How can data science help organizations drive effective digital transformation? A prominent data scientist explains weaknesses in many digital transformation strategy efforts and how data can address the problems.

Dr. Satyam Priyadarshy is a Technology Fellow and Chief Data Scientist at Halliburton (NYSE: HAL). He is also a Senior Fellow of the International Cyber Security Center at George Mason University.

Transcript

This transcript was lightly edited.

Introduction

Michael Krigsman: Did you ever think about the relationship between data science and digital transformation? Dr. Satyam Priyadarshy, tell us about Halliburton and tell us about your role there.

Dr. Satyam Priyadarshy: Halliburton was founded in 1919, so it's over 100 years old. Last year was our 100-year anniversary. It is one of the world's largest providers of products and services in the energy industry. We have roughly about 50,000 people in 80 countries where we have roughly 140 nationalities.

The company helps to maximize the value for our partners across the industry for the lifecycle of the reservoir, as we call it, in locating the hydrocarbons, then managing the production, drilling, and formation. This is the core business and we have been in this business for, as I said, over 100 years.

Michael Krigsman: Give us some insight into what you do as the chief data scientist at Halliburton.

Dr. Satyam Priyadarshy: As we say, data is an important asset for any company. Digital is what enables it to create value. In order to create value from the data is where data science is needed. In 2014, Halliburton was forward-thinking and, when they asked me to join from outside the industry—I joined the oil and gas industry as the chief data scientist—my goal was to build the practice around data, how to leverage data, and create value by leveraging data, which is the historical data, the new data that will come forward, and set up the center of excellence, we call it, with data, data science, and digital center of excellence.

Now, if you think of a center of excellence, most people will think that it does one function but, especially in an industry like that, you have to do a lot of things. One is data science and data itself. Creating value from data is not easy. Most people think just having data, building some dashboards is the way to create value but, actually, data science takes it one step further because the goal is to improve the business on an ongoing basis.

As a function of data science, you have to not just do the value from data but also build a team around it. We have been very successful in building one of the best data science teams in the industry.

Digital transformation challenges and pitfalls

Michael Krigsman: What are some of the challenges of digital transformation? Where does digital transformation tend to fall short?

Dr. Satyam Priyadarshy: One is to understand what digital transformation is. This is a term that people use very loosely in many cases, but digital transformation is about leveraging the digital technologies, digital assets on a continuous basis to grow the business and remain competitive. If you understand that part of the definition, then you will know that it is a step-by-step process in creating economic value.

What do you do? Digital transformation happens when you have digitized or what is called digitization and then digitalization. Then you have a digital transformation in terms of economic value. That means all this is based on data.

What are we digitizing? That is the data, the forms, the data that you collect. Once you have that in place, then you can actually start creating value. Then you create value in a step-by-step fashion.

What happens as well, as you asked the question, whether digital transformation will fail or why it fails? Because people are trying to jump too fast without understanding the full flow of the process. You should know how to leverage the data, how to create the value from it, and then how to actually implement it at the field level.

See, unlike the marketing domain or a retail space, the oil and gas industry is much more complex. You can build, so to say, going into details later on, AI or something, but let's say you have the model. You build the model on the historical data but you have to take it to the field.

What is the process of taking from the lab to the field? That has to be well established and tested out. When you don't take time to actually do this in a systematic manner, that's when most digital transformation projects fail.

The second reason why it fails is, there is a term that I used called the “digital divide.” That means how people understand what digital transformation is at the leadership level, at the operational level, at the actual management level, and the value of creating value from the data and the opportunities associated with it. That understanding is different. That digital divide causes this digital transformation to fail.

How to use data science for digital transformation?

Michael Krigsman: How do we then use data science in the service of digital transformation? Quite frankly, when most people talk about digital transformation, they're talking about things like business models and culture change. The subject of data science doesn't usually come up as one of the top items.

Dr. Satyam Priyadarshy: That's surprising, those who say that, because if you don't do science on the data then how do you actually leverage and build a digital transformation strategy? Yes, the cultural change is an important part. The technology is important. The people part is very, very important. But the actual doing science on the data is very important.

Most of the bigger industries are based on engineering and scientific formulas, so people have been using them for a long time. But now, we have years and decades of data that has been collected for every workflow. That can be looked at in an effective sense and learn from it what gaps have been there, what I call hidden inefficiencies.

Data science is actually very powerful in finding those hidden inefficiencies because if the efficiencies are not hidden, then you must have fixed it. If you have not fixed them, then it's a different problem at the management level. Most of it is to find hidden inefficiencies and, once you find them, to actually figure out a way to incorporate those fixes and then actually create low-cost solutions for yourself so you can actually grow the business in terms of revenues and profits.

Michael Krigsman: What's the relationship then between these hidden inefficiencies that data science can uncover and digital transformation?

Dr. Satyam Priyadarshy: When you have found some hidden inefficiencies, you are actually either optimizing or maximizing your workflow. That workflow change is what leads to the transformation on how you do your business, whether it is customer service, whether it's equipment manufacturing, whether it's testing equipment, or looking at the production side. Then you can actually do this in a continuous form. The more hidden inefficiencies you find, the more transformational things you can implement, and then you can build a highly optimized system across the ecosystem.

What is dark data?

Michael Krigsman: You talk about dark data. Weave that into the story as well, please.

Dr. Satyam Priyadarshy: I give an analogy. All of us have very, very high resolution, high-end smartphones but if I asked you to find a photo of yourself which was taken five years ago, how much time would you take to find it? That data is there, but it's dark. You can't find it.

Imagine now a corporation of any size. The data is actually not easy to find. You know the search engine technology is pretty mature, but you tell me how many big, good corporations are there where you can actually go and type the model number of a, say, centrifugal pump and it can give you the history of it. The data exists either in your databases and data warehouses, on paper, or in somebody's drawer, but the data is not creating value.

The data that is there is actually expensive because it is costing you to keep it. It's not free. No storage is free, whether it's cheaper or not. But when you don't create the value, that data is called dark data.

Michael Krigsman: What do we do about that dark data? Again, what does that have to do with digital transformation?

Dr. Satyam Priyadarshy: The dark data has a lot of learnings in it. Look at a process or a workflow over the last ten years, because many of the industries actually do the same workflows over and over again and things break down. Of course, when you're operationally active, you will actually certainly go, fix it, and move on. The learnings are actually hidden in that data.

The goal is to leverage that dark data and convert it eventually into what we call smart data. What does smart data mean? That means the data can be acted in real-time.

If you are going towards automation in terms of what we call industry 4.0, that means we want to automate everything. You cannot automate and build out systems unless your data is smart. That means you can act on it either through machine learning, artificial intelligence, data science, or any other method that can be acted on. You take that dark data, you convert that to smart data from your learnings by applying data science, and then you transform your workflows.

Convert dark data to smart data?

Michael Krigsman: How do you convert that dark data into smart data?

Dr. Satyam Priyadarshy: It's a process and I've written quite a bit about it. You first start with a small set of dark data. You create some value. You see what's valuable in it, what are the missing gaps in it, and then you start trying to integrate data from multiple silos.

As we all know, most companies or most organizations have their data in multiple silos. When you start connecting them, you build a scalable system. Then you eventually test it out in real-time. Once that real-time feedback is put into where that data ingestion is taking place, at the ingestion point or the generation point, you fix all the issues that are there. Then you actually have smart data.

Michael Krigsman: We have a question from Twitter. Arsalan Khan asks, "Have you been in situations where the data that was collected ended up not having the value?" You've gone through this dark data. It doesn't have value. I guess the more general point therefore is, it's obviously not practical to convert all the data that exists from dark data into smart data, so how do you make those selections?

Dr. Satyam Priyadarshy: As a scientist, you know that every failed experiment is also valuable. Every data has value. The question is how much value it is for you or for that business or for that process.

One has to look at the data. Sometimes people say, "Oh, I collected the data. It has no value there." But people haven't actually even looked at it.

If there is no value, then why did you start collecting it? Why are you spending your capital and operational expense on it? Correct?

The question is to quickly do a phase study. Assess how much value it is. Out of that data, how many elements of that data are really valuable? If this may not have value even in the future, you don't want to probably collect or even generate that data. But if you are generating any data, there has to be some value. Otherwise, we probably started on the wrong foot.

Using data in digital transformation strategy

Michael Krigsman: From a digital transformation standpoint, does that mean that the thinking through of the type of data that you need needs to be part of the strategy, the digital transformation strategy right at the beginning?

Dr. Satyam Priyadarshy: Absolutely. There are four pillars of digital transformation, as I call it. One is what I call the big data analytics and big data doesn't mean just large amounts of data. Small data is a part of it because it is the full continuum of data that we are talking about.

Now the second part is to leverage the computing paradigms that are existing or we call emerging technology paradigms. The third one is to build real-time systems and then the fourth one is autonomous. This is what whole digital transformation is. The foundation lies in the data.

Michael Krigsman: Can you give us any examples? I think this is an area where folks have trouble. As we said before, people tend to think about business models but I think the data aspect can be very challenging for organizations.

Dr. Satyam Priyadarshy: One of the simplest example that I typically give about digital transformation is from the music industry. All the baby boomers and people of my age group would remember that the songs were digitized and put on LP records. Many of the younger generation probably have no idea what they are, but those LP records also required a very big box on which they were played. You think of it that it was sort of digitized data, but it was not very easy to share. If you wanted to share songs with friends and family, you had to invite them to your house.

Then a transformation took place in terms of CD and MP3. It was slightly easier to share the songs because you could borrow a CD and play and then give it back, and then MP3, like iPods and things, that came.

Look at today. The whole music industry is all streaming music. I can actually listen to songs that I like and I can share with Michael my album without actually sending any songs to him. This is called a real digital transformation.

The song remains the same. The data remains the same. The way it has been delivered by leveraging the emerging technology and the data science done on it—what preference of songs, who is liking it and actively shared across the world—is transformation of the industry.

The economic value, as we all know from the music industry in the olden days, is only five, maybe ten record houses and ten artists made money. Today, there are millions and millions of people who are making music, enjoying the music and, actually, the economic value is significant in it. As we all know, it generated a large number of jobs that never existed. This digital transformation is all about creating new kinds of jobs as well as a part of the whole exercise.

Michael Krigsman: That's so interesting. If we think about Spotify or Apple Music, any of these, in fact, what's going on is the data is stored in their repository and they're letting you share the metadata.

Dr. Satyam Priyadarshy: Absolutely. I stay away from the technical term of metadata. I call it the knowledge. I created the album with human augmentation of what I like. Then I can share it, so what we are sharing is the knowledge on the data, that, "Oh, this is a beautiful album. You should listen to it."

Think of many industries. If they really are serious about transforming themselves, then they should be really elaborating and co-innovating solutions where they can actually share the knowledge so that the cost of delivering services can be reduced across the industry. That becomes very important in the energy sector anyway.

Metadata vs. knowledge

Michael Krigsman: I'm quite interested in this topic. You say you prefer the term "knowledge" rather than the term "metadata." Can you explain why? Why do you talk about it that way?

Dr. Satyam Priyadarshy: Because metadata has a connotation in our industry. It's something like what you call how you tag the data, right? But what I'm saying is knowledge that is built on top of the data.

An album is the knowledge, so to say, whereas metadata would be the artist's name, whatever is the genre of the music. That would be the metadata.

Metadata, if I were to buy a mechanical pump and put it in my SAP system, then the purchase order is metadata. The history of the pump that has failed many times, I analyze that failure and create a root cause analysis of it and a regression model on it, then that becomes actually your knowledge.

Michael Krigsman: The knowledge is the information that's really useful for businesspeople in terms of their processes and their shared experiences? Would that be an accurate way of putting it?

Dr. Satyam Priyadarshy: Knowledge is what we actually incorporate wisdom, if you think of it. Knowledge is generated by data and metadata together. Then you have knowledge. Then the wisdom part comes when all of us come together.

Michael Krigsman: Again, link it back to transformation, to digital transformation for us.

Dr. Satyam Priyadarshy: Exactly. Once you have the sharing of this knowledge, it becomes easier to transform, industry-wide or in the ecosystem, your workflows, your business models, your strategy, your future competition. That is very important, especially if we look at an energy sector.

I'm sure you must have heard about this term called energy transition. There is a lot of pressure on zero carbon, negative carbon, and things like that. But what is that?

We are building a landscape now of energy. You have to be a player in it. If one sector doesn't transform and keep up pace with the others, then you will have challenges. This is where the transformation comes into play.

Discovering the value of dak data

Michael Krigsman: Sarada asks, "How do you find the value of dark data? The data is dark and we may not know anything more about it."

Dr. Satyam Priyadarshy: The dark data is created by us, so the value is there. You have to pull it out and do some analysis in what I call proof of value projects. That's how you get it.

Why is it called dark data? Because human beings have created that data, set them under the table, and not shared with anybody. That's why it's dark data.

You can do projects. You can do very basic analysis. You can do simple, visual analytics. You don't have to go and run after deep learning for everything. If you see limited value then you can build more expensive and more compute-intensive models to actually integrate data to create more value.

Michael Krigsman: It always starts with some type of proof of concept to see what the value of that data really is.

Dr. Satyam Priyadarshy: I stay away from that term also, "proof of concept," because concepts are well proven. Artificial intelligence, data science is very well-proven. The engineering process are well proven, so to say.

What we are trying to create is value. Value is the most important part for the business. We call it proof of value.

Michael Krigsman: That's always, at each step, your metric, which is the question. What kind of value are we creating and for who, as well, I'm assuming?

Dr. Satyam Priyadarshy: It's the business, so we are talking about business. If you are doing an academic exercise, that's a different thing. You can write as many papers as you want on the same problem. But in business, you have to actually produce something which has value.

Michael Krigsman: We have another question from Twitter. "Should your data strategy plan for what data you should keep and not bother with storing the rest?" Do you just figure out what you're going to keep and you throw away the rest or do you do something different?

Dr. Satyam Priyadarshy: … (indiscernible, 00:20:36) every raw data is very important. The raw data is the only single source of truth.

As a data scientist and with a career of over 20 years, I believe every piece of raw data is valuable. If anything has been modified, changed one single bit, that's not raw data. That is subject to checks and balances.

If you are generating any data, you should keep it. You never know when it's needed. The goal is not that everything is to be live. The goal is to have it easily accessible so you can always build models and check it back that your model is working under certain … (indiscernible, 00:21:19) conditions back to what it should … (indiscernible, 00:21:22). Just throwing the data without knowing why you threw it or without looking at the value of the data, that doesn't make sense.

Michael Krigsman: We have another question from Twitter, another one from Arsalan Khan. This is a really good one. He asks, "Since data touches every part of the organization, how do you create sandboxes for people who are not in technology to play and experiment with that data and come up with their own thoughts on what kind of data should be collected?"

Dr. Satyam Priyadarshy: Every organization has a culture around it. I think it's a lot of what I call education needed both at the leadership level, at the mid-management level, and individual contributors level, so to say as to why somebody wants to play with this data and what they are going to do. Good governance, good practices about cybersecurity are in place and how this data will be actually used by the people or the group that is trying to look into it, that has to be really sort of codified. Then, actually, it's much easier to create value from that data. That's the first part.

Blocking from the data just for the sake of locking down is not a way to create value. But, at the same time, having the right governance structure around it is very, very important and, of course, requires a cultural change, as you said in the very beginning that digital transformation has a big cultural change component. Creating value from data by anyone requires a cultural change in the organization.

Digital transformation and culture change

Michael Krigsman: Would you elaborate on that? Why does creating value from data require a cultural change?

Dr. Satyam Priyadarshy: If you look at the history—and before I came to the oil and gas industry, I've probably consulted for many industry verticals—it is just like people have been locking down the data and that is the biggest part.

Say, for example, the marketing people have their own data. The sales guys have their own data. Then if you actually add up, sometimes they don't add up to the value that the CEO is looking at. Even if you were to look at sales organizations in some places and having done this exercise a few times, the north region versus the south region plus the eastern region plus the western region, if you add up the numbers, sometimes they don't sync to one.

My goal as a data person, I'm looking at the starting point and the endpoint. They should all connect in some way to actually be coherent. That requires cultural change because people saying, "Oh, my region has been doing great. Yours has not been doing great." That's why it requires cultural change. New thinking is needed in this if we want to seriously adopt digital transformation.

Michael Krigsman: I'm really hung up on this phase that creating value from data requires a cultural change. Wouldn't it be possible to gain value from data? That is basically what you've been doing and there's no cultural change required.

Dr. Satyam Priyadarshy: It's not possible. If their organization was already what I call a data native company, that's a different story, like the Internet companies are purely data native companies. They survive and thrive on data, so their culture is already set.

Let's look at other industries, which are implementing things like, say, IoT sensors and newer technology devices. How do they actually build that culture? "Why should I look at data in the first place rather than just by inclusion? I've been doing this by the same process for the last 20 years. I know how to fix it."

You can see the example of the healthcare industry. They have been integrating data from multiple sources. Now EHRs or EMRs are available. In some places, you can actually look at their reports and things like that. That culture change has taken place. Five years back, it was not the case. Now we are talking in the healthcare about precision medicine.

Michael Krigsman: It's for this reason of what you're just describing that data is at the heart of digital transformation and sometimes we hear the phrase, "Data is the new oil."

Dr. Satyam Priyadarshy: I don't like to call that myself, but data has been there forever. It's an asset, according to me. It's a very expensive asset that we all have because data was generated over the years. So much money was spent in generating that data, collecting it, storing it, putting in data models, putting in databases, data warehouses, so that is a cost to the company.

Michael Krigsman: Could you explain the bias versus variance tradeoff? First, you have to tell us what is the bias versus variance tradeoff. [Laughter]

Dr. Satyam Priyadarshy: Bias is when I look at some data. I usually use a picture about these kinds of things. In fact, you can actually look at my background and there is a nice picture. If you have a biased view, you will only see a few things on this picture. But if you can see everything that is there, you may actually look at these things and say, "Oh, these may look like horses," but you can vary the shape and size of those horses. That's the variance part.

Bias is that I don't see any animal. Oh, come on. A horse can't be black and white, so to say.

When we look at data and then you already have a preconceived notion of what am I looking at, then you are actually generating that graph to look at that data in that way. Then you have got what you wanted. That's what business intelligence was in the past.

Variance is that when I connect the data, I create multiple models, I compare them and then look at it what is the objective function that we are looking at, how it varies by changing one or two of those parameters. How does it vary? That is the variance that we are looking at. Then we can say which can be implemented very easily.

Michael Krigsman: This is a core part of data science. Can you link that back to a practical example of digital transformation?

Dr. Satyam Priyadarshy: Absolutely. You can actually look at, for example, the case study that we have published. I think it's in Worldwide or General Petroleum Technology. I can't remember exactly. We looked at what is called unstructured data from a semi-structured drilling report.

Now, there are a lot of biases in that because people are writing their comments of what they see. Now, when you look at that from hundreds and thousands of these reports, you can actually look at it what may have caused the problem.

When you actually let the system figure out and collate that data in a meaningful manner, then you actually interpret that pattern. Then you say, "Oh, look at it," that your nonpredictive time, in this case, the technical term I don't want to use, but nonpredictive time, it was actually more than what it was recorded because the recording was done by people but the data that was written was different and that actually shows the gap between the two. That can lead to a significant improvement in the process that has taken place.

Michael Krigsman: We have another comment from Twitter. Arsalan Khan says, "The basic idea of the data value you were talking about is, as we move away from data hoarding and competition to data collaboration across the organization, that in turn requires the cultural change at so many levels," going back to what we were speaking about earlier.

Dr. Satyam Priyadarshy: Absolutely. It's a cultural change. That's the important part to create what I call co-innovation, do co-innovation.

You can have multiple silos in an organization. You could have multiple departments, department groups, whatever it is. But actually, when different people look at the same data or integrate the data from different places, there are new patterns that emerge. That is what gives you those hidden inefficiencies. That means you automatically create value for the organization.

The analytics value chain

Michael Krigsman: You talk about the analytics value chain. Tell us about that.

Dr. Satyam Priyadarshy: In the very beginning that you do science on the data. That science on the data is where you actually leverage what is called a complete analytics value chain. Not for every problem, I need to go and build a neural network.

There could be a very simple integration of multiple data sets. You look at very simple visual analytics and it could actually give you insights that can be easily translated and put in practice, so you can optimize your product, service, workflow, or your operations. That small change could lead to significant improvement in cost savings or an increase in revenue or in profit margins. That is what the goal of this business is always.

There may be very complex things where you want to actually leverage higher, more complex neural networks or deep learning algorithms. If you're talking about a lot of images to be processed, we could leverage. There are deep learning things like that.

You have to look at the business. How much of that value do you want to put in the innovation so that your return on innovation is significant? This complete analytics you do if it's simple, visual analytics. You do statistical analysis, you do text mining, you do deep learning, neural networks, whatever you want to call it, but this is a chain, so you should leverage the aspect of that chain to the maximum value rather than just running after the best shiny tool in the process.

Michael Krigsman: Can you put that into context with a concrete example?

Dr. Satyam Priyadarshy: Absolutely. There are various things that we can actually look at. For example, if you think of one of the studies that we did is failure of a pump. Then you can actually look at the data of the history of that pump failure. You can connect it to a data source that was not part of it originally, but you can get public data and connect it with that.

Then you can look at why this pump has failed in a certain region. You can then get more insights and that's an important part. I didn't have to run too much of a deep learning algorithm for it.

When we are looking at large amounts of images, either generated from the raw data or collected from over different periods, then you can actually want to look at some specific aspect in those images, how much minute change is taking place in that image, and then you can actually learn some very complex deep learning algorithms.

Michael Krigsman: In your experience, is this something that organizations really understand?

Dr. Satyam Priyadarshy: My experience over the last ten years, I don't know if I can say that people really understand it. People are always going after the next shiny tool and that is a danger towards the failure of digital transformation, according to me.

Actually, I have eight commandments of digital transformation … (indiscernible, 00:33:23) and that's one of them because, for some reason, we get advice saying, "Oh, this is the latest and the greatest tool, so we should … (indiscernible, 00:33:32)," and that could lead to failure.

You know artificial intelligence was talked about almost 50 years ago. Then there was a period it was very at the peak. Then people couldn't find results because there was no scalable computing. Then people said, "Oh, this AI neural network doesn't work for anything." After 30 years, it's one of the hottest things for people.

Again, how you implement it, how you actually leverage it in a systematic matter is what the success is. It's not about … (indiscernible, 00:34:02).

Michael Krigsman: Well, it's certainly easier to try to find some tool that magically solves all of your problems rather than thinking through the entire strategy.

Dr. Satyam Priyadarshy: Exactly. Anything that is fast doesn't always result in good things.

Advice for using data science in digital transformation

Michael Krigsman: Can you share advice, for business leaders who are listening, on how to use data science effectively in support of digital transformation?

Dr. Satyam Priyadarshy: One of the things that I talk about is that one should think about it not just in a very simple return on investment concept. It is about a return on innovation.

There is no picture here, but if you were to draw a nice picture of how much of data maturity you have, how much risk you want to take, and how much of economic value you want to generate, you can do these things in a very systematic manner. You can actually work on your dark data, construct your data, which is very low risk because that data is already there, and you're not touching any operational aspect of it. Create value from it.

Then you actually can build predictive models. Then you can actually also do R&D work at the same time, which may be higher risk but actually creates more economic value. If you put them on a graph very systematically, there is a methodology to this madness and very successful in it.

Michael Krigsman: What other kinds of challenges or pitfalls tend to arise commonly when business leaders are trying to do these things?

Dr. Satyam Priyadarshy: I think the first one, as I said in the very beginning, is that digital divide in understanding. People are using terms like digitization and digitalization randomly with digital transformation and, as I said, they have very specific economic impact which are associated with them. This leadership aspect is very, very important. Not to be confused by too much of these shiny tools as another example that one should really think of.

Think of these things in terms of your business context and your competitive landscape, not in terms of, well, because somebody has done this that we should also do that. What is my context? The context and customization of your industry is a very, very important part.

Michael Krigsman: To what extent do you need a team of data scientists to do this and to what extent is it business thinking and strategy?

Dr. Satyam Priyadarshy: When we talk about ourselves, we have data scientists but I always call the data scientists as a combination of three people, so to say. Data scientists in an ordinary way, the computer scientist that will do AI and those sorts of things, and then the businessperson and the product or the service person or the domain person.

In some sense, most of my data scientists are trained on all three of them but if that is not available, we bring people from different teams to work together because if you want to build models, you can build models all your life. They may not take you somewhere in terms of business timelines. That's where the business people and the product services or the domain guys come into play.

Don't think pure data scientist as a computer person. The three legs of the stool are very, very important.

Michael Krigsman: At the end of the day, it's the subject domain expertise working with the data experts together that creates the best result.

Dr. Satyam Priyadarshy: Absolutely, and the business partner.

Michael Krigsman: Okay. A very rich conversation. We've been speaking with Dr. Satyam Priyadarshy. He is a technology fellow. I said technical fellow earlier. I apologize for that. He's a technology fellow and chief data scientist at Halliburton.

Dr. Priyadarshy, thank you so much for being with us today.

Dr. Satyam Priyadarshy: Thank you, Michael. Thanks for inviting me. It was a great conversation.

Michael Krigsman: Everybody, thank you for watching, especially the folks who were asking these great questions. Before you go, please subscribe on YouTube and hit the subscribe button at the top of our website to subscribe to our newsletter. We have great shows coming up. Check out CXOTalk.com and check out CXOTalk Story Studio on our website.

Thanks so much, everybody. Have a great day and we'll see you soon. Bye-bye.

This transcript was lightly edited.

Introduction

Michael Krigsman: Did you ever think about the relationship between data science and digital transformation? Dr. Satyam Priyadarshy, tell us about Halliburton and tell us about your role there.

Dr. Satyam Priyadarshy: Halliburton was founded in 1919, so it's over 100 years old. Last year was our 100-year anniversary. It is one of the world's largest providers of products and services in the energy industry. We have roughly about 50,000 people in 80 countries where we have roughly 140 nationalities.

The company helps to maximize the value for our partners across the industry for the lifecycle of the reservoir, as we call it, in locating the hydrocarbons, then managing the production, drilling, and formation. This is the core business and we have been in this business for, as I said, over 100 years.

Michael Krigsman: Give us some insight into what you do as the chief data scientist at Halliburton.

Dr. Satyam Priyadarshy: As we say, data is an important asset for any company. Digital is what enables it to create value. In order to create value from the data is where data science is needed. In 2014, Halliburton was forward-thinking and, when they asked me to join from outside the industry—I joined the oil and gas industry as the chief data scientist—my goal was to build the practice around data, how to leverage data, and create value by leveraging data, which is the historical data, the new data that will come forward, and set up the center of excellence, we call it, with data, data science, and digital center of excellence.

Now, if you think of a center of excellence, most people will think that it does one function but, especially in an industry like that, you have to do a lot of things. One is data science and data itself. Creating value from data is not easy. Most people think just having data, building some dashboards is the way to create value but, actually, data science takes it one step further because the goal is to improve the business on an ongoing basis.

As a function of data science, you have to not just do the value from data but also build a team around it. We have been very successful in building one of the best data science teams in the industry.

Digital transformation challenges and pitfalls

Michael Krigsman: What are some of the challenges of digital transformation? Where does digital transformation tend to fall short?

Dr. Satyam Priyadarshy: One is to understand what digital transformation is. This is a term that people use very loosely in many cases, but digital transformation is about leveraging the digital technologies, digital assets on a continuous basis to grow the business and remain competitive. If you understand that part of the definition, then you will know that it is a step-by-step process in creating economic value.

What do you do? Digital transformation happens when you have digitized or what is called digitization and then digitalization. Then you have a digital transformation in terms of economic value. That means all this is based on data.

What are we digitizing? That is the data, the forms, the data that you collect. Once you have that in place, then you can actually start creating value. Then you create value in a step-by-step fashion.

What happens as well, as you asked the question, whether digital transformation will fail or why it fails? Because people are trying to jump too fast without understanding the full flow of the process. You should know how to leverage the data, how to create the value from it, and then how to actually implement it at the field level.

See, unlike the marketing domain or a retail space, the oil and gas industry is much more complex. You can build, so to say, going into details later on, AI or something, but let's say you have the model. You build the model on the historical data but you have to take it to the field.

What is the process of taking from the lab to the field? That has to be well established and tested out. When you don't take time to actually do this in a systematic manner, that's when most digital transformation projects fail.

The second reason why it fails is, there is a term that I used called the “digital divide.” That means how people understand what digital transformation is at the leadership level, at the operational level, at the actual management level, and the value of creating value from the data and the opportunities associated with it. That understanding is different. That digital divide causes this digital transformation to fail.

How to use data science for digital transformation?

Michael Krigsman: How do we then use data science in the service of digital transformation? Quite frankly, when most people talk about digital transformation, they're talking about things like business models and culture change. The subject of data science doesn't usually come up as one of the top items.

Dr. Satyam Priyadarshy: That's surprising, those who say that, because if you don't do science on the data then how do you actually leverage and build a digital transformation strategy? Yes, the cultural change is an important part. The technology is important. The people part is very, very important. But the actual doing science on the data is very important.

Most of the bigger industries are based on engineering and scientific formulas, so people have been using them for a long time. But now, we have years and decades of data that has been collected for every workflow. That can be looked at in an effective sense and learn from it what gaps have been there, what I call hidden inefficiencies.

Data science is actually very powerful in finding those hidden inefficiencies because if the efficiencies are not hidden, then you must have fixed it. If you have not fixed them, then it's a different problem at the management level. Most of it is to find hidden inefficiencies and, once you find them, to actually figure out a way to incorporate those fixes and then actually create low-cost solutions for yourself so you can actually grow the business in terms of revenues and profits.

Michael Krigsman: What's the relationship then between these hidden inefficiencies that data science can uncover and digital transformation?

Dr. Satyam Priyadarshy: When you have found some hidden inefficiencies, you are actually either optimizing or maximizing your workflow. That workflow change is what leads to the transformation on how you do your business, whether it is customer service, whether it's equipment manufacturing, whether it's testing equipment, or looking at the production side. Then you can actually do this in a continuous form. The more hidden inefficiencies you find, the more transformational things you can implement, and then you can build a highly optimized system across the ecosystem.

What is dark data?

Michael Krigsman: You talk about dark data. Weave that into the story as well, please.

Dr. Satyam Priyadarshy: I give an analogy. All of us have very, very high resolution, high-end smartphones but if I asked you to find a photo of yourself which was taken five years ago, how much time would you take to find it? That data is there, but it's dark. You can't find it.

Imagine now a corporation of any size. The data is actually not easy to find. You know the search engine technology is pretty mature, but you tell me how many big, good corporations are there where you can actually go and type the model number of a, say, centrifugal pump and it can give you the history of it. The data exists either in your databases and data warehouses, on paper, or in somebody's drawer, but the data is not creating value.

The data that is there is actually expensive because it is costing you to keep it. It's not free. No storage is free, whether it's cheaper or not. But when you don't create the value, that data is called dark data.

Michael Krigsman: What do we do about that dark data? Again, what does that have to do with digital transformation?

Dr. Satyam Priyadarshy: The dark data has a lot of learnings in it. Look at a process or a workflow over the last ten years, because many of the industries actually do the same workflows over and over again and things break down. Of course, when you're operationally active, you will actually certainly go, fix it, and move on. The learnings are actually hidden in that data.

The goal is to leverage that dark data and convert it eventually into what we call smart data. What does smart data mean? That means the data can be acted in real-time.

If you are going towards automation in terms of what we call industry 4.0, that means we want to automate everything. You cannot automate and build out systems unless your data is smart. That means you can act on it either through machine learning, artificial intelligence, data science, or any other method that can be acted on. You take that dark data, you convert that to smart data from your learnings by applying data science, and then you transform your workflows.

Convert dark data to smart data?

Michael Krigsman: How do you convert that dark data into smart data?

Dr. Satyam Priyadarshy: It's a process and I've written quite a bit about it. You first start with a small set of dark data. You create some value. You see what's valuable in it, what are the missing gaps in it, and then you start trying to integrate data from multiple silos.

As we all know, most companies or most organizations have their data in multiple silos. When you start connecting them, you build a scalable system. Then you eventually test it out in real-time. Once that real-time feedback is put into where that data ingestion is taking place, at the ingestion point or the generation point, you fix all the issues that are there. Then you actually have smart data.

Michael Krigsman: We have a question from Twitter. Arsalan Khan asks, "Have you been in situations where the data that was collected ended up not having the value?" You've gone through this dark data. It doesn't have value. I guess the more general point therefore is, it's obviously not practical to convert all the data that exists from dark data into smart data, so how do you make those selections?

Dr. Satyam Priyadarshy: As a scientist, you know that every failed experiment is also valuable. Every data has value. The question is how much value it is for you or for that business or for that process.

One has to look at the data. Sometimes people say, "Oh, I collected the data. It has no value there." But people haven't actually even looked at it.

If there is no value, then why did you start collecting it? Why are you spending your capital and operational expense on it? Correct?

The question is to quickly do a phase study. Assess how much value it is. Out of that data, how many elements of that data are really valuable? If this may not have value even in the future, you don't want to probably collect or even generate that data. But if you are generating any data, there has to be some value. Otherwise, we probably started on the wrong foot.

Using data in digital transformation strategy

Michael Krigsman: From a digital transformation standpoint, does that mean that the thinking through of the type of data that you need needs to be part of the strategy, the digital transformation strategy right at the beginning?

Dr. Satyam Priyadarshy: Absolutely. There are four pillars of digital transformation, as I call it. One is what I call the big data analytics and big data doesn't mean just large amounts of data. Small data is a part of it because it is the full continuum of data that we are talking about.

Now the second part is to leverage the computing paradigms that are existing or we call emerging technology paradigms. The third one is to build real-time systems and then the fourth one is autonomous. This is what whole digital transformation is. The foundation lies in the data.

Michael Krigsman: Can you give us any examples? I think this is an area where folks have trouble. As we said before, people tend to think about business models but I think the data aspect can be very challenging for organizations.

Dr. Satyam Priyadarshy: One of the simplest example that I typically give about digital transformation is from the music industry. All the baby boomers and people of my age group would remember that the songs were digitized and put on LP records. Many of the younger generation probably have no idea what they are, but those LP records also required a very big box on which they were played. You think of it that it was sort of digitized data, but it was not very easy to share. If you wanted to share songs with friends and family, you had to invite them to your house.

Then a transformation took place in terms of CD and MP3. It was slightly easier to share the songs because you could borrow a CD and play and then give it back, and then MP3, like iPods and things, that came.

Look at today. The whole music industry is all streaming music. I can actually listen to songs that I like and I can share with Michael my album without actually sending any songs to him. This is called a real digital transformation.

The song remains the same. The data remains the same. The way it has been delivered by leveraging the emerging technology and the data science done on it—what preference of songs, who is liking it and actively shared across the world—is transformation of the industry.

The economic value, as we all know from the music industry in the olden days, is only five, maybe ten record houses and ten artists made money. Today, there are millions and millions of people who are making music, enjoying the music and, actually, the economic value is significant in it. As we all know, it generated a large number of jobs that never existed. This digital transformation is all about creating new kinds of jobs as well as a part of the whole exercise.

Michael Krigsman: That's so interesting. If we think about Spotify or Apple Music, any of these, in fact, what's going on is the data is stored in their repository and they're letting you share the metadata.

Dr. Satyam Priyadarshy: Absolutely. I stay away from the technical term of metadata. I call it the knowledge. I created the album with human augmentation of what I like. Then I can share it, so what we are sharing is the knowledge on the data, that, "Oh, this is a beautiful album. You should listen to it."

Think of many industries. If they really are serious about transforming themselves, then they should be really elaborating and co-innovating solutions where they can actually share the knowledge so that the cost of delivering services can be reduced across the industry. That becomes very important in the energy sector anyway.

Metadata vs. knowledge

Michael Krigsman: I'm quite interested in this topic. You say you prefer the term "knowledge" rather than the term "metadata." Can you explain why? Why do you talk about it that way?

Dr. Satyam Priyadarshy: Because metadata has a connotation in our industry. It's something like what you call how you tag the data, right? But what I'm saying is knowledge that is built on top of the data.

An album is the knowledge, so to say, whereas metadata would be the artist's name, whatever is the genre of the music. That would be the metadata.

Metadata, if I were to buy a mechanical pump and put it in my SAP system, then the purchase order is metadata. The history of the pump that has failed many times, I analyze that failure and create a root cause analysis of it and a regression model on it, then that becomes actually your knowledge.

Michael Krigsman: The knowledge is the information that's really useful for businesspeople in terms of their processes and their shared experiences? Would that be an accurate way of putting it?

Dr. Satyam Priyadarshy: Knowledge is what we actually incorporate wisdom, if you think of it. Knowledge is generated by data and metadata together. Then you have knowledge. Then the wisdom part comes when all of us come together.

Michael Krigsman: Again, link it back to transformation, to digital transformation for us.

Dr. Satyam Priyadarshy: Exactly. Once you have the sharing of this knowledge, it becomes easier to transform, industry-wide or in the ecosystem, your workflows, your business models, your strategy, your future competition. That is very important, especially if we look at an energy sector.

I'm sure you must have heard about this term called energy transition. There is a lot of pressure on zero carbon, negative carbon, and things like that. But what is that?

We are building a landscape now of energy. You have to be a player in it. If one sector doesn't transform and keep up pace with the others, then you will have challenges. This is where the transformation comes into play.

Discovering the value of dak data

Michael Krigsman: Sarada asks, "How do you find the value of dark data? The data is dark and we may not know anything more about it."

Dr. Satyam Priyadarshy: The dark data is created by us, so the value is there. You have to pull it out and do some analysis in what I call proof of value projects. That's how you get it.

Why is it called dark data? Because human beings have created that data, set them under the table, and not shared with anybody. That's why it's dark data.

You can do projects. You can do very basic analysis. You can do simple, visual analytics. You don't have to go and run after deep learning for everything. If you see limited value then you can build more expensive and more compute-intensive models to actually integrate data to create more value.

Michael Krigsman: It always starts with some type of proof of concept to see what the value of that data really is.

Dr. Satyam Priyadarshy: I stay away from that term also, "proof of concept," because concepts are well proven. Artificial intelligence, data science is very well-proven. The engineering process are well proven, so to say.

What we are trying to create is value. Value is the most important part for the business. We call it proof of value.

Michael Krigsman: That's always, at each step, your metric, which is the question. What kind of value are we creating and for who, as well, I'm assuming?

Dr. Satyam Priyadarshy: It's the business, so we are talking about business. If you are doing an academic exercise, that's a different thing. You can write as many papers as you want on the same problem. But in business, you have to actually produce something which has value.

Michael Krigsman: We have another question from Twitter. "Should your data strategy plan for what data you should keep and not bother with storing the rest?" Do you just figure out what you're going to keep and you throw away the rest or do you do something different?

Dr. Satyam Priyadarshy: … (indiscernible, 00:20:36) every raw data is very important. The raw data is the only single source of truth.

As a data scientist and with a career of over 20 years, I believe every piece of raw data is valuable. If anything has been modified, changed one single bit, that's not raw data. That is subject to checks and balances.

If you are generating any data, you should keep it. You never know when it's needed. The goal is not that everything is to be live. The goal is to have it easily accessible so you can always build models and check it back that your model is working under certain … (indiscernible, 00:21:19) conditions back to what it should … (indiscernible, 00:21:22). Just throwing the data without knowing why you threw it or without looking at the value of the data, that doesn't make sense.

Michael Krigsman: We have another question from Twitter, another one from Arsalan Khan. This is a really good one. He asks, "Since data touches every part of the organization, how do you create sandboxes for people who are not in technology to play and experiment with that data and come up with their own thoughts on what kind of data should be collected?"

Dr. Satyam Priyadarshy: Every organization has a culture around it. I think it's a lot of what I call education needed both at the leadership level, at the mid-management level, and individual contributors level, so to say as to why somebody wants to play with this data and what they are going to do. Good governance, good practices about cybersecurity are in place and how this data will be actually used by the people or the group that is trying to look into it, that has to be really sort of codified. Then, actually, it's much easier to create value from that data. That's the first part.

Blocking from the data just for the sake of locking down is not a way to create value. But, at the same time, having the right governance structure around it is very, very important and, of course, requires a cultural change, as you said in the very beginning that digital transformation has a big cultural change component. Creating value from data by anyone requires a cultural change in the organization.

Digital transformation and culture change

Michael Krigsman: Would you elaborate on that? Why does creating value from data require a cultural change?

Dr. Satyam Priyadarshy: If you look at the history—and before I came to the oil and gas industry, I've probably consulted for many industry verticals—it is just like people have been locking down the data and that is the biggest part.

Say, for example, the marketing people have their own data. The sales guys have their own data. Then if you actually add up, sometimes they don't add up to the value that the CEO is looking at. Even if you were to look at sales organizations in some places and having done this exercise a few times, the north region versus the south region plus the eastern region plus the western region, if you add up the numbers, sometimes they don't sync to one.

My goal as a data person, I'm looking at the starting point and the endpoint. They should all connect in some way to actually be coherent. That requires cultural change because people saying, "Oh, my region has been doing great. Yours has not been doing great." That's why it requires cultural change. New thinking is needed in this if we want to seriously adopt digital transformation.

Michael Krigsman: I'm really hung up on this phase that creating value from data requires a cultural change. Wouldn't it be possible to gain value from data? That is basically what you've been doing and there's no cultural change required.

Dr. Satyam Priyadarshy: It's not possible. If their organization was already what I call a data native company, that's a different story, like the Internet companies are purely data native companies. They survive and thrive on data, so their culture is already set.

Let's look at other industries, which are implementing things like, say, IoT sensors and newer technology devices. How do they actually build that culture? "Why should I look at data in the first place rather than just by inclusion? I've been doing this by the same process for the last 20 years. I know how to fix it."

You can see the example of the healthcare industry. They have been integrating data from multiple sources. Now EHRs or EMRs are available. In some places, you can actually look at their reports and things like that. That culture change has taken place. Five years back, it was not the case. Now we are talking in the healthcare about precision medicine.

Michael Krigsman: It's for this reason of what you're just describing that data is at the heart of digital transformation and sometimes we hear the phrase, "Data is the new oil."

Dr. Satyam Priyadarshy: I don't like to call that myself, but data has been there forever. It's an asset, according to me. It's a very expensive asset that we all have because data was generated over the years. So much money was spent in generating that data, collecting it, storing it, putting in data models, putting in databases, data warehouses, so that is a cost to the company.

Michael Krigsman: Could you explain the bias versus variance tradeoff? First, you have to tell us what is the bias versus variance tradeoff. [Laughter]

Dr. Satyam Priyadarshy: Bias is when I look at some data. I usually use a picture about these kinds of things. In fact, you can actually look at my background and there is a nice picture. If you have a biased view, you will only see a few things on this picture. But if you can see everything that is there, you may actually look at these things and say, "Oh, these may look like horses," but you can vary the shape and size of those horses. That's the variance part.

Bias is that I don't see any animal. Oh, come on. A horse can't be black and white, so to say.

When we look at data and then you already have a preconceived notion of what am I looking at, then you are actually generating that graph to look at that data in that way. Then you have got what you wanted. That's what business intelligence was in the past.

Variance is that when I connect the data, I create multiple models, I compare them and then look at it what is the objective function that we are looking at, how it varies by changing one or two of those parameters. How does it vary? That is the variance that we are looking at. Then we can say which can be implemented very easily.

Michael Krigsman: This is a core part of data science. Can you link that back to a practical example of digital transformation?

Dr. Satyam Priyadarshy: Absolutely. You can actually look at, for example, the case study that we have published. I think it's in Worldwide or General Petroleum Technology. I can't remember exactly. We looked at what is called unstructured data from a semi-structured drilling report.

Now, there are a lot of biases in that because people are writing their comments of what they see. Now, when you look at that from hundreds and thousands of these reports, you can actually look at it what may have caused the problem.

When you actually let the system figure out and collate that data in a meaningful manner, then you actually interpret that pattern. Then you say, "Oh, look at it," that your nonpredictive time, in this case, the technical term I don't want to use, but nonpredictive time, it was actually more than what it was recorded because the recording was done by people but the data that was written was different and that actually shows the gap between the two. That can lead to a significant improvement in the process that has taken place.

Michael Krigsman: We have another comment from Twitter. Arsalan Khan says, "The basic idea of the data value you were talking about is, as we move away from data hoarding and competition to data collaboration across the organization, that in turn requires the cultural change at so many levels," going back to what we were speaking about earlier.

Dr. Satyam Priyadarshy: Absolutely. It's a cultural change. That's the important part to create what I call co-innovation, do co-innovation.

You can have multiple silos in an organization. You could have multiple departments, department groups, whatever it is. But actually, when different people look at the same data or integrate the data from different places, there are new patterns that emerge. That is what gives you those hidden inefficiencies. That means you automatically create value for the organization.

The analytics value chain

Michael Krigsman: You talk about the analytics value chain. Tell us about that.

Dr. Satyam Priyadarshy: In the very beginning that you do science on the data. That science on the data is where you actually leverage what is called a complete analytics value chain. Not for every problem, I need to go and build a neural network.

There could be a very simple integration of multiple data sets. You look at very simple visual analytics and it could actually give you insights that can be easily translated and put in practice, so you can optimize your product, service, workflow, or your operations. That small change could lead to significant improvement in cost savings or an increase in revenue or in profit margins. That is what the goal of this business is always.

There may be very complex things where you want to actually leverage higher, more complex neural networks or deep learning algorithms. If you're talking about a lot of images to be processed, we could leverage. There are deep learning things like that.

You have to look at the business. How much of that value do you want to put in the innovation so that your return on innovation is significant? This complete analytics you do if it's simple, visual analytics. You do statistical analysis, you do text mining, you do deep learning, neural networks, whatever you want to call it, but this is a chain, so you should leverage the aspect of that chain to the maximum value rather than just running after the best shiny tool in the process.

Michael Krigsman: Can you put that into context with a concrete example?

Dr. Satyam Priyadarshy: Absolutely. There are various things that we can actually look at. For example, if you think of one of the studies that we did is failure of a pump. Then you can actually look at the data of the history of that pump failure. You can connect it to a data source that was not part of it originally, but you can get public data and connect it with that.

Then you can look at why this pump has failed in a certain region. You can then get more insights and that's an important part. I didn't have to run too much of a deep learning algorithm for it.

When we are looking at large amounts of images, either generated from the raw data or collected from over different periods, then you can actually want to look at some specific aspect in those images, how much minute change is taking place in that image, and then you can actually learn some very complex deep learning algorithms.

Michael Krigsman: In your experience, is this something that organizations really understand?

Dr. Satyam Priyadarshy: My experience over the last ten years, I don't know if I can say that people really understand it. People are always going after the next shiny tool and that is a danger towards the failure of digital transformation, according to me.

Actually, I have eight commandments of digital transformation … (indiscernible, 00:33:23) and that's one of them because, for some reason, we get advice saying, "Oh, this is the latest and the greatest tool, so we should … (indiscernible, 00:33:32)," and that could lead to failure.

You know artificial intelligence was talked about almost 50 years ago. Then there was a period it was very at the peak. Then people couldn't find results because there was no scalable computing. Then people said, "Oh, this AI neural network doesn't work for anything." After 30 years, it's one of the hottest things for people.

Again, how you implement it, how you actually leverage it in a systematic matter is what the success is. It's not about … (indiscernible, 00:34:02).

Michael Krigsman: Well, it's certainly easier to try to find some tool that magically solves all of your problems rather than thinking through the entire strategy.

Dr. Satyam Priyadarshy: Exactly. Anything that is fast doesn't always result in good things.

Advice for using data science in digital transformation

Michael Krigsman: Can you share advice, for business leaders who are listening, on how to use data science effectively in support of digital transformation?

Dr. Satyam Priyadarshy: One of the things that I talk about is that one should think about it not just in a very simple return on investment concept. It is about a return on innovation.

There is no picture here, but if you were to draw a nice picture of how much of data maturity you have, how much risk you want to take, and how much of economic value you want to generate, you can do these things in a very systematic manner. You can actually work on your dark data, construct your data, which is very low risk because that data is already there, and you're not touching any operational aspect of it. Create value from it.

Then you actually can build predictive models. Then you can actually also do R&D work at the same time, which may be higher risk but actually creates more economic value. If you put them on a graph very systematically, there is a methodology to this madness and very successful in it.

Michael Krigsman: What other kinds of challenges or pitfalls tend to arise commonly when business leaders are trying to do these things?

Dr. Satyam Priyadarshy: I think the first one, as I said in the very beginning, is that digital divide in understanding. People are using terms like digitization and digitalization randomly with digital transformation and, as I said, they have very specific economic impact which are associated with them. This leadership aspect is very, very important. Not to be confused by too much of these shiny tools as another example that one should really think of.

Think of these things in terms of your business context and your competitive landscape, not in terms of, well, because somebody has done this that we should also do that. What is my context? The context and customization of your industry is a very, very important part.

Michael Krigsman: To what extent do you need a team of data scientists to do this and to what extent is it business thinking and strategy?

Dr. Satyam Priyadarshy: When we talk about ourselves, we have data scientists but I always call the data scientists as a combination of three people, so to say. Data scientists in an ordinary way, the computer scientist that will do AI and those sorts of things, and then the businessperson and the product or the service person or the domain person.

In some sense, most of my data scientists are trained on all three of them but if that is not available, we bring people from different teams to work together because if you want to build models, you can build models all your life. They may not take you somewhere in terms of business timelines. That's where the business people and the product services or the domain guys come into play.

Don't think pure data scientist as a computer person. The three legs of the stool are very, very important.

Michael Krigsman: At the end of the day, it's the subject domain expertise working with the data experts together that creates the best result.

Dr. Satyam Priyadarshy: Absolutely, and the business partner.

Michael Krigsman: Okay. A very rich conversation. We've been speaking with Dr. Satyam Priyadarshy. He is a technology fellow. I said technical fellow earlier. I apologize for that. He's a technology fellow and chief data scientist at Halliburton.

Dr. Priyadarshy, thank you so much for being with us today.

Dr. Satyam Priyadarshy: Thank you, Michael. Thanks for inviting me. It was a great conversation.

Michael Krigsman: Everybody, thank you for watching, especially the folks who were asking these great questions. Before you go, please subscribe on YouTube and hit the subscribe button at the top of our website to subscribe to our newsletter. We have great shows coming up. Check out CXOTalk.com and check out CXOTalk Story Studio on our website.

Thanks so much, everybody. Have a great day and we'll see you soon. Bye-bye.