top of page

What mistakes to avoid in Data Science & Analytics?


MindSpeaking Podcast Episode 10 - Scott Burk, Professor, MS Data Science @ CUNY





▶️ Watch the podcast with video on YouTube:


Timestamps:

00:00 Introduction

00:24 Introduction of the guest

01:24 Who is Scott Burk in high school?

03:15 What were some of the important decisions that you have made?

06:37 What are the common pitfalls that organizations see or experience when they try to adopt this AI analytics?

08:36 Tips for companies or people that are trying to build AI from building data science

11:21 What is the difference between AI and machine learning?

13:30 Where do you see the market of data scientists takes by moving in in the coming five years?

15:30 Do you think people who have not learned how to code yet should still learn how to code?

16:38 Role of communication in data

21:15 To what extent do people students learn about communication skills and storytelling or developing a business mindset?

24:30 Dale Carnegie & Seven Habits of Highly Effective People

25:36 Something that is not timeless is Technology

27:36 It's all analytics

30:27 Where to follow Scott Burk?

32:06 Conclusion




🎙️Listen to your favorite channel:

Spotify Podcasts

Apple Podcasts

Google Podcasts

Other Platforms




Introducing Scott Burk

Gilbert Eijkelenboom

My name is COVID. I'm driven by curiosity, and my aim is to spread insights that you can apply your life starting today. So, let's do it. Let's start mind speaking. Today in a man speaking podcast, I'm talking to Scott Burke. Scott has a PhD in statistics. He's also a professor of data science at CUNY is 30 years of experience and has authored four books about data science and AI I'm asking him, what are the common pitfalls when companies try to adopt AI Data Science and Analytics, and how to make AI programs successful? He's also talking about presenting to non technical people. And I'm asking him because he has four children and even a few grandchildren, how they look at him. How do you see him as a dad that knows so much about technology? So without further ado, let's dive in conversation with Hi Scott, excited to meet you and exciting to talk to you today.


Scott Burk

Hey, Gilbert, good to meet you, as well. Yeah, I've been bored to this.





Who is Scott Burk in high school?


Gilbert Eijkelenboom

And we're using artists asking people where they grew up and what type of person they were in high school. So can you tell us a bit about that before we jump into the data


Scott Burk

so I grew up in central Texas in high school. I was kind of a rebel. I was very involved in athletics. Not that studious, however. Growing up in Central Texas, it was pretty easy to make straight A's. The curriculum was not overly difficult. So I had a great high school life. Yeah, a lot of fun. Memories. I wouldn't trade it for the world. I probably could have been more studious to prepare myself for the university. But it all worked out. I think I had a really good balance. So yeah.


Gilbert Eijkelenboom

You mentioned you were a bit of a rebel. Do you? Do you still see that side in you when you when you look at your career? Because it can be a very positive trait as well right to stand out to innovate?


Scott Burk

And let's that's a great question. So I I do see it as a positive side. And I think as you get older, you realize when you're being a rebel when you're young, and you're being a rebel, you don't even realize you're actually doing it. So. Yeah. So So it's better to be a rebel when you're a little bit more seasoned. And then when you're a teenager, for sure,





What were some of the important decisions that you have made?


Gilbert Eijkelenboom

yeah. So now you have much more self awareness about the rebel side that maybe sometimes pops up. Right, but there's something positive. It's great, great to hear something about your your background, and how did you end up in data? What were some of the important decisions that you have made or that you've worked? Well, I


Scott Burk

tell people that I started my analytics career actually, as an analytic chemist, which is certainly different than what we talk about in analytics today. But but the quantitative is always been there. I I went undergrad as a chemistry biology double major. And then my first big was analytic chemist, and sort of programming a lot at the lab that I was working for, and then got involved in computer science, and then from computer science. In the lab, started a graduate program in business finance. quantitative side of finance, not the qualitative side, capital markets, that sort of thing. So then I went to Texas Instruments, and did a lot of computer science. And then after about five years ago, Mitch wanted to go back to school. And that's when I really started focusing in to mathematics. And I, since I had my master's degree, I was able to teach at Baylor University and work on a PhD at the same time, and it was in statistics. So that's really when I started really kind of the nuts and bolts of my career, but I believe in lifelong learning the whole time and had that I was working professionally. I was I started a program. Actually a little bit of an interesting story. I went to a class for what's now IBM modeler. So it was SPSS at the time. And so the instructor we were, he said, we were we were learning SPSS, but he said, Hey, look, I've got this really cool software that we just bought Clementine and and it was the first visual visual platform with objects that you could drop down into Canvas. And so he goes, You know, this is a very expensive software. I won't, I won't put the number, but it was you know, in the high rates, you know, 10s of 1000s of dollars. He said, If you take a class at Connecticut State University, in the software for 200 bucks or something like that, and I said I may just enroll taking classes and ended up getting the, you know, degree and data mining. So it's really seen the statistics versus view of the world and so but again, I believe that tools are, are to be used. And you need a broad tool, chest to pick and select. So yeah,





What are the common pitfalls that organizations see or experience when they try to adopt this AI analytics?


Gilbert Eijkelenboom

right. It's a really interesting journey you've you've made and, and eventually, you wrote several books on data, AI, data science. And one of the one of the series is named, it's all analytics. Right? And that's correct. And you'd also talk about common pitfalls about that, that organizations see or experience when they try to adopt this AI analytics. What what are those common pitfalls, what do you see there?


Scott Burk

So one of the pitfalls is that organizations are very functional. Many organizations are very functional, right? And so you have different departments speaking different language. So really, you know, an example of Casey Panetta had an example where might be the department I think, in Spanish, the HR department might be speaking Portuguese. The you know, marketing department is English, etc. One So, one of the pitfalls is that every organization should have a baseline of illiteracy or analytics literacy, where people can communicate right because organizations are definitely moving towards the more analytics, right AI and analytics are driving a lot of ROI in companies and it has such a potential. Really, we're just scratching the surface of what we can do. So we need a foundation and be able to speak to that. And then really, there's there's three pillars, that kind of in any organization be successful, but one is the organizational design itself. How the organization, what the culture is of the organization to support that and then another piece is the data design piece. And then the third piece is analytics design. Those three have to be better fit together to be useful in organizations. Yeah,





Tips for companies or people that are trying to build AI from building data science


Gilbert Eijkelenboom

yeah. So I certainly see how those three pillars are important than do Do you have any other ways or tips for companies or people listening that are trying to build AI from building data science at Definity? What else do you see as important?


Scott Burk

Yeah, I actually have I could go on a long time on that question. So again, so the first thing is to have executive support, right. I've been involved with many different companies over the years. And one of the things that is missing sometimes is, you know, the corporate objective will be, you know, flavor of the month and they'll roll out things and when you when it comes to a you've got to ingrain analytics and AI into the culture itself, right. So there's a couple of ways to do that, to enable that. First off is really is at the executive suite level and we have a new book coming out. Specifically designed for the executive and leadership on that. So it's a little bit of a different thing. But you know, Design Center of Excellence, have a cross functional assignment structure in organizations. So you might have to have you know, the functional units themselves, but the best design is if you have like very low center of excellence, this is one design that you can have that matrix matrix seat. It's a matrix organization that combined across all those functional units right to get the COA you know, you can have best of breed and then you can embed those those people across within from a project standpoint. Another design is a center of practice. Some organizations are really doing this well where the design a it's not necessarily functional. It is more of like minded people coming together to solve and to educate. I just went back that you got to see the same language. One of the ways that you can help enable that is through a community of practice where you have functional groups coming in to Common Core that Common Core has very specific objectives. The leadership actually incentivizes that through the organization. And so that's another way that you can do it as well.





What is the difference between AI and machine learning?


Gilbert Eijkelenboom

So let's go ahead. Now what I see with many people in the business, you talk about speaking different languages, and it's important to understand what you're talking about, right? It's stable, and what I see here is a lot of AI and machine learning, floating around and many people are confused with the terms and what is difference? What simplest way of explaining the specific nation you've heard or used. distinguish the two.


Scott Burk

Yes, so there are a lot of terms that that float around in addition to those two, how everything from statistics, mining, machine learning, AI, etc. But the question is the difference between machine learning and AI right okay. AI is an overarching term. Well, I'm not gonna give you a simple answer, and maybe I'll get to a simple answer. So that's the 1950s. Right. So AI was developed in in the 50s. And it tends to be an overarching term, right? So it includes machine learning, machine learning is a subset of AI. So that might be the simplest way to put it. You know, all ml machine learning is AI, but not all AI is ml and so, AI can include much, much more everything from decision based rules, right. So, you know, rule where machine learning is about algorithms trained to for to answer specific problems.





Where do you see the market of data scientists takes by moving in in the coming five years?


Gilbert Eijkelenboom

Right, right. I think that's pretty, pretty clear and pretty simple. I've seen a lot of different explanations on the internet and everywhere. I think this guy is pretty close. But um, we mentioned a bit about the past about the 1950s, but also, to look ahead, not 20 years ahead, but let's look years ahead. Where do you see the market of data scientists takes by moving in in the coming five years


Scott Burk

we've already seen some great advances in Pat, you know, advanced pattern recognition, right image recognition, I think that's just going to continue to democratize right. So at the station, this group of engineers they were not trained data scientists. Obviously they were smart. had formal training but they sought to develop some really fancy libraries. It's because we have so much knowledge, it's available at people's fingertips, right everything through, you know, YouTube, the MOOCs that the massive online courses that are available, etc. I think the next advance some companies are doing is streaming, right so video, right So video is much you know, it's taking that image recognition and pushing it in. And the obviously have to have technology to support the video streaming, the information that's coming in the recognition. So an example of that might be, you know, smart cities where you have cameras along the streets that are identifying potential collisions or wrecks or problems. So I think that's going to be smart cities. It's going to be a big, big opportunity across the globe. So in streaming in general, right, so just faster and faster data with quicker and quicker decisions. I think we're going to see that over the next five years, plus people that never had the ability because of the lack of coding experience or a lack of background, they're going to be enabled to do things that they never thought that they would be able to do before in the next five years.





Do you think people who have not learned how to code yet should still learn how to code?


Gilbert Eijkelenboom

Do you think people have not learned how to code yet should still learn how to code?


Scott Burk

Absolutely. I think it's, you know, I, whether it be mathematics or coding, I think that this skill is useful, but more importantly, it's the way your mind starts to rewire. To think, right. I mean, you know, if you get into coding and everything, it's a beautiful, right. It's a way to get lost very easily because it's it's so well, it can be frustrating, but it can be very gratifying as well, right? You get in the zone, essentially. And so, yeah, I think it's a good skill, but it's also good just for your brain. Yeah, your mental process and making you smarter overall, outside of coding.





Role of communication in data


Gilbert Eijkelenboom

Yeah, it helps you to think that it is to me I guess, and, and I speak a lot about the other side as well as, combined with technical skills, analytical skills, also communication skills. What do you see as the role of communication skills in in data?


Scott Burk

Well, it's it's paramount, right? Because if you can't communicate results if you're, you know, analytics professional or a data scientist or modeler, there's no way that you can, you're not going to get very far without those communication skills, right. So, and there's different levels of those communication skills, obviously there's just the basic communications. But there's the storytelling, which is becoming very popular. We write about that. The dashboards, the visual, you know, especially business leaders, they they want to see, dashboards, they want to see see it visually, right. So, visual storytelling is is extremely important. But I'm also I, with a word of caution, you have to be careful that we're not learning and or we're not just cherry picking what the story, right we have to give the complete story as well. And I have a quote that I really like is torture the data long enough and it will admit to anything. And you could do that very easily. And in analytics and AI you can cherry pick, select the parts of the story that you want to tell leave out parts of the story that you don't want to tell, go into the boardroom and, and really kind of mess things up for the long term.





To what extent do people students learn about communication skills and storytelling or developing a business mindset?


Gilbert Eijkelenboom

I think is becoming a really big topic in wedding data and also AI you know what you do with the data how you represent it, because for sure, you can really mislead people and guide them towards the decisions they actually don't want to make. But just because you're showing the data in a certain way. It's it pushes them in a certain direction. Right?


Scott Burk

Right. Absolutely. Absolutely. And then the last piece is to make sure that the the math supports the assertion. What I mean by that is humans. The reason we love visual dashboards is the fact that over 1000s and 1000s of years we've we've evolved into being able to quickly recognize patterns and see things in in pictures. So that's, that's beautiful. At the same time. We also make up patterns that we think are in the data and you know, they're not really there. So, again, I think you need the mathematics and the statistics to say what is truly meaningful and what's not. And then a combination of the two. That just to earlier point that just make sure that you're telling the right story, right. I mean is it is it and there's a difference of statistical significance and practical significance. And along the side with that is we try to tell stories in our books, right? So instead of trying to make it just you know, here's AI, here's, you know, machine learning, here's whatever, we have little vignettes and the reason for that is people people love story. And it makes it more interesting. And yeah, one example is where Netflix developed the you know, the million dollar prize, they had a million dollar algorithm, they didn't they knew that worked. And so this is where the the objective significance or statistical significance pairs up against the practical significance. They never they never even used that algorithm. Because once they did the math that they knew it worked, right. It was a million dollar algorithm. But once they did the analysis of how much it would cost to put an engineer into practice, it just wasn't worth it. So you so you need both practical as well as, as mathematical significant.





Dale Carnegie & Seven Habits of Highly Effective People


Gilbert Eijkelenboom

The practical side is a big thing what I see with many Xi learning AI projects that they fail because people produce something that might be good on paper, but people just don't use it because they they don't understand it. Or they don't appreciate it or it's not in line with what they need or it doesn't fit into business processes. I think that's such a big, big important of big factor of being a data scientist also understand that side. We've talked about communication skills a little bit about data storytelling. You teach data science as a professor at CUNY. And I'm wondering, to what extent do people students learn about communication skills and storytelling, or developing a business mindset?


Scott Burk

So they they learn it within the projects and everything that they have in the different different classes. There is a there is a visual bi core specifically towards you know, developing dashboards and analytics and best practices, the ways that you want to do that with different tool sets. But kini is very much project driven. So, you know, we, you know, you have to present your results. And so I think that's probably where it's done more than it's really a practice area for getting into industry, right? Because you're gonna have to do that as and this is a master of science and data science. So these data scientists are going to have to present to their managers to executive leadership, their results and pitch their story right that pitch, you know, what, what their, what they've done and and why it's meaningful to the organization. So it's really kind of embedded it's other than the dashboard, visual bi course there's not a dedicated storytelling course. But again, it's part of, it's embedded in pretty much each class.


Gilbert Eijkelenboom

Now fantastic. It's great that they get such practical experience through that project work and, and practice those communication skills or storytelling or visualization. And I think it's a really important important piece that that people need to do develop.


Scott Burk

Yeah, you know, I mean, retrospectively, I didn't get a lot of that in my in my coursework, so yeah, we're learning we're getting better.


Gilbert Eijkelenboom

How did you catch out? I mean, how did you adjust or when did you notice a notice hate is important or I I'm missing a piece.


Scott Burk

So it was, it was Trial by Fire basically. And between a combination of Texas Instruments semiconductor company, and a healthcare company I worked for was really the healthcare more than anything else, because it was non technical people that you were presenting to so there was there. It really challenged you. And one of the things that I did personally and I would recommend it to anybody, is I took a Dale Carnegie class where I learned to do you know, public speaking you know, and how, you know, it was designed. They've been around for years and years and years. And basically, I never thought of it this way. But looking at it, essentially that that course is about storytelling. It's really about creating the presentation and there's different objectives. But if it were to be boiled down, it really would be a course about storytelling.


Gilbert Eijkelenboom

Yeah. Awesome. Dale Carnegie, I know has a they have a lot of really powerful courses. I really liked the book of Carnegie. It really changed kind of changed my career. I read it. When I started my career to the very first in the in the holiday, I went to Cuba, I brought a Book Seven Habits of Highly Effective People so I slept slightly different of course compared to what you're you were saying, but it's written by, by Stephen. Stephen Covey. Exactly. And it really changed my perception. I thought it was a book about business because it was about effective people, right? But it's more about how you look at the world and how you see other people. So I see quite quite some overlap in data skills that data scientists need and what is presented in a book and apparently also, what you learned in the Dale Carnegie course.


Scott Burk

Yep, those are two of my favorite books. And I try to get back to those every couple of years and read them again. Because they're just timeless. Just just gre