitely want to learn more. So, I reached out to Babur and I asked him to join us to talk a bit about what he’s doing and share a bit of his vision.  

So, Babur it’s great to speak to you. 

Hi Ed, it’s a great pleasure to be here with you. 

Terrific, let’s start off with a bit of your background, could you share what had brought you to where you are today. 

I’m a computer scientist by trade, I studied computer science at University of Texas, Austin. Right out of school one thing led to another, I always found myself working for start-ups; either I started or someone that I know very closely started. So, it’s been over 20-years and spread to six start-ups on various industries, but at the core its always an innovative service or product with a unique application of latest advances in software. 

Could you talk about the business problem that you see, that has led you to find your company? 

As you know, we are in the space of industrial digital transformation, we help our clients and we help industries in general to go to a post-digital state. That journey for each company is actually a roadmap of hundreds of new use cases, so they will develop, test, deploy and maintain hundreds of use cases, and new use cases, in the course of three to six years. So, the business need and the business problem is the speed at which these use cases could be both strapped from data, from tacit knowhow, and put in use. So, we see digital transformation as a business problem that companies need to develop hundreds of use cases in the course of a few years. The need is, can you accelerate the development of that many use cases, with as little human need as possible, as well as creating a new type of data from these new use cases, stored and consumed in a very different fashion than the traditional data siloing approaches. 

That’s really interesting because we’ve been having a lot of conversations about digital transformation, and one of the common themes is that every project is really a bespoke undertaking. We have not yet reached a level of maturity that has led to being able to have replicable templates easily, and it sounds like that’s the direction that you’re going. Could you talk about how you think about that? 

If you look at a very large industrial organization, their ambition is to achieve some business returns measured in multi-billion dollars in annual scale, at the end of this journey. Of course, there is not a single silver-bullet use case for you to do that, there’s not even three or five use cases, if you look at the half-year to a year, they study how they would transform themselves, it’s a series of hundreds of use cases spread across the entire enterprise. For example, in the oil and gas space, in a single company that would be dozens of use cases, in each of the exploration department, drilling department, production, refining, shipping, trading etc. So, the ability to build this many use cases, and when its original data that’s going to bootstrap these use cases, coming from hundreds if not thousands of different data silos behind an organization firewall, it is extremely important not to repeat the old habit of siloing in this new world.  

Siloing your data first inhibits one of the foundational reasons people do digitization, second it slows down the process of developing things. We believe, and many of our clients choose Maana because they share our vision, is that digitization, although you do hundreds of use cases, it’s not the same old way of doing use cases, so it’s not more data churned by more software. There’s something fundamentally different in this digital world because we attempt to deploy a new category of algorithm in the industrial world, called machine-learning category, the machine-learning is at its best when its tackling data without silos. If you look at pure digital companies like Uber, or Amazon.com, and the word we believe digital becomes because of that, is that these companies have no data silos, everything is in one place in one shape, and so every local decision therefore could be globally optimized.  

So, the whole pursuit as we see it, is, use these digital use cases to bootstrap this singular place where the companies no longer do these cases in one-off applications, as well as the way that the data they generate doesn’t create silos anymore. You see this trend Ed, you guys follow this very closely, the natural place to do digitization is on a cloud, it is far more cost-efficient there to be able to do that. Secondly, you see that the company that’s embarking on digital transformation, once it picks its cloud or multiple cloud providers, but it picks the cloud, to host these digital ambitions, the next thing they do is create a data platform; this data platform brings data from hundreds, if not thousands of different data sources across the enterprise into a single place, so the very first act is de-siloing data. That’s making data available to these new use cases, and our functional role in this digital stack is to make that data which is now de-siloed, now useful. So, we provide another layer of de-siloing the knowledge potential in that data, so we add another layer of de-siloing on top of it. 

So, essentially what you’re doing is, you act to extract the appropriate data, and then be able to combine data from different sources in a context and a structure, that allows the business value to be much more easily accessed to outsiders that weren’t part of generating the data in the first place. 

That’s correct, allow me just to qualify a little bit. When a large organization de-silos data into their data platform, in their digital stack, which is the foundational layer of any digital stack, is that any functional layer on top of it should not copy the data anymore. The data sits where it is, and knowledge – if you could take the dictionary term of knowledge, that’s how we refer to it, and mathematically how we define it, is that the knowledge is when data is doing something. So, the data itself sits in the data platform, the knowledge is the mathematical modelling of answering a specific question. So, our computational knowledge is basically at the moment of your question, such as, ‘Where should I drill the next well?’ ‘How long will it take to drill this?’ ‘Which type of crude should I process this week in my refinery?’ so, the knowledge layer that we provide uses the data in the data platform, which is already de-siloed, to specifically answer these questions. 

I’d be interested to learn a bit more about how you decide and define the knowledge that you are essentially effectively generating from this raw data, and if we go back a bit to put some context on it, the whole concept of knowledge management as a category of software goes back a few decades, this concept of capturing the tacit knowledge that resides in an organization; there have been a number of different approaches in many regards, I remember earlier-on you had companies that were in the search market for instance would just use indexing technologies as a way to capture, and it was very much a brute force approach, and a lot of that technology got subsumed into other solutions. I’d love to get more of a sense of the approach you’re applying to generate knowledge from data, in a contemporary fashion. 

We approach from a very operational angle; if you take a large oil and gas company and across its enterprise it has about 500 major operational decision points, where every day several 10,000 people at these operational decision points would make 100,000-200,000 decisions. So, the use cases that you would see in a digital road map involves, if not all, then most of these decision points, decision improvements. So, how can we use our data traditionally that you see so many different places, and bring it about to answer a specific operational question? These questions are almost always the same which people ask over and over, so each operational decision point is a digital use case, and each defines a problem space, if you may say a search space.  

So, a decision point would be to answer a very specific question, ‘Which oil wells should I abandon this month?’ So, this is no different than asking Google, ‘What can I eat today?’ because that question is submitted to Google several million times every year across the world, so answering that specific question is in a computational knowledge graph in Google. It’s not a random query because they know that query is going to come, and they know about you, they may figure out where you are, they may know your eating habits, dietary habits, and based on that they will attempt to answer or make a recommendation to you. Of course, in the industrial world the world is not that simple, but it is still a fixed problem domain which could be mathematically modelled. 

That makes a lot of sense. Let’s talk about your approach to the software that you’ve created and talk about some of the components to the software, and ultimately what are the outcomes and the experience that your users have by employing and using your technology. 

We call our product Maana Knowledge Platform, it’s very high-level, it provides its end-users to build this knowledge infrastructure, and using that infrastructure to build these use cases which we call knowledge applications. Each knowledge application is a recommendation app to a specific work decision workflow, and the end. The end users are a group of people made out of subject matter experts, data engineers, or data scientists, one or two developers, it’s almost always five to ten people of different backgrounds coming together to write a knowledge application, to improve decision-making at a fixed operational workflow. So, we provide a tooling around the core of our secret source, a tooling for this group of people working on the same application, and to develop a use case. So, there’s development tools as well as knowledge infrastructure building tools. 

We borrowed a terminology from the industrial IoT world which we have nothing to do with, it’s a digital twinning. In the IoT world, companies like ABB who was one of your guests, I’ve listened to his podcast, they built digital twins of the physical assets. So, by looking at the history of the asset, the digital twin could at any given time predict the future behavior. In our world what we do is, we use our technology to provide a digital twinning of a business operation which may or may not have assets in it. So, it is that operational workflow, then you write these applications to improve decision-making. 

I think that’s a really clear metaphor, this concept that digital twinning really helps to visualize the process that you’re undertaking, and you’ve developed some unique IP around this as well. I’d be interested to learn a bit more about the mono-computational knowledge graph, could you just describe a bit of that and how it works? 

Yes, as it says, it’s computational, so it is not a graph database because data sits in a data platform, on a data lake or a graph database. So, it’s about storing computations about models, to answer a question. A question is subdivided into dozens, sometimes hundreds of smaller questions, whose answers eventually lead to make a recommendation, and so the computations about all of that is networked into around the data, related and relevant to compute. So, we built a graph representation of how a group of human minds think at that operational point. Then it becomes another colleague of real subject-matter experts working with them to answer that question. 

What you’ve described is in many regards intelligence augmentation, very much focused around specific processes. 

That’s correct, some of the big consultancy firms like our partner Accenture, I believe is the name father called Augmented Intelligence. That’s exactly our platform which allows its end-users to build this augmentation, to help them make better decisions. They’re actually heading another colleague, but this one happens to be living in software and data, but it works with them to answer that question. So, what they’re doing is, in a very unique way our interface works with them, they’re not even aware of the mathematics etc. behind it but they pass through the system, their mono-system, how they would be thinking in a situation at hand, and then it gets modelled. So, you’ve got all these alternative ways of how people tacitly may respond, as well as whether that approach by human thinking has proven to be good or bad. All the history is in the data, so that you enrich and provide another point of view to the tacit knowledge by machine learning from data, then you blend those together. So, you’ve got the people experience, and actually what’s really happened coming out of data, and you make a more experienced aid which is now working side-by-side with the subject matter experts. 

I think you’ve really outlined the vision of what a lot of people see in many regards as the future of work, which is highly trained people working with specialized computerized assistance. I’d love to go back to the example of freestyle chess, which is these leagues of chess experts using computer augmentation, or computer assistance to compete against each other, and apparently the combination of human and AI, or we’ll say applied AI, has been beating both the fully computerized Deep Blue type programs, as well as humans themselves, so to get to a better outcome than either relying on the black box, or the human robot. 

That’s correct. We belong to a group of companies, small and big, trying to find how to make this. Because at the end of the day if this is possible, if a large organization have several hundred very important operational decision points, where several thousand people under employment every day around the world takes tens of thousands of decisions, and 10-15 percent improvement in such decisions is worth trillions of dollars within a single industry, annually. But to build this, the only way this future is possible is if systems like these, these augmented capabilities could only be built by the modelers which are the subject matter themselves. If you wait for highly specialized experts, scientists and deep architects to come and build these point by point, it is an unattainable work. So, the idea here, can a software platform like Maana make that gap disappear, for the subject matter experts working with their data science, or algorithmic scientist clients, then they can collectively build this. 

So, if you only rely on scientists, this will never work, because they won’t scale, nor their engineers, nor do they have full understanding of the business problem. So, here is the key, even developing these applications, not only if they’re the end consumer, even developing, the key here is the expertise in the minds of the people in the company’s employee base. We regard to that knowhow far more important than petabytes of data. 

I think it's telling that you’re bringing up the role of subject matter experts, and what I wanted to dig into is the focus on the energy industry, the energy vertical as, I would say, the initial industry you’re focusing on. Could you talk about the specific problems in that industry and what led you to pursue the problems in this industry, at least as your initial focus? 

 in an asymmetrically different way than the rest of our pursuits.  

As a smaller company, even smaller then, we made a fateful decision to focus, despite the fact that none of us knew anything about that industry. So, we made a lot of bad decisions in Maana, but we found the best first decisions was that focus, and that once we focused we saw within that industry a single company could have nine different industries, trading is different to drilling, exploration is different to producing it, so that provided Maana the opportunity to verify our technical claims that our platform could be agnostic to the kind of decisions that are being made on it. So, an oil and gas company could actually provide Maana, so you see our product used by traders, you see it used by drillers, you see it used by health & safety environment specialists, you see it used by duo physicists, so that was a very important verification by us to see if we can really be agnostic to these vertical decisions. So, we don’t see the world as industrial verticals, we see them as decision verticals. 

That’s a great insight, and it would seem that that does lay the foundation for a lot of the successes you’ve built so far, the successful implementations, to apply to different industries or different verticals. 

Again, the claim we provide to our customers is that once they’ve built their digital technologies stack, and almost always on cloud, and for reasons almost always on Azure, and that’s not a marketing statement for Microsoft, but that’s a reality; so we then come and provide them, ‘Okay, you’ve done the hardest part, you’ve de-siloed the data that you have been siloing the last three decades, now let's give you some computational capability to extract knowledge, without moving your data or creating new silos’, so this then becomes your operational decision point. So, the companies start seeing that having a knowledge layer in their digital stack after they completed their data platform foundation, would actually accelerate the development of those hundreds of use cases. So, they start coming and seeing it, and that started I guess our positive reputation within the energy industry. 

Could you talk about some of the success that you’ve had so far, or any examples of uses that have even surprised you, from your clients? 

Some of these use cases, for the first time we hear our product is used in a decision environment, so I’ll give you a number of them as very interesting, because these are problem points, that if you do a small improvement you actually help human civilization, so we take pride in that.  

One would be in the world of shipping, this is when a vessel learns that it cannot make a scheduled port call, and it has a number of fixed hours to find an alternative port and reroute its logistics. If you think of 70 percent of the world’s trade is carried in maritime around the world, and if you could help some of the major players of that industry to assess and remedy this problem, you’re actually significantly contributing. So, that’s one area that our product has been used. 

Another is reduction in parts sent to the field. Large companies send parts to their assets in the field to do the initial installation, annual or regular maintenance, inspections, repairs, etc., in some heavy industries one out of three parts sent were never used, others significant losses; so can you improve that, can you help a field technician that has just received an electronic ticket from a service center to make the right parts selection, so the part waste will be reduced? 

Another is to help people who are taking over shifts in large industrial operations, a day shift, a week shift, or if it’s an oil rig it’s a shift change every three weeks, so that when a new shift is coming in and taking over from a departing shift, how can you inform the new shift of all the health, safety, and environment risk they may be facing, so that the operation during that shift doesn’t have any incidents? So, it’s a wide-range of things. 

That’s great. Who would you consider your competition, or what alternatives do you face when you’re working with a potential customer? 

Our biggest competition Ed, as you mentioned earlier in our conversation, is the traditional instinct to solve these problems as a one-off, so let’s go tackle this use case in this fashion, they’re all in your road map, you’ve got to do these use cases, but what you do is you either tackle them one-off, or you give a little bit of some of them to some vendors, a few to another vendor, and everybody does them in their own silos. So, that traditional approach to doing this many use cases in a very large organizations is our number one competition. The second thing is the competition from the cloud providers, similar sounding capabilities. 

I would love to quickly ask you about the business model, and what type of a model do you approach, how do your customers pay for the software? 

We assume that a customer or a prospect customer we pursue to become the vendor, will have at least dozens, if not hundreds of use cases in their digital road map, we would always assume half to a third of them would be an extremely good fit to tackle with Maana. So, our business engagement model assumes that fact that it’s not a one-off case, or a two-off, it’s a sequence of use cases, the client will use Maana to address, develop and solve. So, we give our customers an annual development license so they can develop these use cases, it’s enterprise-wide, there’s no limit on how many people can use it or how many use cases they could develop. Then we partner our customers’ success in the digital transformation, and then we have a production version of our software because the use cases developed with Maana in day-to-day operations needs to run on Maana, so there’s a production version, but instead of charging one big production license, we charge per use case production license, and per use case that actually is in production. 

The time it takes to develop something, pilot, test, and go to production, is usually a several months’ gap. So people pay us as their used cases see day-to-day operation and production. 

As you look forward, in your mind what are some challenges that you’re focused on, and what are you most optimistic about, looking forward? 

The single biggest challenge I think we face as a small company, growing, but it’s still small, and as you know we’re focused on very large industrial players, so our sales cycles are long. It takes about a year from the very first meeting to selling your first enterprise licence, and in that one year you have your sales, pre-sales, maybe a proof of concept test, so our biggest challenge right now, or opportunity, is, can we start to shrink the sales cycle? We see that taking place as more and more best practices are taking place in this digital transformation journeys, and people are abandoning less and less POC approach and going to do the real thing, since there’s some experiment. But as a company my biggest challenge is the sales cycle. 

Secondly is the talent needed to grow my company, so that remains to be a significant challenge for us from day one, but we manage to address it. So, what we did on the talent side, we’re a Silicon Valley based company, and another very important fateful decision we took is, we spread our organization nationwide, and worldwide. So, today all our engineering that develops our product is now in Bellevue, Washington, and all our customer facing engineering is in Houston, Texas, and in London, and in in Dhahran, Saudi Arabia. We have addressed finding the right talent problem on the US West Coast, by going elsewhere, where equally or maybe better talent exists, and that’s been working really well for us. 

And the question about what you’re most optimistic about? 

I think in general, the industry is awakening to make use of their data. Truly, data was something generated and stored, and a group of people tucked away somewhere in an organization looked at it if you wanted some deep analysis. So, if this digitization succeeds, all of these augmented AI capabilities, or AI capabilities should be as simple as bringing a solution in an excel worksheet to an end-use. So, I think that’s my great optimism, I see that is most likely going to happen, and if it can’t happen this time it will be postponed for another 40 years probably! 

It’s been terrific to learn more about your vision and the work the company is doing. I always find it fascinating to solve longstanding business problems, and just picking away at ways to solve problems and create new solutions. With that, we’ll wrap things up; again its been Ed Maguire, Insights partner at Momenta with Babur Ozden who is CEO and Co-founder of Maana. Babur, thank you so much for the time, it was an absolute pleasure talking to you. 

It’s been my pleasure Ed, thank you for having me. 

[End]