This episode features an interview with KJ Gupte, Data Science Lead at Tradeshift, a cloud-based platform for supply chain payments. KJ has had nearly 15 years of experience in the industry. Before Tradeshift, KJ served as Data Analytics Manager at PwC where she led a team of engineers and developers in making data-centric products for customers like Apple, Google, and HP. She is also a graduate of the Harvard Business Analytics Program, and says her experience in the Data Science Pipeline and Critical Thinking course felt like rewiring her brain to think about data in new ways. In this episode, KJ discusses translating technical data for her stakeholders, building credibility in data among the C-suite, and being a trailblazer in the supply chain economy.
This episode features an interview with KJ Gupte, Data Science Lead at Tradeshift, a cloud-based platform for supply chain payments. KJ has had nearly 15 years of experience in the industry. Before Tradeshift, KJ served as Data Analytics Manager at PwC where she led a team of engineers and developers in making data-centric products for customers like Apple, Google, and HP. She is also a graduate of the Harvard Business Analytics Program, and says her experience in the Data Science Pipeline and Critical Thinking course felt like rewiring her brain to think about data in new ways. In this episode, KJ discusses translating technical data for her stakeholders, building credibility in data among the C-suite, and being a trailblazer in the supply chain economy.
Quotes
*“I think what has helped me drive success is being extremely collaborative with the team and extremely transparent. Because many times when you are transparent and you exchange your thoughts, that's where you get the most gains. So I might be thinking about data from an angle and the users might be thinking about it from a very different angle. And a lot of exchange needs to happen. Because not everybody looks at the data in the same way. Everybody comes with a very different skill set, whether you are an engineer, whether you are a data scientist, whether you are a CEO, CFO, everybody has a core skillset that they bring to the table. It's not always the same or uniform. So collaboration and transparency is always the key.”
*“We are absolutely data-driven right now. I'm very glad I was hired at the point that I was, because that's when we were scaling up. And that's when we were sitting on like chunks of data that was so potent that it was just a matter of using it. And in fact, we are sitting on buyer and seller supply chain data, which is the core of the supply chain disruptions that are going on. So we are in so many ways trail blazers to the whole supply chain economy.”
*“The leadership knew that there was a lot of work that needed to be done in data. So my core job was to actually convey that message. Initially I spent a lot of time giving presentations on just what we had, not going into complexities. But, um, presentations around, okay, this is what we have. This is the story that data is telling. And just making people curious, you know, making people interested about data. Because if, if your data is not telling a story, it is just numbers.”
*”When it comes to data, things are needed as of yesterday. And the moment you get data, it is stale. You have data right now, but it has gone light years ahead. So you have to be extremely fast and evolve with the data. So that was definitely a challenge, but what helped me is I had a lot of visibility. Right from the time I was hired, I was directly working with leadership. I was working with engineering to see the lay of the land, I was working with my CEO, my CFO, I was working with the head of engineering to see what they were talking about, to understand the language.”
Time Stamps
[8:34] Driving success through collaboration and transparency
[14:28] Tradeshift's journey to becoming data-driven
[19:05] Getting visibility on the freshest data
[24:08] How to build credibility and trust in the data from company executives
[29:24] How to deriving more value for customers
[34:47] Under Pressure: How KJ Gupte makes difficult decisions
[38:39] Advice to budding data scientists
Links
Thanks to our friends
Truth Be Known is brought to you by Talend, a leader in data integration and data integrity, enabling every company to find clarity amidst the chaos. Talend Data Fabric brings together in a single platform all the necessary capabilities that ensure enterprise data is complete, clean, compliant, and readily available to everyone who needs it throughout the organization. Learn more at Talend.com
Producer - Hello and Welcome to Truth Be Known, stories about how modern data leaders seek truth in an uncertain world. Today’s episode features an interview with KJ Gupte, Data Science Lead at Tradeshift, a cloud-based platform for supply chain payments. KJ has had nearly 15 years of experience in the industry. Before Tradeshift, KJ served as Data Analytics Manager at PwC where she led a team of engineers and developers in making data-centric products for customers like Apple, Google, and HP. She is also a graduate of the Harvard Business Analytics Program, and says her experience in the Data Science Pipeline and Critical Thinking course felt like rewiring her brain to think about data in new ways. In this episode, KJ discusses translating very technical data to her stakeholders, building credibility in data among the C-suite, and being a trailblazer in the whole supply chain economy. But before we get into it, here’s a brief word from our sponsor...
And now, please enjoy this conversation between KJ Gupte, Data Science Lead at Tradeshift, and your host, Rob Norman....
Rob Norman: Welcome to Truth Be Known. I'm Rob, and today I'm joined by KJ Gupte. KJ, it is great to have you on the show.
KJ Gupte: Thanks Rob. It was really an honor to be invited on the show and talk about my favorite subject. So yeah, very excited.
Rob Norman: Excellent. Okay. Just for those who may not be familiar with you and your work, can you just give us a little bit of background?
KJ Gupte: Yeah, absolutely. So, I'm a Senior Data Scientist Tradeshift we are a late stage startup. In fact, we just closed our funding round for 200 million. So very excited about it. We are a SaaS company. Yeah. Thank you. We are a SaaS company. So been here for last five years in the role of the Senior Data Scientist, now I'm the Data Science Lead. And I've been a manager in the data analytics and forensics practice at PWC before I joined Tradeshift. And before that I was a software engineer for Infosys. So it's been a while I've been in this field, like almost 15 years now.
Rob Norman: Fantastic. And just for those to get a bit of context in terms of what Tradeshift does. So congratulations on the funding, that's amazing. But what does Tradeshift actually do?
KJ Gupte: We are a supply chain platform. So, our customers are mostly buyers and sellers or kind of. Interacting with each other via invoices and orders. That is our primary focus. Of course our platform does a lot for sellers as well. We have a lot of more products and incubation, but in a nutshell, that's what we are.
Rob Norman: Gotcha. So basically a trading platform.
KJ Gupte: Absolutely. Yeah.
Rob Norman: Got it. Okay, fantastic. Can you tell us just a bit more about your role? What's your role as a data scientist? Where do you spend most of your time? Could you give us just a little bit of insight into your role, what it entails whether you have a team, just give us a bit of background to that.
KJ Gupte: Yeah. So as a data scientist, my role has evolved a lot. So initially probably what I'm doing right now is very different from what I used to do when I started off. But as a data scientist, that's what you do, right? Like you understand data. Mostly you have the kind of skillset to understand data that most of your users and your stakeholders don't. So your primary focus is to actually drive that understanding to your audiences in a way that they understand. So I get technical when I'm understanding data, but then I have to define the data elements in a non tech technical story kind of a way to my audiences so that it makes sense to them. The reader has to tell a story. That's really the, the nutshell of my role. Right. But initially I started off with, oh, understanding the data landscape of the platform, understanding where the data's coming from, understanding the anomalies and you know, what the issues could be in, in the data, understanding the infrastructure, and then coming up with key performance indicators that could be used for, you know, highlighting specific aspects of the performance of the platforms. So define it with the user. So Cape KPI that works for one team might not necessarily work for another team. So you have to be in that focus, like you have to collaborate with your users and see what they're looking for, define that, and then model that in the data that you have. So I spend initially a lot of my initial time in understanding the data. Then I spent a lot of time collaborating with my teams to define the right kind of metrics. Then I spent a lot of time with data engineers to come up with the right data, to put in the biplane before I start modeling it. And then, then my core job was to write code on that data too. So, you know, to make that data kind of tell a story and then deploy it to my users. Super interesting. And are you doing this with a team underneath you or you solely responsible for for the data science at Tradeshift?
So I am an individual contributor in many ways, but that said, I also have to kind of lead the project many times because when you collaborate with people, you kind of also ask for help from them. Right now, I'm sitting with finance and leading a lot of operating metrics, right. So for that, I will have to constantly work with finance to see what they are looking for, if they have any issues that they are seeing, if they are even understanding the story that we are trying to tell, right. So it's kind of like both ways. It's like, it has to be a two-way street. If that answers your question.
Rob Norman: Yeah. Yeah, no, that's great context. And I think, you know, just segwaying into the end users, cause you mentioned finance there. So you have your business stakeholders that you're you're engaging with and helping them to understand the data, but also help helping you working with them so that you understand what they need, right. So, so your business units are principally finance, is that the function you primarily support or is it broader than that?
KJ Gupte: No, it is way broader than that. So right now it is like, It is finance on PayPal, but like, it's kind of like a dotted line between product and finance data. It's always that way. So initially when I joined Tradeshift, I was part of engineering, because a lot of work needed to happen within engineering to make the data available. So my primary focus was to interact with most of the engineering and product manager of teams there. Then I moved into strategy, because then we had data that we had to play strategically to, you know, drive decisions. So I did a lot of work within strategy for projects that were driven by strategy. And then now being put in finance, again, my role still is cross-functional and still working with all the teams, including customer success. But because right now we are in a place where finance is driving a lot of metrics, we also were seeing these metrics for business use as well. And also these are the metrics that we are presenting to the board, it made most sense for me to like sit under the CFO and like work with these numbers, but that actually always evolves and changes. So it's not like a fixed role. As I said, data scientist is a very like, you know, broad role.
Rob Norman: It's a broad role. Yeah, absolutely. And and when you're working with finance and you've already mentioned a few of these, in order for a data scientist to work effectively with their end users their stakeholders what would you say are the key things that will drive success in that relationship?
KJ Gupte: The biggest thing is to be extremely open and collaborative. So that is something that it's has been challenging, and most of the software companies where, you know, engineering is siloed, individual contributors are siloed and then one does not know the other. I think what has helped me drive success is being extremely collaborative with the team and extremely transparent. Because many times when you are transparent uh, and you exchange your thoughts, that's where you get the most gains. So I might be thinking about data from an angle and the users might be thinking about it from a very different angle, right? And a lot of exchange needs to happen. Because not everybody looks at the data in the same way. So when there is an exchange of information, when there is exchange of perspectives, that does vary, you actually drive the actual matter, the actual subject. So that has helped me in my projects a lot. Because everybody comes with a very different skill set, whether you are an engineer, whether you are a data scientist, whether you are a CEO, CFO, everybody has a core skillset that they bring to the table. It's not always the same or uniform. So collaboration and transparency is always the, the key.
Rob Norman: Yeah. And it's really interesting that you're mentioning how you know, business users come at this sometimes from a different place when they're looking at data than yourself or than data scientists will look at it. So getting everyone on the same page, I think, is what you're saying, right? So you're all speaking the same language. You're all looking at the problems in the same way.
KJ Gupte: Yeah. Like to start with, we are speaking in different languages. Then when we start exchanging ideas, that's where we come up with like a common language. And then we are like, okay, now we get it. Now we are, you know, pursuing the common goal.
Rob Norman: Yeah. And how do you then translate that back into the engineers? Okay. Now, the requirements and business need is understood. Now, got to turn in towards the actual data itself, work with the data engineers, make that data available, visualize it in a way that our, you know, stakeholders can understand. How do you drive success so that ultimately you can deliver on what you've agreed with your business stakeholders?
KJ Gupte: So, one thing that has helped me in the past is I was one of them. I started off as an engineer. I started off writing chunks of code for Qubole back then. I don't even know if we would write that in that language, but the one thing that helped me is experience of wearing their hat. So experience from where they were coming from. So it made a lot of sense to completely get away from the non technical aspect after I had the problem statement, and then going to a mindset where everything was going to be technical, everything needed to be translated into technology, whether it was code, whether it was infrastructure or whether it was just the problem statement. So experience is what helped, having technical knowledge definitely goes a long way.
Rob Norman: And if for the audience that's listening, if there were some sort of practical tips in terms of how you might recommend ensuring success. So, so for you, success has come from experience makes a ton of sense, and it's a real advantage that you have in being able to work from, from experience in the sense that you were a data engineer, so you understand what they need. But if you are wanting to give recommendations for those that are data scientists that are turning back towards their data engineers to, to translate that business requirement, are there any additional things that you would say, yeah, these are key success points that they need to,
KJ Gupte: Yeah.
Rob Norman: that you recommend them then following.
KJ Gupte: Yeah, for sure, because, you know, data and like the entire data industry has evolved so much that a lot is being made available or being made easy for users, you know? So even if a few don't have a technical background, if even if you just go into the weeds of data to understand how data is, like, saved, how data structures are created and, you know, how, data is housed within a platform, and then just like use data sets in your day-to-day, have an analytical mindset and think about it in terms of, okay, so probably this is if I had data for this problem, how would I solve it? If you just bring it into your mindset, all you need is an analytical mindset. Because that's an engineer's mindset, right? If you're going to that like in that direction, it makes it very easy for you to approach an engineer and talk with them. Because they literally, they kind of they're in their box thinking about things in very analytical and engineering tones. So when many times when they're interacting with business comes from a totally different place. So it helps. And again, I don't know if I'm answering your question, but just to give you an example. Sequel is the most used language right now. And it is really not as hard as it used to be. Like all there are base tools where, you know, you can understand core sequel foundations. And even if you understand core foundations of the lay of the land of data, it's easy for you to approach data with engineers, you know what I mean? So it's kind of very easy like there are SaaS based tools, there is Tableau, which is kind of a a great version of Excel where you can kind of literally do small things in a big way. So it's a matter of understanding these tools being the great picture about data that was not available in the past. There are many SaaS based tools these days. So if you bring these tools into your day to day, if you just, you know, use them to solve like small, simple problems, it will make you very data savvy, which will help you to understand or talk in that language.
Rob Norman: Makes a ton of sense makes a ton of sense. How far along the path to being data-driven is your company? Because everyone's on that journey. So I'd love to hear from you. Would you say your organization is data-driven now? And if so, how did you crack the code? And if not, what would you say is missing?
KJ Gupte: So we are absolutely data-driven right now, like as we speak. I'm very glad I was hired at the point that I was, because that's where we were scaling up. And that's where we were sitting on like chunks of data that was like, so potent that it was just a matter of using it. So we are absolutely you know, data savvy right now. And in fact, we are we are sitting on buyer and seller supply chain data, right. Which is like the core of the the supply chain disruptions that are going on. So we are in so many ways trail blazers to the whole supply chain economy. And we kind of, I'm glad that we are making use of the data and we have identified that our data is, like, meaningful enough value. But it was a journey to get there, you know, because I think right now everybody wants to be data driven. But by the time startups get to a place where they want to be data driven, they are already at a place where they are scaling up. The product has evolved. They have users. Now it's just a matter of scaling up. If at that point, if you start becoming data driven, it is a little too late because most of the data that you want comes from your foundation, from your code infrastructure. So if you are kind of trying to make decisions out of data, gather data at a scaling upstage, it's kind of hard to make foundational changes for your reporting. So I always recommend that I anytime you think of a new product, think in terms of data. Think about, you know, what is it, how much of it of data you can gather from your product, because that is what is going to take you a long way. Again, but going back to your question, are we data driven? Absolutely. Like I think of that's the core of our organization, our leaders, our leadership is very data heavy, so they focus on the teams to be. to have data skills at least basic data skills for their teams to make sense out of data. And now we are also kind of being asked data questions by like the world economy forum. We have our world rate index going on. So, we are at a good place when it comes to data.
Rob Norman: That's amazing. And yeah and to a degree, it sounds like, the company knew it needed and wanted to be data-driven and it's sort of part of its DNA, but you had to be there to drive that journey towards where you are now, which is data-driven.
KJ Gupte: Absolutely. So I I was like, I would say the first or the second data hire. So there weren't any data engineers or data scientists, but like the leadership knew that there was a lot of work that needed to be done in data. My core job was to actually convey that message. Initially I spent a lot of time giving presentations on, you know, just what we had, like not just going in, not going into complexities, but presentations around, okay, this is what we have. This is the story that data is telling. And just making people curious, you know, making people interested about data. Because if your data is not telling a story, it is just numbers. And you know, many people who are non technical are really thrown by numbers. So I had to spend considerable amount of you know, chunk of my time and building stories, using Tableau, kind of driving insights from data, asking questions, making people ask questions. Many times I used to always be referred as the professor, the data professor, because that's one hand that you really have to, you know, bear when you're the torch bearer of data. And you want people to understand what you're saying. So I think that is part and parcel of every data person's job, whether the person is a data engineer or data scientist, or, you know, even director of data, they all have to have the professor's heart at some point to, you know, drive the data message.
Rob Norman: Yes. Yeah. A hundred percent. And it sounds like, your leadership knew they wanted to be data-driven, but they didn't know exactly how to be data-driven and it was up to you to help them to understand how to be data-driven, the art of the possible. And then also establishing what are the priorities? Because if they know what the art of the possible is you've conquered that barrier, which I think in some organizations can even be that. It's like. You know, really pursuing the desire to be data-driven. I think that's probably less and less these days, but when you've got that situation where you are acting as the professor, how did you align around some key initiatives and prioritize certain things over others?
KJ Gupte: So that is of, that is a biggest challenge, right? Because when it comes to data, things are needed as of yesterday. And you'll know the moment you get data it is stale. You already have data right now, but it has gone light years ahead. So you have to like be extremely fast and you have to evolve with the data. So that was definitely a challenge, but what helped me is, one, I had a lot of visibility, which was great. So right from the time I was hired, I was directly working with leadership. On the one hand, I was also working with engineering to see the lay of the land, but I was working with my CEO, my CFO, I was working with the head of engineering to see what they were talking about, to understand the language. And once you have that kind of visibility, you very immediately start to see what your audiences is looking for. You know, these were the leaders that were driving the company, right. So that helped me to look at the picture from their perspective. Every time I came out of a meeting with the leadership, I was not just a data scientist with two eyes. I was actually a scientist with like six, six sets of eyes, right. So that definitely helped me. So one of the, one of the challenges is to prioritize once you have like a chunk of projects that, you know, a lot of work needs to be done the first thing you look for is impact. Impact goes a long way. So sometimes there's low hanging fruit. You can make like, you know, you can make nice reports out of it. But they're really not impactful. Like the KPIs are good, but you don't really know how to drive them. So it's like the leadership tells you. If the leadership is telling you that, okay, this is the metric that everybody's looking for, that is the metric to go for. And many times it takes a while to come to that place. So you have a lot of other metrics going on, but then there is the ski metric, which is the yard not star goal. So in the past, there has been one project where know, it was great, we had a metric, we had a reporting model for it, but it was not very flexible. It was not answering all our questions. And what we realized is if we build a new model, and if we kind of make it flexible, that is going to go a long way which was not a low hanging fruit. We had to keep a chunk of work at the backburner just to give traction to that project. Are you able to share what that is like just to bring it to life?
Oh yeah, absolutely. So like our right now our key performance indicator and just one of them is the cross marginalized value, which is like the total dollar spent between the buyers and sellers. So that always gives us a story of where we are heading, you know, like what kind of transactions and what are the, who are the buyers and what are the markets and stuff like that. So we always use that as like one of our metrics that leads the direction in which the platform is heading. We had a great model for it. But that model was only answering basic questions. We were not able to get into the root of, you know, whatever issues, we were not able to get into the root of why there were anomalies or what were the outliers doing differently. So it only made sense to recreate the model, which was like starting from scratch. I knew that once we do that, the impact will be even greater, because probably the GMV was answering catering to five people. But if we changed the model, it would have catered to a hundred people. So what we do, so the impact of the new model was way higher. But it helps. So what we did was we just kind of put everything else Uh, or on the stand still, then we just decided to recreate the whole model. We put all our heads down, all hands on deck. We took some time. We tested the model, but once we were able to get there, it helped to solve so many problems, it made things flexible. It solved a lot of problems that we were just, you know, pushing for some other time, pushing for some other time. So those are the projects. And for that prioritization is focus, because say we would have not prioritized that probably even today, we would have stuck with the clunky model and we would still have bins. You know, we would still answer the same way that we were. So I think those things like I said, impact is the number one key to prioritization. Second is like transparency. So my leadership was very transparent with me about their goals. The leadership was transparent with the company about, you know, what they want to do. And that way they got help from all the teams. So it's always, like I said, it's a two way street collaboration is like, you know, it was a long way.
Rob Norman: Totally. Super insightful. Sometimes when you put fresh data in front of execs, they will say, where does this come from? Can I trust the data? How do you address that? How do you make sure that what's in front of them it's like, yeah. Yeah, you're good. You can trust it.
KJ Gupte: Oh, my God, Rob. Like I think that is like, that was the only thing that was said to me, the initial, you know, my initial time in the company, right? Like, because initially there was nobody to put the data in front of the execs. And then there was me. And then every time I did that, they were like, okay. I mean, how can you be so sure. You know, are we really like, is it really that clean? And so it's not like a silver bullet where I say, and then they believe, right. It's not like that. So even like the way I mentioned my earlier project, it took a while to, to test the model, to see the trends in the model, what really builds credibility and trust is the trends. If you see that, okay, the trends are kind of stable, there are fluctuations, but most of the times the trend is stable. If you look, if you take a subset of their data, and if that data is also telling a similar story and then also highlighting anomalies, that also is credible. So it's not always one chart that makes their data credible. It's always kind of supplemental information alongside it, right? So this is the chart. But to make sense out of this chart,. This is what you have to see. A subset of these are the anomalies. This is what they did. And this is what, you know, this was the market. This was the market condition around it. So a storytelling approach is what worked for me. Initially, there were a lot of failures. So initially my initial meetings failed because I was very confident about my data, but the audiences were not. So I used to always be sent back to, you do okay. Yeah. Let's try again. So there were failures in my projects as well, but the storytelling approach helped. And while I was building the story, I was also getting a little gaining trust in my own data. So for me, it was like, it helped both ways because I was the more confident I felt, the more confident my audience felt.
Rob Norman: And that underlying confidence in your data, was that principally due to the test and learn approach you took on the model or were there underlying data governance foundations that you put in place to say, yes, I can definitely trust the data in my model.
KJ Gupte: Yeah, absolutely. So initially I was working on a model that was created by a tool. So there was a tool and there was like a local further tool, which kind of did not help me with building my own, gaining my own trust. Because every time I did a deep dive, I found something. I went down a rabbit hole, I found something and I was like this does not look good. But once I started, once I recreated the model from scratch, I knew that I had written my source code that was actually doing the leg work for the data. That source code was then, we had built it together, right? So the engineering team had, was also part of it. They had given me a lot of direction in where the data was going, how, what were the issues while you're not gathering data and stuff like that. So once we together built a source code that we tried and we tested and we peer reviewed like bunch of times, and then we put it in test for like almost six months to see whether our trend were kind of, you know, similar and well, whether our correlation analysis was similar to the past data and stuff like that. So a lot of work went to build that foundation. So like you said, yes, governance. Go more than governance, it was due diligence. It was a lot of analysis, a lot of like going down rabbit holes, finding things, writing them down. And then at the end of the day, we came up with a white paper. To write all the stuff that we had found. So now we have like following a solid foundation and documentation in our source code and in the testing that we did, which was all together that helped us build trust. And now that we have a lot in place, there is, it is very easy to convey that credibility to our leadership.
Rob Norman: That's great. So now you're in a situation of the company being data-driven. It sounds like it know it's data-driven across finance. What are the big buckets of where data is really playing a key role driving the business forward? You mentioned finance, you mentioned gross margin. You mentioned even building, data as an asset for external use and demonstrating the underlying data that the platform has that can be made available outside of the company. Those are two, are there other big buckets where now the data that the companies moved to that data-driven place and is leveraging?
KJ Gupte: Yeah. So like, now initially we started off with seeing how customers are engaging on the platform. Right? So most of the data that we started building was based on on their engagement on the platform. Now we are actually leveraging that data. And actually also making it available for our customers to see how they did. So we are actually using a lot of our data to provide to our customers and helping them with their analytics as well. So that is the key. The other key is now is, and I cannot talk much about this because it's
Rob Norman: Only as much as you can.
KJ Gupte: lot of this. Yeah. But yeah, We are kind of drawing correlations between between the finance or other, the revenue and the oh, you know, spend aspect, right. So like, what is the correlation between them? Like, do our customers meet the kind of level that they think they would when they sign up with us? Oh, you know, we do a lot of, we are doing a lot of modeling to see if there is more value that we can provide our customers based on their engagements on the platform. So now we are kind of focusing on data to make the journey easier for our customers, you know, to make it easier for our customers to be proud that they are on Tradeshift. Of course they are. But right now, like, you know, the customers don't have the level of visibility to their data as much as we would want them to. So the, of course we are trying to make that one of our biggest projects.
Rob Norman: So. Let's let's shift gears just slightly. I think we've covered a ton of ground and that was super interesting. You're a data science leader, at Tradeshift. What, what in your mind makes a good leader?
KJ Gupte: One is being extremely transparent about your goals and having like a collaborative mindset is what helps. The other thing is I would like to use the word empathy, but empathy is not the way it is defined. Empathy is basically in a way that you can think outside your own head, like many times as individual contributor's, data people are very much in their own heritage doing their own analytics. But to really drive success, you have to look at your world from the audience's point of view. Right? So the way I said, every time I came out of a leadership meeting, I had six set of eyes, not just one. So I think a good leader is one who can who has the ability to ask for their audience's eyes. Not really eyeballs, but you know what I mean, eyes and gain their perspective and really not be afraid of asking questions. Like, you know, what are you looking for? I'm thinking of this differently. What is your direction? How can you help me? And really being in that place of asking for help, instead of telling people what to do.
Rob Norman: Yeah. And certainly when, you know, I've worked with data science teams in the past. And I remember in the beginning when we were starting to, they were starting to produce a lot of data around performance marketing and, you know, delivering us insights. And we were starting to ask questions about, could we see this? And could we see that? And and it was a real collaborative partnership. But I also remember at some stage there was this shift from that position to actually the data scientist saying, you know, what I'm seeing in your data is this, and these are the trends that I'm looking at and have you considered it? It was almost as if the shift got to the point where the data scientists understood the marketing, the business, their stakeholder as much, almost as much as the stakeholder themselves when it came to the analytics side of it. Right? So that, that was an amazing shift that I saw. And it was almost like a level of literacy and understanding of the business side that was really mature. And that happened over time.
KJ Gupte: Yeah. And that almost always happens because you are so much into like problem solving there that you just happen to look at the data from their standpoint. Right. So, I'm sure what might have happened with their data scientists is initially they might have started off with their data. And then after they started seeing trends in specific areas, they got interested in that. When they got interested in that they started having an analytical mindset around that. And that's how, you know, they started viewing the audiences or the markets. Right. So. It's kind of the way you say you're collaborative, what helps us sometimes as a data scientist, you catch things differently than your users and you need to tell them this is what you need to see.
Rob Norman: Yes.
KJ Gupte: Because you are looking at the report for the first time. On, you know, the, on your Monday morning, you're seeing a ton of numbers. But the data scientist has worked on it for like past two, three weeks. And they have looked at the trends and a different way. So it's always like, okay, they probably, at this point, know what you need to look for.
Rob Norman: Yeah. So, so that empathy that you talk about is critical to move to a point of yes, collaboration in the beginning, but actually later on partnership, like true partnership in the sense of.
KJ Gupte: A true pro.
Rob Norman: Yeah. Providing insights that, you know, the business stakeholder would perhaps get to every time, but like, you can show it immediately because as you said, you'd been working on the data for three weeks and you now understand it, to a, I would say greater degree than the actual business stakeholders. So you can point them out and add value instantly.
KJ Gupte: Yeah. And the confidence that comes from it is like very contagious, you know? So it's if the data scientist is telling you something, you ought to just know that it's probably what it is.
Rob Norman: Yup. Yup. We're going to shift again, cause there's a nice little segment that we have in this podcast called under pressure.
Rob Norman: So, your role is, has constantly evolved. You mentioned that since joining Tradeshift, but obviously over your career at PwC can you tell me what's the most difficult decision you've had to make in your career and did data shape that decision in any way?
KJ Gupte: Yeah. Like I think of when I was approached by Tradeshift, I was a manager at PwC in their data analytics practice. Right. And data was like, has always been my passion. So when they approached me, it was a startup. I was not really thinking about startups as much. But when I looked at the product and the role, especially like when they, they hashed out that this is the role and this is what you will have to do in that role, I had done some of it in PwC, which was interesting. But working in, on a SaaS product and a SaaS product that sits on a ton of data that is so like, you know, not leveraged as much and you would be the one to do it. It was a hard decision, but I think that was data was a very big part of their decision. Like I, I looked up and then I saw all the KPIs that could come out of of data like this. And, you know, also the partnerships, I said, like the visibility was great. It was kind of a big risk because as a data scientist, you are on your day one, you have, oh, you are supposed to face your audiences. Like I remember on my first first day, first hour, I got an email. KJ, can you know, get help us with these numbers? So I knew the risks with the job, but again, like I said, data was like the biggest thing we would we had great infrastructure to start mining the data that we had created. I didn't know whether I was doing the right thing, but my career and my like expertise and data kind of led me to take that decision.
Rob Norman: Yeah. And it's always tough when you're affectively leaping into the unknown, but also in this case, you're, you're effectively pioneering to to to a degree, right? Because you're moving to a space which is completely new. And yeah, you didn't really have a blueprint to follow to, to any degree.
KJ Gupte: Yeah, you said it. And many times that happens because even now, if you look out in the market, right, like their data, the job of a data scientist, not really like certain people. You know, like people know that they need a data scientist, but they don't really know what they need their data scientist to do. Most of it comes from the data scientists, like, you know, they have to go in their data. So it was a very tough decision. But yeah I'm very happy about
Rob Norman: Yeah. And a tough decision, because at the same time you can see the opportunity. Cause how amazing is it to be able to shape it entirely, but also that's like, wow, I have to get so comfortable with ambiguity and the risk associated to that. I was reading the apple quote from the, the ads in the seventies, which is like, you know, when they were launching, I think it was a brand ad, and it was like, "Here's to the misfits, here's to the ones that break the mold," et cetera. Right. So it's, those are the ones that, that, that change to make change in the world. Yeah. And that's what the startup culture is about. Right? Like you live in the unknown and hopefully you, oh, make a change. And I think as a data scientist, you have a lot of potential for impact and to make a change.
Totally totally. Closing out with a couple of quick-fire questions.
Rob Norman: Best piece of career advice you've ever gotten.
KJ Gupte: I think you, you nailed that one with the quote. I'd like for data scientists I would say have an open. open mind. Uh Be ready to fail because like we were told that the Howard Business Analytics Program that I took that most of their data science projects fail. The, the ones that succeed are the only ones that you know about. I mean, it's, oh, there's so much learning at the end of it that you always evolve in your role in, in the company data and you know, and all the problems that you're solving. So my advice to all the budding data scientists out there and, you know, budding CDOs out there is to be ready to fail. And then learn from it.
Rob Norman: Yeah. I think that's great advice. And that is exactly how we learn. The best. We don't really learn a great deal from our successes. We learn when things go wrong and that's when there's a lot of introspection in there and then we move forward from there. So that's awesome.
KJ Gupte: And other people also start taking interest in you when you fail, Because they want to know, oh my God, what happened to you?
Rob Norman: What happened to you? yeah. Who do you look up to in the data space? Are there thought leaders, influencers that you keep tabs on social?
KJ Gupte: Oh my God. Yeah. So the one person I look out for is Professor Datar who is not on social media, I wish he was like, there are a lot of references to him, but he's not, but he was my favorite professor in the program and just the,
Rob Norman: That was at Harvard, right? That was when you did your Harvard program. Yeah.
KJ Gupte: Yeah. Yeah. That was at Harvard and I think he's like the epitome of what a scientist should be, not just are data scientists but a scientist. Right? Like he had that empathy or, like out of the box, thinking that he kind of brought out in us. And he actually made us answer our own questions, which is like a task. And I think the data scientist many times does that. Many times when you are like, you know, in the dark and you know, the blind is leading the blind, you kind of help your audiences too. So he is something that I look up to in his critical thinking, in his. In most of the articles that he's written. The way he guides businesses to approach data, to approach the people, driving the data. He has a lot of emphasis on people. And then once you identify your people, identify what ticks them and define, like, you know, what is their motivation that's what is your answer to most of your challenges. So yeah.
Rob Norman: So to check them out, it's Doctor Datar?.
KJ Gupte: Professor. Professor Datar, yeah.
Rob Norman: Professor Datar, D-A-T-A-R, right?
KJ Gupte: Yes.
Rob Norman: Yeah. Yeah. Brilliant. And he's Yeah. He's a professor at Harvard.
KJ Gupte: Yes. He taught our course on data analysis and critical thinking, which was the hardest course. It was extremely hard and he cold called most of us. So yeah, we will never forget Professor
Rob Norman: Okay, Amazing. And are there particular online sites you go to? Resources like places like online? I don't know, chats, communities, forums particular websites that you go to stay on top of your game? That other a lot of online data forums and discussions. The one that I go to very often is other I'm part of is the Howard business review discussion forum. So that forum is like really good to understand where the businesses are heading, what are they looking for? And, you know, things that we don't really think about, like anomalies, like we would have thought this, but oh my God, this happened. So discussions like those are always good in the Harvard Business Review Discussion forum.
Nice. Nice. That's a great recommendation. We could chat for so much longer, it's been awesome having you on, kJ. It's been a real pleasure. If people want to get in touch, reach out. What's the best place to get in touch with you? Is it LinkedIn or Twitter?
KJ Gupte: I'm very active on LinkedIn. Unfortunately I'm not active as much on other social media as much as I am on LinkedIn. So yeah. Find me on LinkedIn.
Rob Norman: Amazing. KJ, thanks so much for joining today's episode. It's been great having you on and for our listeners, really looking forward to the next episode. So see you all very soon.
KJ Gupte: Thank you so much, Rob, it was a pleasure.
Rob Norman: Likewise, cheers to KJ.