#30 Data Science at McKinsey

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 in this episode of Data Framed, a DataCamp podcast, I'll be speaking with Taras Gaurishni, a Senior analytics manager at McKinsey and head of data science at Quantum Black, a McKinsey company. Taras and I will be talking about the role of data science in management consulting, what it takes to change organizations through data science, how the different moving parts of data science have evolved over the past decade, and in which direction they're heading. We'll see the impact that data science can have not only in tech, but also in such various verticals as retail, agriculture, and the penal system. Taras will also talk us through the five steps required to change organizations through data science, all of which are necessary. Can you guess what they are? I'm really excited to have Taris on the show as DataCamp has had a long relationship with McKinsey, including that McKinsey uses DataCamp for training. We're also trying something this week for Data Framed listeners. The opportunity to try datacamp for yourself. All you need to do is email salesatacamp.com that's salesatacamp.com with the subject line podcast and redeem your free two week trial. I'm Hugo Bown Andersen, a Data Scientist at DataCamp and this is Data Framed. Welcome to Data Framed, a weekly datacamp podcast exploring what data science looks like on the ground for working data scientists and what problems it can solve. I'm your host, Hugo Bown Anderson. You can follow me on Twitter hugobown and datacamp@datacamp. You can find all our episodes and show notes@datacamp.com Community Podcast hi there Taras and welcome to Data Framed.

Hi Hugo, I'm really glad to be here.

It's great to have you on the show. I'm really excited to be talking today about management consulting, analytics for businesses and data science. So you're a Senior analytics manager at McKinsey and I want to know what that looks like on the ground. But before we get there, I'd like to know what your colleagues think that you do.

Yeah, you know, my colleagues, especially those that don't do analytics themselves, don't always have a very clear view. Most of them think that I spent most of my time building models. Some of my analytics colleagues think that I spend most of my time talking to clients. But the truth is actually a lot.

Broader than that and I'm excited to get to that. I'm wondering for the non technical people, your colleagues, what does building models mean to them? What do they think building models looks like?

Yeah, it's one of those sometimes scary, sometimes black box type of process where they see it as a whole bunch of random data from different client systems coming in and the outcome comes out that hopefully tells them some insights into what drives performance of the clients or what they should advise clients to do. But for most non technical people, most, but not all, this process is not very, very transparent.

Yeah, and I suppose they think there's some sort of computation, involves some sort of mathematics, but they don't have a strong sense of what that may mean.

Exactly, exactly. But you know, there are some people that do a lot of analytics work even if they are not data scientists themselves and for them they actually know exactly what's going on, they just don't write the code. But it's actually impressive how well they understand and that they underlying data science.

And actually, you know, I think McKinsey does an incredible job and a lot of work in thinking about how to explain technical model building, data science, analytic techniques, data to a broader audience. I saw recently there's a great interactive web based explanation you have called An Executive's Guide to AI which really explains the nuts and bolts of what type of models are considered artificial intelligence these days.

No, that's exactly right. You know, we actually work on that and one expert from our team really draw development of most of the content for the guide because we get this question asked all the time and we cannot use technical language to executives because they need to understand it. At the same time we really want to be factually accurate, things like that. Even explaining what machine learning, what deep learning is in simple terms is very, very important for us. And I think it's actually very important more generally so that people who can make decisions actually understand what, what they are trying to accomplish and they feel comfortable actually acting on it.

Yeah, I like that. And I like the fact that a subtle overtone of what you just said then is that the machines don't necessarily make decisions yet I Mean, we do have certain types of learning algorithms, reinforcement learning in particular, but the fact that when we're looking at a predictive analytics, we've got AI making predictions, but then it's down and up to human responsibility to then make decisions and take action.

You know, it's a great point. I think it's one of the key things to understand is how different is model prediction from the actual decision. You know, model prediction may say that, you know, a customer is likely to buy certain product if, if it's offered with a certain discount, or a customer is likely to leave the company because of, you know, interruptions in services. Right. But that has nothing to do with a business decision that a senior executive needs to make, which usually is, you know, what type of discounts to offer to any particular customer. Because there is a lot more that's going on and there is a lot more strategic as well as operational practices. So sometimes people think that predicting an outcome through the model is the same as acting on it. But there is like it's just the beginning. There is a really, really long chain of thinking, usually qualitative thinking, that needs to happen before you actually act on decision. Again, not always. In some cases, decisions need to be made in real time. Cases for example, like banking fraud or credit card fraud, and then algorithms actually act. But for management consulting, when you talk about strategy and really high impact decisions, models need to be interpreted. We like to use word translated and the implications need to be assessed and pressure tested and before anybody can make any decision because the stakes are just too high.

I love it and I look forward to returning to these ideas when we talk about specific verticals and industries where you see kind of the most demand for data science, analytics and management consulting. But let's get back to you before, before we get there. So you told us that your colleagues generally will think you build models, talk to clients, but you also told us that what you do is, is a lot more broad than that. So. So what do you actually do?

Yeah, so I spent time in probably four or five different areas as somebody who built and is leading a fairly large data science team. Probably the most important part of my work is to make sure that we have the right capabilities that we can serve our clients on what their needs are. And that involves quite a lot of strategic thinking in terms of what data science, what problems data scientists will be solving tomorrow or next year, and working backwards from there, what people you need to have in the team, what technologies you need to test out, what use cases and what clients you want to work with. So there is a little bit of that. A second piece is extremely important for anybody which essentially piece around relationship management. For us that means working with clients, managing and strengthening relationships so we can actually deploy data science to tools and drive real change. But also internally, right, As a leader of data science function, you need to work with technology, function with data, function with business development, function with marketing, function with individual businesses. Because data science is by nature so embedded into the rest of the organization that you cannot be effective unless you spend significant time, 30 to 50% of your time just managing the relationships, right? And then you know, it's still people, it's still people role. So you need to help people be successful. And that starts at just the first interview that, that, that you interview somebody and hopefully if they get hired and then it's onboarding, then it's finding the right projects, then it's doing mentorship on the project, handling conflicts, helping people, people grow, no matter kind of in what role they are and how experienced they are. Because that's, that's the only way, you know, consulting is people business. And the only way to create impact is by mentoring, developing people and creating opportunities for them. And then of course there is an actual data science part where I really, really enjoy doing, but for me that usually involves working with data scientists from the team, helping them structure the approach and then helping them kind of if things don't work, to understand why so that they're able to actually build models and deliver the outcomes. I still sometimes try to actually be a hands on data scientist and write code and build models myself, but it's really, really hard to find time, I'm sure. Last piece is of course, whenever I serve clients, it's making sure that they can digest the models that we are building for them. They know what the models mean, they can act on it, they can implement them, they can actually very importantly measure the outcomes post implementation. So they can see, yes, this really, really works. And this leads to actual tangible outcomes, benefits, either financial or other benefits to their organizations. Because that's probably like the heart of what we do as consultants in data science.

So really that last point speaks to something very well, essential to the job that you do, which is taking the results of the data analytic and data science process and as we said, turning them into decisions and actionables. But even before that, I suppose you need to translate the business problem into a data science question, you know, solve it to a certain extent and then translate it back to an actionable. So you have this bi Directional process in and out of data science, right?

No, that's absolutely right. There is always a process of, you know, even understanding the business problem, right. And then translating that into analytics approach and in the consulting settings that's actually not that straightforward because at the beginning when we just start the project, we don't always know everything that we would like to know about the client business and especially about kind of data environment and how they make decisions. So the process becomes very iterative, right. Based on what we know, we structure analytics process, figure out which models need to be built at what level of granularity with what data, then we go, then we try that and while we are trying that, we see what works with what doesn't, we show it to clients, we get their inputs and based on that we refine analytics approach until we actually get to something that's as meaningful and as impactful as possible. So that kind of translation happens at the beginning, but it continuous also throughout the project.

So does the question ever arise is data science or data analytics even the appropriate approach to the business question posed?

Absolutely. And I think it's the right question to ask, right? If you can solve the client question with the simpler approach, there is no need for data science or advanced analytics. So you know, I think advanced analytics is a great tool, but we shouldn't try to force it to be used everywhere. And if simpler approach works or if there is not enough data, then actually some high level strategic qualitative thinking or interview based approach or case study based approach works much better. On the other hand, there are clearly cases where you have to use fairly advanced modeling to get to understanding of the environment or market or business that's deep enough and that's accurate enough and that's the cases for data science. By the way, what's really interesting is that it happens both ways. Sometimes you start advanced analytics project that you think you would build a lot of models and as you learn client business or data availability more, you realize that that's just not going to work. So you go back to fairly basic qualitative analysis and sometimes you start a strategy project that has no analytics component. And in the middle you realize that there is so much to be learned from applying know advanced machine learning that you kind of can rescope the project and bring that in. And that's why you know, it's, it's so important to have people who actually involved in the project who actually understand data science as well as understand business question and then they can make these decisions on what's the best approach to, to go forward.

This actually speaks to something else that you know, different people on your team will have, have different skills. And you mentioned that you lead a fairly large team of different data scientists. And my question there is, have we as a community figured out what data science teams should look like? And what I mean by that is, you know, with backend engineers, we know what those teams generally look like and how they work together and best practices. Have we figured this out in data science yet?

You know, it's another excellent question. I don't think we have. And you know, I believe that it's not just McKinsey or consulting. I think it's in general, it's evolving very, very quickly. Our own journey was that at the beginning we had more people who were predictive modelers and who had very kind of broad backgrounds so they could build a wide range of predictive models. Increasingly, what I see is that on one hand you need to start specializing much deeper. So you need to start having people that only do NLP or only do deep learning, or only do anomaly detection. On the other hand, analytics expand beyond predictive modeling. For many questions that we need to answer now, somebody who understands complexity theory and can do real advanced simulations is actually invaluable. And I only see this process of changing the needs for data science skills accelerating with every year. For us, not only we haven't figured out the steady state, but I don't think there is a steady state. Maybe for some organizations that have much more fixed business model that need only certain type of skill set, maybe that is slightly more steady, but even then I do not believe that was given how quickly data science changes. You can be static, you can be static about it. You really, really need to constantly adding skill sets and moving along with the field.

And something you've spoken to there is that not only is the skill set changing so rapidly and the techniques, but even what happens on a daily basis, as we see more feature selection, automated, more data manipulation, data munching, data cleaning, automated machine learning, for example, we're going to see what data scientists do on the ground evolve incredibly quickly.

No, that's absolutely right. I think it's amazing to actually have tools now that we haven't had even two years ago, where even for computing infrastructure, you can go to Google Cloud or AWS and spin the system of clusters with all the software that you need in one click of a button, which was completely impossible before. And that changes the speed with which we work and just the nature of the models that we can build. And how easy it is so much that suddenly it opens up a lot more new possibilities. So there is certainly strong influence of technology on what skill set data scientists need to have as well.

We'll jump right back into our interview with Taras Gorishni after a short segment. Let's now jump into a segment called Freelance Data Science with Susan Sun, a freelancer who has worked with the New York Times, Cooney, Google, and General assembly, to name a bunch. Susan, what are we going to talk about today?

Hi Hugo, Today I want to talk about how to get clients as a freelance data scientist.

Great. Do you want to start by telling me your approach to getting clients?

Actually, every time I tell people that I freelance, this is the first question they ask. How do you get clients? It was my biggest fear when I was switching from full time to freelancing about getting clients. All the self help books, all the freelance workshops, they all say the same thing. When you're freelancing, you're supposed to be on the hustle. Everywhere you go, you're supposed to whip out your business card to give a very pithy 30 second elevator pitch to some corporate big shot in a suit. For people who are on the more introverted side, this is pretty much their worst nightmare. In reality though, I never once gave an elevator pitch. 9 out of 10 times I even had a really bad habit of leaving my business cards at home. I'm not saying that all the freelance advice from the books are wrong. I just think everyone has his or her own preference for how to be evaluated and it's tough to put your best foot forward in a networking session. I personally, I prefer to let the work speak for itself instead of trying to sell myself in a networking session.

And how do you get your work to speak for itself?

Well, to get ready for this podcast, I took a quick tally of all my clients and tried to trace back how I met them. 100% of my clients, past and present, I connected with them through a professional recommendation. So this is with someone who has worked with me on a past data project and understands my work ethic. I've been recommended by my ex coworkers, ex boss, even ex students, and definitely some folks from data volunteer programs. I think this is unsurprising actually. In a job hunt, the candidate who comes through a referral is usually much stronger than someone who applies through a job board or cold drops a resume or business card. And I think this is because the person who's doing the referring is usually staking his or her professional reputation on the line by making that referral. This means that the referrer has already assessed the candidate's professional and personality fit before making the introduction. This makes for a much higher chance of a good match, and it cuts down on the time and the resources spent on finding that right candidate. Now I know there are a lot of freelance job boards out there. I haven't spent much time engaging on those. For one, they're pretty notorious for the cutthroat pricing wars among the freelancers. And for another, the trust goes both ways. Just like how a client prefers somebody that came through a referral. I much prefer working with clients from recommendations because there's already a foundation of trust.

Right? That makes perfect sense. Do you have any recommendations on how to grow your network?

Well, if you live in New York City like I do, then you're definitely in luck. The city is great for the data community. There are a lot of conferences through the year. My personal favorites are PI Data and PI. Gotham Strada is definitely the most well known one because it's so huge and is really network friendly, but it tends to be on the expensive side. If you're on a budget, I would recommend trying some data related meetups hackathons. You can browse through what's available through meetup.com or eventbrite. There are also a lot of volunteer programs for pro bono data science work. I got started with Datakind and recently joined Data for Democracy, but there are many, many more inside New York City and all over the country. Even outside of New York, there are a lot of choices. If you don't see the meetup that you want in the city that you live in, you could definitely form one yourself. Data science boot camps are very popular these days. You can participate in the boot camp as a student, a teacher, or simply stop by and listen to one that free talks or information sessions. General assembly for one has a great alumni network. I actually found one of my clients through the GA network. And finally, there's no need to limit yourself to only the data community, the freelance community, not just the data. It's very extremely supportive of each other. And if you're just starting with freelance, it's helpful to attend some local freelance events that are cross disciplinary. It helps to meet people who have the same questions that you do.

Thanks Susan for that informative introduction on how to get clients as a freelance data scientist. After that interlude, it's time to jump back into our chat with taris. I want to jump in now and find out through the lens of your work at McKinsey which verticals and or industries do you see the most demand for? Data science, data analytics, management consulting at the moment?

Yeah, if I look at the moment right, I tend to think about three broad buckets of industries, right. The first one is the one that really interacts with consumers directly. So industries like retail, industries like telecom, some media, some banking. And for them the need to set up, it's actually really driven by personalization. They need to heavily customize their products, their marketing message to individual consumers ideally, and do it in real time with as much data and analytics as possible is really key. The other bucket is essentially organizations for which risk management is really a big deal. So insurance for example, insurance is all about pricing risk properly and managing claim processing very efficiently, setting up the rates very efficiently. And for them you can't do risk management without quantitative methods. So that's another really interesting application of analytics. And then finally many other organizations, they generate a lot of data from their operations. For example, think about a semiconductor fab that has literally hundreds of different pieces of equipment and each piece of equipment for each process step, making chips generates real time data with millisecond precision from all the sensors. It's extremely complex, the data is very large and all real time. And stakes are really high to optimize manufacturing and do it well. So I think for those organizations there is a huge need for actually really, really advanced analytics to be driving better decisions. Another piece similar is genomics and bioinformatics. Huge amount of data coming in. Another piece is healthcare where you have now structured data on healthcare claims, but also unstructured data from medical testing, images, et cetera. This, this is a little bit more where data is there but analytic methods are not always there and data has not been used yet fully. And that's, that's another area where there is increasing, increasing pool for data science.

There is so much interesting stuff in there that I'm kind of at a loss to figure out quite which direction to go. So you have you stated industries, verticals that deal with lots of consumer interactions, those that need really good, excellent risk management. You also mentioned the industries where data's been generated. Semiconductors is really very interesting because I think a lot of people, when they think about data science, they think about data science in tech where we kind of know what we're doing a lot of the time, but when we've got real time data flowing straight in and we need to make decisions straight away or automate those decisions, that's a very different game, isn't it?

Absolutely. And you Know, we always talk about kind of digital born companies and they do amazing stuff, don't get me wrong. Right. But if you're a completely digital company, it's actually easy, right? Because all your information is digital. You can access it and it's, you know, you can test your ideas really quickly and it's just natural. Now try to actually do analytics if you're a semiconductor player or if you're agriculture provider. Right. Some of my most fascinating experiences were actually in kind of old school feels like agriculture. But imagine that you have a tractor driving in the farm. Tractor has a bunch of sensors. Now tractor has a video cameras that can film the fields. You have a drone that flies on top and takes real time videos of the crop. And then all of this needs to be transferred in a very, very low cost from the farm to central processing unit or all the computations need to be done on premise on tractor as an edge system. And you need to optimize extremely complex set of real physical equipment actions. It's actually really, really hard. And you need to do it in such a way that the farmer would actually not mind using the advanced analytics to improve his yield of crops. So I think very often we don't give enough credit for traditional industries that operate in the physical world because it's so much harder. And to me, that's where a lot of value of analytics in the future will be coming from.

That agriculture example is so fantastic because it speaks to another very interesting concern. I think that, okay, let's say you have a drone taking footage and you want to do some image analysis, pattern recognition, object detection, bounding boxes, whatever it may be, you might say, oh, I can just throw a huge convolutional neural net at it. Having said that, if you want to do this in real time in the drone, sure, if you've got, you know, clusters in the cloud, you can do this. But if you want to do this in real time in the drone, it will actually change what type of model you build to do it.

Right, Exactly. Because literally in that time you start thinking about time it takes to classify a single image and you actually literally start counting how many milliseconds your convolutional network can run. And difference between, you know, 100 milliseconds and 300 milliseconds means certain speed of flying for the drone. And you just cannot go sometimes with the 300 because otherwise to take image of, you know, 100 acre field will take you forever. And so you go with simple, simpler models that can work in the edge computing environment that may not certainly be as accurate as possible, or you spend a lot of time thinking through the, your deep learning architecture on how to keep accuracy up, but you know, reduce computing time down dramatically. And that's another fascinating problem that's very, very different from what conventional data scientists would have to deal with.

Right. So now I want to jump in and find out about organizations that you've worked with or industries. I do understand that there are certain, there's a lot of privacy you need to respect with respect to your client base. But I'm just wondering if you tell me about a few key examples of organizations and or industries that you've worked with.

Yeah, absolutely. So let me switch gears and talk again about another really, really low tech client. We work with correctional facilities and the problem that they had was problem of violence. The inmates were sometimes really violent and they were seeing increases in violence and the administration could not understand why. In some ways, think about it, right? It's a very challenging environment because it's always high pressure and the stakes are really, really high. Literally people might get injured or killed. On the other hand, you have ability to act and you have a lot of data that you can use to make better decisions. What we've done is we looked at what actually drives violence in each group and it's not necessarily about individuals. It's about if you put certain number of individuals in the same cell, it's the relative composition of people that go in there that increases or decreases the violence. Then if you start thinking about it from the quantitative perspective and understand what drives that, you can actually reduce violence quite dramatically. And again, it's a great example where looking at industry that traditionally does not use analytics leads to the huge impact and it really makes a difference in a major way.

So that sounds like a success story. And I do think when an organization has challenges or problems that data analytics can help to solve, there is some sort of barrier. The value needs to be created and it needs to be demonstrated. So I'm wondering what does it take to change an organization in terms of their decision making process to change them through data science?

Yeah, it's a great question and it's a, you know, it's a very, very challenging task because many things need to happen and all of them need to happen. If one link is missing, chances are it's not going to work. So what I mean, specific, first of all, you need to have a vision for what analytics should do for organization. Why do you actually want to have data science or analytics in the organization and people need to truly believe that it will add value and it is connected to the business strategy, not just in analytics team, but throughout the organization. Secondly, you need to have a support from very senior executives, ideally C level, to actually create that excitement in organization, to make sure we have the right visibility, right funding and right resources to do it. And to actually help analytics or data science team work with other parts of the business. Because without that, very often data scientists just build models in isolations and models never get used. Then you need to have the right data and right data environment and that is relatively slow process that could be expensive. So you always need to actually start building that environment, but you shouldn't wait until that is done before you start deploying analytics. So, so you kind of want to identify areas where you can drive real measurable results very quickly so that there is excitement in the organization about the analytics and more and more people want to try it. That's what we mean by use cases. You find out where there is the biggest value for analytics and you go and you build the models. You start making decision different. You capture that value for like 3, 4, 5 use cases quickly. And then everybody wants to do it for their own part of business. And then you need to make sure that it actually sticks. So it's not something that you've done. Once people use it for three months and then they went back to the old ways to make something stick. You need to redesign the processes of making decisions that usually involve some kind of technology solution, software interfaces to make analytics digestible. And it involves retraining people and it involves new ways to measure the outcomes of those decisions. And then finally people who really or organization really great analytics, they always do that. You need to kind of change the culture of organization so that every time you run a business meeting and you want to propose something, people ask what data do we have to act on this proposal? Is it really backed up by hard numbers, well designed models or not? And once you get to that level, then analytics truly becomes part of the company DNA and you know, you accomplish what you're trying to do. That is multi year journey for most organizations. And it's not easy.

Right. I want to zoom in on this idea of early stage value extraction in particular through several use cases. I suppose this essentially is having a few proof of concepts, demonstrating their value and gaining trust of people at different levels throughout the organization.

Yes. No, that's fair. Think about it from the perspective of kind of business executive who doesn't necessarily understand technology. Right. It's a buzz. Everybody talks about it, but. But you want to know if it's, if this thing is real or not. And what real means is, you know, if I start doing what the models tell me to do, is my business doing better or worse? You need to convince somebody to give you a shot, which usually happens through the proof of concept. And then you need to very rigorously measure what happened so that it's so clear that there was impact and this impact is directly attributable to analytics. That use case then does two things for you. A, the business partner that you work with becomes a champion for you and uses more analytics within their own part of business. But B, it creates much broader visibility that other people see it and other people say, you know what, this thing actually really, really works and I'm just going to go and try to do it. And then finally what's important is that if you can show that you spent $2 million on a use case, but it generated $50 million in revenue or in cost savings, suddenly you can claim that analytics become self funded. And so organization can allocate a bigger budget now to do another three or four use cases and to fund new technology or new software or new data management systems. And that's how you get going in real organizations because you always need to justify the budget that you spend on certain activities.

Right? So you mentioned if one of these moving parts is missing, the change in the organization will very likely fail. And you know, there are a lot of these moving parts. You mentioned creating the vision for analytics, having strong support particularly at C level, early value extraction, process redesign, culture change, data foundation. It seems like a lot to, to get working together. How often are you able to do that or how often do you see it all work together?

You know, it's a great question, right? Another question to ask. Another way to ask this is how many organizations have actually been able to achieve it? And there is not many in each industry. There is maybe two, three, four organization that are ahead of the pack and sometimes none of them actually achieved this wide acceptance of analytics in every decision making that we talked about. So it's actually tough. But what we are seeing is that where clients are now is very different from where they were two years ago. Everybody's moving in the right direction, right? And everybody, or not everybody, but most companies are making significant progress. They're just not all of them are there yet. And I think this journey will take a little bit longer for them. There's just too many things need to happen. And each of these Steps is actually quite a long step.

We'll jump right back into our interview with Taris after a short segment. Now it's time for a segment called Insights From Computational Education. Today we welcome Neil Brown, who works as a research fellow in Computing Education at King's College London. Let's start at the top. What do you think is the most important finding in education research, Neil?

So, for me, I think the most important result out of education research is that skills and knowledge are domain specific.

Let's unpack that a bit. What does it mean for skills to be domain specific?

So a lot of people have this idea that there's a general skills, things like problem solving, and that if you sort of do things like Sudokus or you learn to play chess, or even if you learn to program, that you're training up your problem solving skill and that this will make you better at sort of solving all problems. But what education research finds is that this is not the case. So if you learn to play chess, you tend to get better at playing chess, but it's completely independent from improving your ability to program.

Well, that's interesting because you hear a lot about these brain training apps and things like that to keep your mind active, but what you seem to be saying is that they don't really help to train your brain, so to speak.

Yeah, that's right. So if you solve a particular puzzle, like Satan, Glex, Sudoku, then you're improving your ability at that specific puzzle, but you're not improving your ability at at totally unrelated puzzles.

So what other implications do you think this result has?

So, one area where I think it's quite relevant to technical people is this spate of interview puzzles that appeared a few years ago.

All this how to weigh a 747 plane and things like that.

Yeah, that's the kind of thing. So once you realise that problem solving isn't a general skill, you realise that asking questions like this doesn't make any sense. So if you want to ask somebody how many tennis balls can you fit in a car, and they try and sort of work this out, you're testing how well they are at things like numerical estimation or spatial skills. Maybe the estimation bit is useful if you want to do project management, but if you're looking to find out about how well they program, it's not going to tell you anything. So really, this sort of result from education research just makes things very simple. If you want to improve people's programming, you should train them in programming. If you want to test how good they are at programming, then give them a programming test. There's no need to approach it in a roundabout way with all these other sort of complex problems.

Great. So you're saying that this has implications for interviews and apps? Are there other areas this comes up?

Yeah. So I find that once you know this, you start to see it sort of everywhere, this kind of reasoning. So, for example, a common belief in the UK used to be that teaching Latin would encourage you to think more logically, to improve your logical thinking. And Latin grammar is probably close enough to English, French, German grammar that the knowledge may transfer and improve your grammar. But learning Latin isn't going to help you do better in maths or in programming or anything like that. Conversely, learning programming isn't going to help you do better in the other subjects. They're just separate things.

Interesting. Thanks, Neil, for elucidating the domain specificness of both skills and knowledge. Do you have some links about this that I can include in the show Notes?

Yep, sure, yeah, I'll be happy to share some links with you.

Sounds good. Look forward to speaking next time. Time to get straight back into our chat with Taras Gorishni. So, Taras, you've been levering data science to help Fortune 500 companies, for example, improve their performance for over a decade now. How have the different moving parts of data science evolved over this time?

Yeah, you know, it's fascinating. I would say that there is so many changes that happen in so many ways. The simplest one to see is a software, right? 10 years ago we were using mostly SaaS and then I know open source really became prominent and people moved to R, then to Python and there is just continuous emergence of new software tools and just very, very easy to see and track. Second piece I think is for algorithms themselves. Ten years ago we were still doing a lot of BI and we were doing a little bit of statistical modeling, but it was mostly linear models, it was classical statistics. Then machine learning became much more prominent. And then very quickly we started being able to work with unstructured data through deep learning, through nlp. And to me that's the next wave that's happening now is moving to the unstructured information. And then another change that I noticed is that in what domains analytics is used? Ten years ago, market analytics was big, risk analytics was big in financial sector because financial sector really needed that. And then some of the heavily operational companies would use supply chain inventory management, but those were three big areas. Now if you look at where analytics is used, it's literally everywhere. It's In HR organization function, it's understanding what people to hire and when they're likely to leave. It's every interaction you have with customer, it's customization, it's just so broad and so that's relatively new. Right. This thread of analytics into every decision making is something that happened in the last 10 years. And to me that's probably the most fascinating part.

I think the HR example is really interesting and we've seen increases in machines helping out with the hiring process, for example. Of course machines can encode human biases. Machines can create their own types of biases as well. And I'm wondering what your view on these types of biases that may occur, what the major challenges are and whether how important it is to have a human in the loop as much as possible.

Yeah, this is another great question. Right. I will tell you my own perspective and I do think that there is many different points of view on this. I think that human plus machine works a lot better than just machine alone. Some of my colleagues that are data scientists in tech firms actually disagree and they see their role as fully getting human out of the loop and fully automated everything. I think it's just difference in the nature of decisions that we are trying to solve with analytics. But for me, on one hand it's really, really, really good to design algorithms with as little bias as possible. And bias usually just comes from your input data. So you need to think about is your data actually representative or is it biased in any particular way? But on the other hand, once you get your model predictions, what do you do with it? We still come back to the issue of translation and to act properly on output algorithms. That's where human really, really, really, really adds value. So if I go to HR example, for example, I spend a lot of time looking at McKinsey Aon recruiting and understanding what performance characteristics that people display during interviews are predictors of their long term success. But it's not like I take output of predictive model and allow algorithm to decide who's going to get hired or not. It's just for us to inform our recruiting process, to run it better and to focus on the relevant things. But humans still needs to process, digest, assess how precise the predictions are, assess false positives and true positives and based on that, design the right recruiting process. You can't just take human out and let algorithm decide. It would not work very well for much more qualitative strategic decisions that you need to make.

Yeah, and you actually said something very interesting. Well, a lot of very interesting things in there. But something I want to zoom in on is you said, you know, a model may, for example, optimize the long term success. And we may not even know what long term, how to define long term success correctly. That may be something that evolves over time. Trying to figure out what we're actually optimizing for or what we're learning is just as important as implementing the algorithm itself.

Absolutely, absolutely. The performance metrics matter tremendously. And sometimes we see if you change one metric to the other, suddenly the models change a lot. A lot, a lot. And that's also why it's so important to kind of keep an open mind and have human involved because you actually sometimes we build models for different outcomes. Outcome might be your, if I take slightly more traditional commercial example, outcome might be your revenue growth, might be your profitability, or it might be your market share. And then you realize if you build three different models for three different outcomes, you get three different sets of drivers, and if you start analyzing the differences between them, you really get to much deeper understanding of your market dynamics because you look at multiple dimensions of the problem. And that's absolutely critical.

So you gave some really nice insight into how you've seen the different moving parts of data science evolve over the past decade and more. How do you see the different parts of data science evolving in the future?

Yeah, it's fascinating, right? For one thing, I think open source is here to stay. I just can't imagine doing analytics without open source, given how complex it gets and how increasingly powerful it becomes. We all have to share the code, share algorithms, and it will remain relevant for a very long time. For second, I think unstructured data will become more and more important. It's already most information that's generated has been unstructured, but we are just at the beginning of being able to use, been able to use it. I think the advances that can come with improving, for example, NLP techniques will be truly transformative and it's actually beginning very, very real. Even the work we do now in NLP is so different from what we could do two years ago that if I just fast forward that pace of development, within the next five years, we'll have a very, very different face of analytics just because of that. To me, it's also really a lot about interpretability. Take deep learning, for example. Amazing techniques allows you to do things that are not possible completely. Analyzing videos, analyzing images were not just possible to do well before deep learning. At the same time, the applicability of deep learning in truly strategic Decision making is not that high primarily because of that. But I spend a lot of time looking at symbolic feature engineering. So something that can automatically discover features just as deep learning does, but represents those features as symbolic equations. And that becomes extremely powerful because then you understand what's going on and what drives your outcomes and you can actually explain it to the executive. So to me, I think there'll be a lot of advances made in a representation of models and in making models more and more and more transparent while using complex models, because simple models we have done it is really now about complex models, but making them understandable. That to me will be another big change that hopefully will happen in the next five years.

Great. So having seen how the different parts of data science will evolve in the future, in your mind, how do you see the use of data science within organizations evolving in the coming years?

It's another fascinating question, right? I think the data science will become democratized and part of it is that, you know, isolated data science teams that we have now in many organizations will become much more embedded with the rest of the organizations. So literally data science will have connections to, you know, every business, every function, and as a result it will be impacting a lot more decisions. Another piece is that I think data science will be used by some of the non data scientists. Just like when Excel and spreadsheets came around, suddenly business analysts were able to use it and you didn't need to be a programmer to do it. With good graphical interface, with algorithms that are much more intelligent, you can actually have people who are business managers or business analysts to begin to use data science tools. For me, Alteryx is a great example of somebody who's in Tableau, two great examples of somebody who's done that. For data management, you suddenly don't need to write SQL code and it's very easy. And I think we'll see a lot more of that in just core areas of data science as well. So that there'll be hard problems and real data scientists will be solving them and there'll be more standard problems for which we'll have really, really good tools and lots of other people will be able to use them.

So this is somewhat of a controversial question in a lot of circles, but my question for you is can now or is there a future in which data science can be done in a gui?

Yeah, I mean, look, we're all data scientists, right? And the reason why we don't like GUI is because it's just inefficient way to work. But at the same Time. If all you do is multivariate aggression and all you need is look at your p values and look at your regression coefficients, why would you as a data scientist want to do it yourself? Writing code in Python? Just give a GUI to business analysts, let them upload it, build in right check and balances to make sure that the right significance is done, and have business analysts take introductory course into statistics and you're done. I think data Camp, for example, has done a lot of democratizing that knowledge because online learning data science is so powerful. And I think it's not just data scientists who take the courses. Right. So at some level, you know, we as a data scientist don't always want to kind of let it go. Right. And sometimes we just don't trust other people to do it. Right. But for us to really continue to have impact, we need to work on really hard problems with continuously new methods and approaches. And that means stop working on simpler problems that are very easily solvable that somebody else can do. To me, I just see that as a natural progression.

I love it. So I'd like to know what one of your favorite data sciencey techniques or methodologies is.

Yeah, so you know, I'm huge fan of autonomous feature or hypothesis generation. Again to me, kind of ability to find features that drive outcomes, but yet unlike deep learning, ability to express those features through symbolic equations is hugely powerful, at least in a strategic decision making where we need to have more understanding than precision. I've seen this space kind of evolving very, very rapidly in the last two, three years, from pretty much zero from before that to actually now having companies and products that can do it quite well. And I think we're just scratching the surface of that. So for me, the interface of powerful future engineering and interpretability is definitely kind of my favorite area in data science. Now another one is which my second favorite, it's actually the science of complexity. We talk about building models, but if you start thinking about simulating behavior of complex and nonlinear dynamic systems that sometimes behave in chaotic ways, but they're still guided by rules that can be learned and can be developed, it's fascinating and its impact is huge because our life is essentially a complex system. Every organization is complex system, human body is complex systems, most engineering processes are complex systems. So ability to actually do it well with precise mathematics is absolutely fascinating. And again, I'm beginning to see some data scientists and some software tools being developed that can handle complexity in, you know, without excessive amount of effort, I would say. And to me that's, that's Another emerging trend that I think we'll be talking a lot more five years from now.

I really like that the favorite techniques and methodologies you mentioned are both very forward looking because you know, the automatic feature hypothesis generation is something which is now and in the future will change what all data scientists do on, on a daily basis. It will automate a lot of the, the drudgery, right. The, the 80% so that we can focus on the far more exciting and, and creative work.

Absolutely, absolutely. I have no doubt of that because it's already happening and we will just continue to do this in more classes of problems and with more impact.

So Taris, to wrap up, do you have a final call to action for our listeners out there?

My final call is really things are moving so fast now and all of us as data scientists, if we are doing now what we were doing a year ago, we are really, really falling behind my call for action and that's something that I try to do every day. Continue to learn new things every day, keep in touch with new software development, keep in touch with new algorithms, understand mathematics and just deploy it. Because that's, to me that's the biggest fun part about BIM data scientists is this learning and its application to new domains, new areas all the time. And it's very, very nice that we actually have to do that ourselves to stay relevant. So don't do the same thing over and over. Keep looking out for new opportunities.

I couldn't agree more. And I think the need to learning new things, especially as you said in a field that's moving so quickly, you also mentioned to, you know, learn some of the math as well, which is incredibly important and also to not be so scared of math because math can be overwhelming in a lot of ways. So when you're, when you're writing code, learn bit by bit the basics of what you're, what you're doing, maybe try to re, implement a few of your favorite algorithms, whatever they may be in terms of learning new things. I may be biased, but I think Data Camp is an incredible platform to do that on. I also think McKinsey has a lot of, of fantastic online resources. As I said, this interactive executive's guide to AI. You've actually got a whole bunch of stuff, commentary, stuff released recently on the economics of AI, which I think would provide a wonderful counterpoint to people actually writing code as well. So I think all of that is really fantastic and we'll include a link to a bunch of those resources in the show. Notes.

Perfect. Could not agree more.

Absolutely. So Taris, thank you so much for joining me on the show. It's been an absolute pleasure.

Yeah, pleasure was mine. Thank you so much for inviting.

Thanks for joining. Our conversation with Taras Garishni we saw that the industries in which there is currently the highest demand for data science management consulting include those that deal with lots of consumer interactions, for example retail, telecom, banking, then those that need excellent risk management, for example insurance, and then those that deal with lots of data such as healthcare. We also saw that to change an organization through data science there are many steps and all of them are necessary. If one is missing, the change will likely fail. The steps are create a vision for analytics, have strong C level support, extract value early on from at least several use cases, do process redesign and culture change and build data foundation. And don't forget that this week we've got a special treat for dataframe listeners. The opportunity to try datacamp yourself. All you need to do is email salesatacamp.com that's salesatacamp.com with the subject line podcast and redeem your free two week trial. Also make sure to check out our next episode, a conversation with Alan Nicoll about chatbots, conversational software and data science. Alan is co founder and CTO of Ruza who build open source machine learning tools for developers and product teams to expand bots. Beyond answering simple questions, what does the future of chatbots hold for data science and society at large? Find out next week. I'm your host Hugo Baun Anderson. You can follow me on twitterugobown and datacampattdacamp. You can find all our episodes and show notes@datacamp.com Community podcast. Sam.