#81 The Gradual Process of Building a Data Strategy

Hi, I'm Maggie, VP of Curriculum at datacamp, and I know firsthand how overwhelming it can be to keep up with the rapid changes in data and AI. But staying ahead doesn't have to be hard. That's why our curriculum team has curated the best learning content to help you and your team stay ahead. Covering everything from data and AI literacy for busy executives to cutting edge generative AI for developers, and everything in between. Whether you're upskilling yourself or your team, we make learning practical, engaging and flexible. It just works. Power your data and AI transformation journey. On datacamp.com.

You're listening to Data Framed, a podcast by Data Camp. In this show, you'll hear all the latest trends and insights in data science. Whether you're just getting started in your data career or you're a data leader looking to scale data driven decisions in your organization, join us for in depth discussion discussions with data and analytics leaders at the forefront of the data revolution. Let's dive right in.

Hello everyone, this is Adele, Data science educator and evangelist at datacamp. One thing I love about hosting the Dataframe podcast is that I get to speak with really smart data leaders who can break down the different levers for becoming a data driven organization. And you can argue once you put all these levers together, you have the key components of a data strategy. This is why I'm so excited to speak with Vijay Yadav. Vijay Yadav is the Director of Quantitative Sciences and Head of Data Science at the center for Mathematical Sciences at Merck. He is a seasoned data leader who drives the analytics strategy and roadmap for Merck's manufacturing teams and owns the development and deployment of advanced analytics capabilities throughout Merck. Vijay has over 20 years of experience working in pharmaceutical and chemical manufacturing and has deep insight on developing data strategies that scale. Throughout the episode we speak about his background, the key components of a data strategy, how data culture and people are arguably the most important aspect of a data strategy, how to structure upskilling programs for data science and more. Now let's dive right in. Vijay, it's great to have you on the show.

Thank you Adel, so great to be here.

So I'm excited to discuss with you, you know, all things data strategy, how to operationalize and build data culture, how to accelerate data science within an organization. But first, can you give us a brief background about yourself and how you got into the data space?

Yeah. Thank you Odell. It's great to be here by trade. You know, I am computer scientist and I started my career in IT space and Data has always been part of. I consider myself as a data practitioner even though in my college days I was specialized in artificial intelligence and numerical methods statistics. But I got to apply those techniques later in the career. But I can tell you that throughout my career data has been part of my passion and majority of the solutions. I have been able to apply some of the AI and machine learning techniques to deliver some business value for the company. Currently I'm with Merck. I'm heading the data science team and primarily focused on developing the AI and ML solutions to make manufacturing modern kind of capability enabled area.

There's definitely a lot to unpack today and what I'm really excited to talk to you about is really the different elements of a data strategy and how to get there. You're a very experienced leader in the data science space and you've worked a lot on data transformation projects. And I especially love this topic of data strategy because it really allows us to dig through basically all of the elements of what makes an organization data driven. And I think that's very exciting. So I'd love it if you can walk me through kind of from a high level, the different components or elements of a data strategy.

Yeah, no, I mean that is, that is great actually. And that's the way you want to think holistically, what the data and analytics strategies would look like. In my experience, there are six pillars of that strategy. The number one at the top is basically data analytics strategy itself. Who owns that strategy? Who develops that? Right. Who are the players basically want to do that? So that's the first component, the very strategic point of view that as well. What are the business objectives? End of the day, I think you want to have your strategy based on the actual business strategy where business wants to basically go. Right. So that's the first component. The second component is operating model. This is the new area, Right. What are the different players? Who plays the role? What is the roles responsibility of data and analytics organization? What is the role of technology function? What is the role of business functions? Different business function we have, whether it's a data product team, whether it's engineering team or design team. So there are multiple players. How we come together, orchestrate in a way that it's a cohesive in order to make this strategy and execute and drive the value out of that. So that's the second component, the third component, I will say the data assets itself. So how do we go about building the data assets? You have to really think about it's. We all call it data is oil and data is gold, it's an asset. Right? So how do we take that and make that. We can basically drive the value, preserve it. Right. To do that now, it's not you take all the data and then you dump into a data lake, right? That is not the strategy you're putting in the data. What I mean by data strategy is that what can we do that is accessible in an easy way? It is governed that only people who need to access, they have to have access. It makes sense for someone to look at that. It's more like democratize the data itself. How do we do that as well? That's a component. The next component is the infrastructure, the platform itself. We have a data. How do we apply the infrastructure? Put together infrastructure that is scalable and it can take the speed. The business wants to do that. So infrastructure, whether it's a platform, data platform. And also when we talk about the data science field, if data scientists are working, what kind of tools and technology they're using. So we got to put together a place where it can be shareable and scalable. The next component I would say is the data product management. This is fairly new concept. If you talk about the Google and Facebook and Meta now, the concept the data product is it's really in Silicon Valley is very prevalent. But what it is not prevalent is in other industries, whether it's a pharma, whether it is other ones. So how do we bring the concept of product? When I say products, a data product, the concept is that it's not that you take initiative, you start and finish, you have to develop that and you want to make that product accessible to the entire company. And that's how you have to treat as a product. So what kind of methodology we need to basically bring on that as well. Right. And the last one, I would say the people and culture, even though it is the last one. But I'll keep at the last. Right. So how do we organize this organization itself? Data and organization, what is the culture? And culture, I can tell you, is the biggest component in the strategy at all levels. And we'll talk more as we as we go. But at the high level, these are the six areas that I think are really required.

That's awesome. So there's definitely so much to unpack here. And I'd love to go through each element, by each element and dig through and kind of pick your brain. And I said I love this conversation around data strategy just because how holistic it is and because we get to tackle all of these different elements. Let's start off kind of with the ownership of the data strategy you mentioned. That's one of the first elements when you think about ownership. You know, oftentimes I see with an organization there's often a trade off or kind of a free. Especially as they go about their first data analytics journey at the beginning within their data analytics journey around who owns the strategy. And it could be there's often a debate about whether it's IT or the data team or the business functions in some sense. Who do you think should own the data strategy and its execution?

Now that that is such a great and important question. Adil, I can tell you that the company who are company who are doing well those company the business owns a strategy, data strategy and the company who are not doing well. It's a very traditional way somebody else owns it. Could be technology or someone else, right? In my opinion experience I can tell you is that high success you will get when the business owns a data strategy. So what I mean by business owning that. So think about who is the user of the data. It is a business. Business knows that what basically good looks like. Now we may not be capable but that's where we want to put together who can, who can represent the business what good looks like to them, right? To do it. The drive the value, right? If the business is the one, this would be the one driving what. What do I need really to do that? Not you know, it organization traditionally that that's what it is. And also the ownership in terms of drive the value. So let's say we develop a solution, right? We deploy it. End of the day business is using it. Who makes the commitment that hey we developed this tool. Now here the value we driven, not it, not any other function. Business is the one who's going to make the commitment that we define what success looks like. We are using it. And here is the value of delivering. So you can take that value and go to the next level basically do that. So in my experience and opinion and very strong opinion is the business is the one that should basically own that strategy. Now business is the business of doing, you know, technical if pharma company making the medicines you need people data in kind of expert strategist really to represent business. And they are the one basically on behalf of the business basically want the strategy to drive it.

What you touched upon here is really that data science and data to a certain extent is not a traditional technology project, right? It's not something that you take from point A to point Z. It's a mindset shift. It's kind of a new methodology for solving business problems.

If you talk about the business of data, that business of data is feeding, enabling your main line of business. Right. So if you think of vertical of the business of data now when you talk about the business then you have to have a customer. So your business is the customer of the business of data and this is.

Why it should be owned by the business, Is that correct?

100%. That's exactly right. You summed it up really, really well to do that. So one comment I wanted to just make it so the way I think today and those are the differentiating factors. So any company is at least they are in the business of two things. One is the primary business they are in. So if you're a pharma company, you are making the pharma product, that's your primary business. But you are also in the business of data. Business of data is a concept where and you're driving value out of that. So I think the, the mind shift here is that you're not only the business of what you're doing, but you're also business of data. If you want to take your organization to the next level of innovation.

Yep. And in conversations with the C suite especially, how do you convince them that hey, the data strategy should not be owned by IT because probably C suites are very comfortable in handing off these technology driven projects towards IT and not towards the business owners necessarily. So how does that conversation go and how does that kind of mindset shift within the C suite happen?

That is such a great question. And I can tell you it's not easy one and folks at that level, right. They have been operating in a certain way. What I would say basically makes the case for them is that the fact that the business understand what good looks like. Right. IT organization is not the one. They are expert in something but they are not close to what the business basically looks like. What I have done in my experience is that showing that show the success that hey listen, when we came in and we represented the business, we talked to the business, we defined the ownership of the business. Here's what we delivered and here's the value. Now you compare that outcome with any other solution that you have basically done and you can see the difference. Just assigning that ownership driving business is driving that. You can see the difference. Right? So there's no straight answer. I think it takes time. So sometime you saw the success that this strategy, if you change it then I can tell you that this has proven better value than traditional way of doing things.

Given that the business should own it, what is the role of IT then in executing and ideating the data strategy?

If you're saying data is business, then what it does, it has a big role to play. So think about, we talked about one of the pillars is the enablement, right? It's the data and analytics enablement, right. So the infrastructure, right, it is. So for example, if you need architecture, if you need to architect a solution, that's a purely technical role that has to be played by that, right? Or you need a platform, the IT platform that we need for that it. So I would say that the delivery, once we define what success looks like, the business says here's good looks like and here's my vision for this particular solution. I'm going to lay out the vision. This is what good looks to me. Let's define those pieces and then it you come and enable us wherever that technology component basically comes into play. So I would say it still plays a big role, but I think the business is the one at the driver's seat. That's the piece. That difference I see there, that's awesome.

So we covered kind of the ownership of data strategies first component. Let's move on to the second component, which is the operating model, right. How do you organize data teams? How do you organize the players, as you mentioned them, within this organization, within this new practice, within an organization to deliver value, Right. So in terms of structuring the operating model for a newly formed data team, what are the different options that you can pursue?

So one of the things I have seen in my career as success is that one of the in the operating model, the sponsors play a big role. So what I mean by that, so the sponsor are the one basically your champion for the business case. Now when you make the business case to develop a data product using AI, ML and any analytics product should do that, it has to be based on the value that is being driven. Now the business sponsor is really the key for you to take represent you at the senior level. Not only that, if we involve the sponsor, then they are the one making the change management possible. So it's not that I go and develop a model here it is and somebody just adopts it. There's a change management process that has to take place in order to deploy the new solution. So if you have a sponsor, then you can basically take it. So that's number one in any operating model. Every time you're launching a data product solution, your sponsor are really playing the key role there as well. Right. So that's one piece to it right. Now talk about the product management piece. Right. So who sets the vision for the data product? What good looks like? Right. So normally the product team is the one vertical that basically do. And I have done that successfully in multiple roles that I have basically played. So that is one role that is basically someone who is defining the vision, representing the business here at the roadmap basically looks like. And this is how we are basically going to do that now.

Kind of the roadmap of the data team.

That's right. Yeah, that's right. To do that now it is a play. It is, you can say that's where they're coming to play as a team, playing the role in delivering the IT infrastructure, the platform architecture, anything compliance for example in these days, I think data compliance is really part of that. Right. So we have to have that element into that as well. And so we define who is doing what. Right. So clearly well defined. The other element is that the business subject matter experts. Now if we are taking the, if we are any saying that hey, business is owns that strategy, right. So the business has to allocate the resource as a subject matter expert. So as the project is going on, development is going on, we have someone who has the expertise to basically provide the insight what good looks like and things of that nature. Right. So there's another business plays a role. So we talked about the sponsorship, we talked about the it, then we got a product team, then we got business side subject matter experts. And I think somewhere in the organization is a change management comes into play. So who plays that role? Right, that's another element of that. And on top of that think about the priority. So who defines which projects would we take that is it a product team or someone from the business that hey, we got 10 things here, here is number one. So someone has basically a priority person that you're basically running it and also you want to drive the value out of that. So somebody should be responsible to track the value into the process.

Even kind of before thinking about how to define a team, then companies and organizations should really think about who's owning the process, who's setting the vision and the roadmap of the data team, who's sponsoring the change management, ensuring that the data team is supported. Even before thinking about who to get as a data scientist or how to build the data science team.

Absolutely, absolutely. That's really important and I think that's a major problem right now we have is I have been in a situation where when I Engage. Nobody knows who is doing what. And I think I have seen the success. If you clearly define the roles and responsibility and a very clear map, then I think you have a high level of success in that scenario.

And without that base layer that you're discussing of setting the roadmap, setting the processes, you're setting up data scientists to fail because they don't know necessarily what they are building or who they are building for. Is that correct?

100%. I think one of the element, and we'll talk about that is that to me I think customer experience or user experience is the key. What I've seen is that we design traditionally people try to define a solution and try to find the user after the fact. That is not the case. Your solution should be really tailored to the business and the user. You need to understand the user base and you're designing a solution of that as well to do that. Yep. Go ahead.

Awesome. Now that you've like we've have this operating model relatively set in place, how do you organize the data team? You know some organizations, they opt for like a center of excellence centralized data team. Other organizations, they opt to, we take a data scientist, we put them in the business and they have skin in the game and they're part of the business. Which model works for you that you found works for you most between a centralized and decentralized model and a follow up on that is what are the trade offs between both?

I would say the mixed model. So think about there are analytics needs that are different in nature. The complexity is different. Right. So if you embed a data science team or data analytics professional as part of the business who are doing, you know, let's say dashboarding or any type of insights that we can basically create in a simple way doesn't need advanced analytics app. Right. So I think you can embed those as part of supporting day to day and they understand the local needs, they are part of that. I think the major piece you need to centralize and what I've seen basically working is that you take that one complex piece because you need to bring the different skill set together. And if you're fragmented, you may not be able to drive the value. So I would say it's a mix of the two. You support the local where things can be supported by not too complex, very local. But when you take the major transformation initiative, I think having a central team with a different skill set, I think that that's really, really important.

Yep. And especially early on the analytics journey, it's very important to Kind of set the stage with a centralized model and then you can over time embed and become much more hybrid. Is that correct?

That's 100%. I think that's, I think that's the key point is that a lot of organizations are still maturing. You don't know what the central basically looks like. So I think there's a whole data culture piece. You want to make people aware that hey, the data can drive the value, right? So that is part of the culture. When you go and work, assign the people in bid as part of the local. So people are getting aware, they're getting to know as the time goes by, maturity level up, now you can centralize because now the company is much more aware of what data can deliver for them.

You know, the third component and fourth component that you mentioned at the beginning was kind of data asset management as well as data analytics enablement. I think this is where a lot of it plays a massive role. So let's start off with data asset management. Oftentimes it's the least exciting element of a data strategy, let's say, but it's probably one of the most important because it's really the source of truth and kind of determines the truthfulness levels of the insights one captures. So moving on to the other components of data strategy, of course we can dedicate an entire episode towards asset management, data asset management and data governance. But I wanted to gain your insight on how important should data asset management manifest itself within a data strategy and.

How so it's critical. It is the critical component of the data strategy. One thing that you see that we complain, oh my God, data is not clean, the data is scattered. Is it desperate? If you think about the reason behind why we have so much problem in the data, what's wrong with that? Why did we not think about the data in a structured way and we can reference it. One of the element that is really a success is that you want to give the context to the data. So what that basically means is the metadata. So think about, I have a piece of image, right? I'm just showing you the piece of image which has something to do with let's say vision inspection. A product inspected, you take the image of the product and machine is telling whether product is defective or not. Now I want to attach that image, one piece of information with everything else outside it. Where it came from, which site it was, which plant it came from, which machine, right? So the up collect that data in a more hierarchical way. The organization, the site, the product and any other metadata you have imagine if you can attach the metadata along every piece of information that exists in the company. That's when you'd make the power really to go and filter the information, extract it. Now you have defined it. So giving a context to the data is really critical piece. Now that is not easy task by any means. It is the complex, it can take time. But I think if you think on those lines, it's really, really important. One biggest challenge that we are having is that anytime if I'm a user on the shop floor and I am, I enter the data through my own hands on the keyboard or the machine that I'm looking in front of me, I can see the data that data captured and sent to the data lake. Now imagine what exactly happened in the process. The data was just so close to me, I just saw it, I just put my finger. All of a sudden it went to the lake. Now I have to go figure out where it is. So it went away from me. Now the data strategy has to be. You want to bring back the data to the user where it belongs. And that is a huge undertaking itself. So rather than sending the data lake, how can we bring it? And that's where strategy belongs that can we break the data into domain specific data, for example. So if I work in the shop floor, I'm not interested in supply chain, I'm not interested in something else. Can you give a context to the data? Just I understand myself bring it closer to me. If you just show it to me, I understand what exactly it is because I don't have to search for that. So you make a data more specific in a democratic way that I understand really well. Don't give me the c ocean me to float and try to find out to do that. And also of course the data governance is another challenging piece. So we don't want to give you access to everything to everyone, right? So if you have a con, if you have a give the context of the data, we talk the metadata, you should be able to control that. Because if data belongs to certain plants, certain site now I can say hey, you are owner for this particular place and I can assign attribute everybody else is basically out. So I think the metadata and giving context is a really important factor to make the data governance much easier to do that. So answer your question. It's not easy, but it's really critical piece. And their technology now is enabling in that space. Actually there are multiple technologies coming up how to give the context to the data things of that nature.

What are the first steps that you want to formulate as part of a data strategy to ensure that at least you're on the right path when it comes to data governance and data quality.

So if you think about how do we get the data, data is nothing, but it is outcome of a process as the process is happening. So think about, we are manufacturing medicines. So as the process basically moves from left to right, underlying all the data is getting captured. So what is happening is that as the process moves, your data is generated in the process. Now what is happening is the region that we have a lot of gaps because your process is broken. So if your process is broken, that's when you delink your data set and that's where you have the challenge. So the first thing is that you really need to look at your business process a high level. You don't have to go in detail and you want to make sure that your processes are linked together. So your data, underlying data is linked together as well. Now you can say that, yeah, that's possible, Vijay, we can basically do that. But we are talking multi site, multi country. How do we do that? Right. And that's another player that, that's why we said the metadata is really important to do that. So when you define the process, you almost literally have to give more like a process id, site id, a product ID or whatever the other metadata is, give the unique definition to each data element and you give that uniqueness to when you capture the data, each piece of that, and that is how you can segregate and put the data governance overall.

So we talked about, how do you make sure that your data is at least properly governed and you're on the right way to creating a high quality, standards of high quality around data. Now the next step of this and the next piece of this is data analytics enablement. How do you make sure that people are able to fish for insights in an easy, straightforward and streamlined fashion? Right. So we've talked about setting this foundation, but ultimately you need to enable folks to work data with real time. I'd love it if you can walk me through what organizations can do as part of their strategy to enable people to work with data at scale, or what are the different tools, analytics enablement that they need to think about to be able to operationalize data at scale.

Think about, there are two, maybe three buckets actually. So the people who are very technical people, data scientists, right. And then you can. So they're purely technical. We can talk about that in a moment. The second category of people is the citizen data scientist. So these are the people who are not necessarily a trained data scientist, but if you can give them a tool, we have a concept, AutoML. Right. So AutoML is really great way to get started where you don't have to understand the underlying technical details of the algorithm, but you can visually you can really look into that and you can basically drive the value. And the third element in the bottom is that people are not really into that much of data science. So can we train them, come and upskill them are very basic, you know, tools, whether it's a spotfire or qlik sense to do that. So I think in order to move the needle in upskilling people almost, you have to take these three categories of people to bring to the next level of maturity. Right. So having a training, a mentorship program and training programs an ongoing basis so we can talk more in detail is really key. Auto ML is a really great capability and in fact in Merck we are, we are pursuing that avenue. Data IQ is one of the tools that we are basically using for that and it serves the scientists also citizen data scientists as well to do that. And then of course a very technical side of it that how can we train the different languages and more upskilling, more technology, we train those people as well. So I would say that we need to focus all three areas.

Yeah, and what about kind of the descriptive analytics area to a certain extent, kind of empowering the insight layer, not necessarily just the machine learning layer. Where is the room for business intelligence tools within that empowerment?

So I think tool like data camp. Right. We are using that tool really to in a data literacy program. We also have some of the minitab, for example, a jump. So some of the descriptive kind of tools that we are basically using to upskill people for the basic hands on kind of work to do that. So there's a group of some of the tools that we can bring to the bearing actually to upskill people for the basic fundamental as well going to execution.

We talked a lot about these different levers. Data analytics, enablement, data asset management, making sure that the infrastructure is there, who owns the strategy, how to set the operating model. These are all big strategic decisions that require a lot of change management, stewardship over execution of data strategy given the amount of these layers that organizations need to push and pull on. I have a couple of questions around execution. One, how should leaders go about prioritizing these different levers? Do you begin working on infrastructure first, then enablement or is it all in tandem and then which parts of the organization owns these different parts of levers. So we can start off kind of with the execution process and timeline. How do you prioritize these different levers over time and where does the work start when you go for execution?

You know, the culture is big part of that. Right. So one thing that we have really see is does everybody in the company sees data as an asset and how do they see it? So somebody hasn't worked in the shop floor for all their life, right. And if they have not seen how the data can deliver the value for them. So I think the education and literacy program is definitely something that you want to get started all across the company. Right. That doesn't need any. There's no predecessor for that. I think the more people are changing, the culture pieces is changing. I think you will. You see a lot of things fall into place slowly. One thing I noticed, my experience is the people are really powerful creature. We all talked about the different structure. I think that if I have to keep everything at the top, I will keep people at the top. If people understand the value, they are motivated, they can get things done. Even the technology is not there. So if I want to deliver the business for value for the business, yeah, ideally it will cool. If I have a platform, a model deployment, I can manage the lifecycle, I can version control all those things. But listen, if I'm motivated and if I know there's a value in this one, even I don't have those technology components, I'll still able to do that. It's going to be slow, but I'll still deliver the value. So I think what I would see is that putting the infrastructure we don't have to really make unless you put this, we are not going to deliver the value. So there's a long hanging fruit all across that we can deliver without having the heavy lifting infrastructure. And you want to drive the value, create the awareness here, the value as the people are maturing their thinking and maturing, even the business senior leaders think is changing. Now you want to scale it up, right? And that's when you can take your time really to put the infrastructure in place. As the maturity basically goes, I would say that it's not that unless you put everything in place, then only you can drive the value. I think you can start with low hanging fruit and just keep on building upon it until you get to the stage where you need to scale it up.

So in some sense there's a lot of value that can be done in kind of the culture, the skills level, the enablement and then over time you start scaling these different levers like infrastructure, governance and be iterative in your approach.

Is that's what exactly I'm saying. Yeah.

In terms of who owns these different parts. So we talked about the kind of who makes sure to a certain extent that all of these different pieces are being executed upon.

So we talked about business of data. So what I was referring to that is that's a data and analytics organization within the business. So this organization basically is part of the business reporting somewhere. The structure to do that we can call CDO or CDAO or call it any other names but that's basically self standing vertical within that organization. So for most part actually we've talked about all the different components. Majority of that is driven by at least owned by that organization. But the it comes in the place. We'll talk about that as well. So I would say it is the mixed responsibility. There's not one that everybody wants it. There's it has a role to play. There's data and analytics organization has to play. There's a business sponsor that they have to play subject matter experts and the business owners and business units coming to place. So I would say there are multiple owners. But overall I think the driving force is data and analytics organization. As we talked earlier.

Definitely. So we kind of highlighted in our conversation kind of the iterative nature to a certain extent of executing on a data strategy. How do you kind of enable that mindset shift as well of hey, data science is an iterative project. It's not necessarily something that you do. It's one and done. Data strategy is something that you continuously do basically as long as the technology exists and that you can take value out of it. So how do you adjust the process when executing the strategy and these data projects to account for the iterative nature of the data game to a certain extent.

So here's what I've seen work really well. There's no magic wand for that. How you do it convince people. So let's say if I am and I have done it almost every case very successfully so you don't tell you so what I mean by that you show the value. So let's say you got a concept and I took the one. I joined Merck, for example, I took one concept and I basically had the proof of concept for a solution that was really it is solving the business problem. So I said, I presented to the senior management this is what value is basically driven. They can clearly see that value that can be basically driven. Now I said if you want to scale it up this is what you need to make the investment. So I think it is the time where if you can see the value, if the stakeholders can see the value, they are willing to come with you. So you have to show the small success, whatever the success it is. And when people see the success, they are willing to come with you. The other element I would say is that I've seen is in the senior leaders. If you can see that, you can help them in their success, whatever the objectives. So if I'm a head of business unit, if I go in front of them that, hey, listen, here is something that I want to come and solve your problem. And if they can see it's helping them successful, then now basically they are on board to do that. So you have to take that element. So this whole showing and coming in front of them and showing the value, if you can sow a small value and they say willing to scale, I think you have with them. So it's a very gradual process, people understanding the value, more organic way. That's the one. But at the extreme side, if someone comes, says, you know what, we are going big on data, here is the money, just take it and show the value. That is not the scenario most of the cases, right. Every time a company is investing in something, they are looking for roi. And we know that ROI cannot be that easy for data projects. It takes some time. And if somebody is looking ROI in six months and a year, probably that's not practical. It's a gradual. You can go that way, or if somebody is willing to invest the big money, then you can go that way as well. But I would say it's a culture and mindset and showing the success slowly and delivering the value. I think you can bring business on board.

Definitely. Culture and mindset is super important when it comes to enabling analytics across the board, enabling the scalability of data science and any data strategy. Right. So I'm excited to really dig through that. Of course, the big elephant of the room that we're talking about here is data culture. The majority kind of organizations have yet to crack the data culture problem. I'd love your perspective on not only the importance of data culture and how it enables data strategy, but how a lack of data culture stops data strategy.

Oh, my God. That is such a critical question, I cannot tell you. And I had to navigate this and I think to a very high rate of success, I was able to get in this. So there are multiple components to it. So when you talk about the culture, culture is a big word, right? It may Mean different things to different people. So let me break it down what I mean by that. So the one of the element is if you start the senior management we talked about earlier is senior management thinking the old way that hey, somebody else owns the data strategy, not us. It traditionally has been a kind of driver for IT solutions. Now if you can shift that mindset that yes indeed, data belongs to the business strategy, right. If you conceive that's a huge change, that's a big undertaking itself to do that. So how do we make that happen at that level? Senior management do understand the value of the data. But I think understanding and doing are two different things. Right. So can we convince them in that? And again, we talked about the showing the success is that. So that is at that level, middle management, again, if we can show the part of the success that somebody can see, wow, yeah, I can do this. I never imagined and I have multiple examples where people never imagine that's possible using the data. But once we saw the small prototype, it's a well, how about this? How about that? I love to do this. Now they're part of the journey now to do that. Now they're becoming part of that. Right? So I think coming back to that piece, showing the success and coming with that and going hand in hand, I think that is really big component of that. The other thing I touch upon that so how do we upskill people? So we. One of the things that we did in Merck, we ran hackathons. Now hackathons is you take some business, most challenging business problem and bring people together and you solve the problem. One of the thing that I brought in in that concept was I brought in participants who are not going to contribute, they're just going to watch. So in each team I just put together business folks who have some interest in data. I said, you just watch come to this hackathon. These meetings just observe who's doing what really and how they're thinking, how they're talking. And I cannot tell you is that after that hackathon was over, these folks wanted to be engaged part of the data science initiative going on. And they will volunteer their time to work on other projects to do that. So how do we go and create the literacy and the value of that element of that. So I would say it takes time. It's a gradual upskilling. We also have some of the tools like data camps, minitab or jump. How do we bring organize on a regular basis the training and awareness program for that as well. And the last component which is really what we call a digital mentorship. So one of the things that we implemented here at Merck is that there are people who are willing to switch their career and they want to learn more about data. What can we do for them? So we established a digital mentorship program where we identified the people who are willing to mentor some people in data field and then we surveyed the people who are willing to learn things and we basically paired them up and we had three months program, actually highly, highly successful program and we're able to scale it up, 20, 30 people in one batch. Now imagine if you have that kind of scale and solving the problem and bringing people on board, be more aware how they can solve the problems. It is a huge impact. So we can scale it up actually even more. Have a digital mentorship program for that.

100%. Completely agree on your first element of showing a win to galvanize the data culture and to show people the value of data science. There's nothing like showcasing a prototype or a low hanging fruit of like hey, I just automated like 10% of your workflow and saved you like time to a certain extent to do something more meaningful with data science, to be able to showcase the value of data science and to showcase the importance of data. To kind of harp on your second point on upskilling, how large is the intersection to a certain extent between ups and data culture in some sense? Is data culture synonymous with upskilling and data skills or is there an additional cultural element, behavioral element that kind of can't be filled out with an upscaling program?

To a certain extent, yeah, so. So there are some mix of almost everything really. So there's definitely a. Not upskilling but there's the mindset itself. Right. So the mindset is really, really the other element of that and that's a totally different upskilling is probably will not change that. If somebody has working in certain way, then they are thinking certain way, they're making decisions certain way, how do we change them? So that's a totally different area that we want to handle. And hopefully I think if the company at the top level is pushing in that around and we're forcing people to think different way. So I would say, you know, leadership by example. So if you want to change people behavior, I think it's hard to coming from the top. And you don't have to just talk about that, you have to demonstrate that. So if, if I'm a senior leader, not only that, I I just talk about how useful data is basically leading by example showing this is what we have basically done. Making the investment that's really key. We can all talk about all data and analytics strategy. There's certain level of investment that has to be basically made to do that. If you can show that example coming from the top, other part of the organization basically come into the fold. So in any change this is what we're talking is the change management. Right. So we're changing people how we're thinking and their ways of working. So upskilling is a big component of that change to do that. Now in any part of the organization there are people who are motivated, they want to do things different. What I have done, if you want to galvanize the team of people, you got to go after the people who are really excited, bring them together, create the energy actually and one of the element that I did was when I joined Merck, we created something called Data science knowledge network. And what the idea was that you create the community of people, not only pure data scientists, but also subject matter expert. These are the people highly, highly motivated. They want to do things out of their regular job. They're not data scientists. Right. So how do you bring that kind of culture and you create an. When people see other people doing something different, they say wait a minute, let me see what other person is basically doing. So you can bring that way creating the network of people talking and showing the success, coming together, learn and grow and give them opportunity. I can tell you that I've given opportunity for people who are part of some of the hackathons as some part of this one. I said come if you want to have an experience, I'm willing to give you the project. So on top of a regular job and I've been able to basically mentor and coach these people and that's a hands on experience. That's a huge difference that can make in someone's career.

Yeah, that's actually how I got into data science at certain very similar dynamic where I was also working in organization. A non data science job. But I was very excited about it. I was got to be able to contact the data team and then yeah, that was my first experience with data science and professional setting and kind of harping on that last point that around community, I really want to expand into that. How important is creating community to galvanize data culture? And secondly, what are the most effective tactics that you found to kickstart the community of practice around data skills and data culture?

It is highly important if you think about outside world, if you create a product, what the companies are doing almost everyone, they want to create the community around their product. So if you are selling a pen or you're selling a phone, you know you got a software piece of software. Almost everybody wants to organize a group of people, community of people. What they are basically doing is that they're talking to making these folks, all the users of the system talk to each other and learn from each other that way. Number one, your product is what is getting out and you're getting a lot of feedback from that. So think about if I create the user community and a lot of people are talking about a lot of other things, can I get insight out of that? What are exactly they're talking now? That becomes ground for me to go and mine the data to develop my product to the next level. Same thing here as well. We can apply the same. We create the community of people. They're talking, they're developing and other people are basically looking into that. And we can, in this community you can find out what is important to them, who is facing what kind of challenge. So I think organically you're capturing the thoughts of people. When you bring the community, that's the best way to get that information. You will not get any other means really to do that. So I think that this community is the best way to data literacy program and doing that. And I'm happy to talk to anybody who wants to get some guidance how to create community. I know we have a limited time here, but I'll touch upon that as well. But if audience wants to really get some brainstorming on that, how to do it, I'm happy to share that. But primarily there are four components to any community. Number one is knowledge sharing. So when you bring there everybody at the different level of skill set, when we interact, the people are basically learning and sharing, right? Number two component is that problem solving. So we are coming together. It is not just we're just talking. Why don't we come together, take a business problem on top of our day to day job and come together and solve the problem. That's the innovation hub actually you can call it to do that. So you can create the community to solve the real problem. Hackathon is part of that community as well. I can talk about. So number three is we talked about is that upscaling. So when you have that community, you understand who needs what kind of skill. You can capture the data now based on what they are saying, where the feedback is. You can create your training program, our workshop for that kind of skill gap that you might have. And the last component I would say is the digital mindset. So when you people say coming when you look at things, how people are functioning, how they're working, you get more. It's more like a snowball effect basically to do that. So community around that as it get bigger and bigger, they do that. And now when you build the community, you want to make sure that that community is represented by all faction of the business unit. It's not like some data scientists sitting in their room. You want to bring the sponsorship from the different business function, they are promoting it, that people need to participate in that. And also they're participating in hackathon. Are they participating into learning programs? So there's a whole development piece that can be used by engaging the different business functions.

That is so awesome and so comprehensive. And I know that we're almost reaching out of time, but I want to end with one final question. Vijay, do you have any final to action before we wrap up today's episode?

No. Thank you so much. I can tell you the folks data journey is not easy. I would say that all the component that we talked about that those are the key critical component, there's no one fixed formula. I think you need to consider some way and depending upon your situation, you can pick and choose where to basically start. Culture is the biggest element of that, I can tell you that. So if you can work on that, it's slow process will get there, but it's going to take a lot of effort and speaking up. Right. So if I'm data scientist, if things are not working out, then you got to speak up and see what do you need? Do you need to make it more successful? Right. So I think you have to create the voice of change in the culture element of that and upskilling. I would say the people are the highest element of that. So if you can empower people, I think you can get a lot more done that we can't imagine even what people can accomplish if they're empowered to do things.

That is awesome. Vijay, thank you so much for coming on Data Framed.

Thank you so much, Adam. So nice talking to you.

You've been listening to Data Framed, a podcast by Data Camp. Keep connected with us by subscribing to the show in your favorite podcast player. Please give us a rating, leave a comment and share episodes you love. That helps us keep delivering insights into all things data. Thanks for listening. Until next time.