A passion for maths and solving mathematical problems led Niclas Thomas to a PhD in machine learning with a focus on medical research. But then a conversation with a recruiter steered his career towards data science in the retail sphere. After stints at Tesco, Sainsbury’s, and Gousto, Thomas is now head of data science for Next, the clothing retailer.
In this interview with Real World Data Science, Thomas reflects on his career journey so far, from hands-on coding work to team leadership and management. He also argues for the importance of communication and storytelling as part of the data science skill set.
Transcript
This transcript has been produced using speech-to-text transcription software. It has been only lightly edited to correct mistranscriptions and remove some repetitions.
Brian Tarran
Niclas Thomas, thank you for joining us today. I hope you’re well.
Niclas Thomas
I am indeed thanks. Thank you for having me.
Brian Tarran
Today we’re meeting because we want to find out a little bit about your career in data science, how you got into it, what you’re doing now, where you see both your career and data science as a profession going next. So do you mind– can we start by giving us a brief introduction to who Niclas Thomas is?
Niclas Thomas
Yeah, of course. Yeah. So I’m currently working as head of data science at Next. My background is academia originally, a maths degree. I did my PhD in machine learning, and more in medical research, so more of an applied machine learning position where the idea was to try and predict, ultimately, predict disease from a given sample of data from blood – can you actually predict future disease? – which I think is a really interesting area; I love medical research. And then [I] switched over to more commercial role and worked in several retail data science roles: so, Tesco, Sainsbury’s, Gousto, and then now, as I said, currently head of data science at Next, where I run a team, and I imagine most listeners will be familiar with what Next do: a retail, a clothing brand on the whole, where the idea is, obviously to sell some great stuff, great products and put the right product in front of the right customer.
Brian Tarran
Can you tell us, what does your job involve? What are your sort of main tasks and responsibilities in that role?
Niclas Thomas
Yeah, so I suppose I’m lucky enough to have been head of data science at several different companies: Sainsbury’s, Gousto, and Next. So it’s always interesting to compare the role of the head of data science in each of those three. At the moment, I think there’s a core focus on, well, ultimately making sure the teams are efficient as possible. And that really means just making sure our tech stack – what tools, what programming languages, what software we use on a day to day basis – is set up for success and make sure the team have what they need to be able to do the job as efficiently as possible, whether that’s using Python or R, whether that’s how we develop code, and how we work with other people as well, being a big part of that, then. So how do we work with other software engineers? How do we work with web developers, then, to make sure that the work we do actually gets in the hands of the business and ultimately in the hands of the customer. So that’s one aspect: it’s just making sure the team is set up for success, both in terms of the ways they work and what tools they have to work as well, then. I guess the other side of that coin is what we actually work on. So understanding the value of potential work we could do, and helping the team understand what that value is, and, and ultimately giving direction of what things we want to work on next. Obviously, that’s not my decision in isolation, but understanding on the one hand, what other stakeholders want to do, what my superiors wants to do, as well. And trying to put that all into the mix to understand these are the next best projects to work on given a finite amount of people to work on these problems. And then ultimately, then, the last part, then, is ultimately helping the team deliver those projects, those products as well then, which usually means calling on my experience of having solved these problems myself, either directly when I was earlier in my career or indirectly through leading others then or, you know, being the head of a team and working with some other great people and to learn from their experiences as well.
Brian Tarran
What does data science mean to you, personally? I’m not asking you to define it for everybody. But for you, what is what is data science?
Niclas Thomas
Yeah, I wish I’d come up over the years with a great definition of this. But yeah, I mean, really, it just, I mean, at the very highest level, it just means using data to drive business value, I suppose, as I guess in my– which probably reflects the fact that it’s more of a business role that I have. But I think that in its broadest sense, I think that’s true: using data to drive insights and make decisions for the business. There are more, I guess, detailed definitions of that. So, for example, the way I’ve always differentiated between data analytics and data science is that if you want to make repeated decisions on a daily or weekly basis, then that’s when it becomes more about a data science question versus a data analytics question, because data analytics is generally about answering large one off ad hoc questions, rather than making the same decision over and over again and using methods appropriate for that. But, ultimately, that’s what data science means to me, I think: making repeated decisions using data and the scientific method to use data for good.
Brian Tarran
And so what do you think is your most important skill as a data scientist given that definition that you have of data science?
Niclas Thomas
In my role, I suppose communication ultimately becomes the most important thing. I’d say definitely earlier in my career, and I think if you’re the person actually delivering and implementing the algorithm, I think that the technical skill set obviously is really important then. But ultimately, I almost see my role as the head of data science as a hybrid– as a link between my team and the rest of the business, then. So it’s really about being able to, on the one hand, translate technical concepts into non technical descriptions of what we’re actually doing, making sure the rest of the business can understand and vice versa, then making sure I understand the business process and business terminology well enough to be able to translate that for the team, as and when needed, into a vision for a project, a product, then, and develop a strategy for that. So I think that the communication both in the strictest sense of being able to talk that through with, with my team, with other team members, with stakeholders, as well, but also more in the looser sense, then, of being able to define that strategy, being able to define what the roadmap for a particular project or a product might look like.
Brian Tarran
Can you talk us through your so your education and your training that led up to your kind of first data science job, your first data science role.
Niclas Thomas
I suppose the first time, the first time I– actually, I’d never heard about it, I think, when a recruiter approached me. This is probably going back into 2014, when I was maybe eight months into my postdoc after my PhD. I think– obviously it did exist before that, although I suppose the terminology wasn’t quite as widespread going back almost 10 years now where the term is a lot more rife. So my original background, I did a master’s in maths originally, four years. And then I remember being– the last year of that, then, I was applying for a few jobs, and I applied for one at the Met Office, where the focus obviously was predicting weather, forecasting. And I wasn’t successful in that job. But I did notice that the, on the job spec at the time, it was PhD preferred was one of the specs on that role. It was probably the first time I thought about taking on a PhD as more of a career move rather than as the natural progression to an academic career, more of a business career move if you like, then of actually how it can help you in more business settings. So that was at least when I decided to do my PhD and thought it’s certainly not going to be– and this was back in 2008, so at the time of the financial issues at the time when getting jobs was harder anyway, so it felt like a win-win of doing something that would be– I was clear I wanted to work in a data role of some sort. And that combined with the fact that I thought it would be a good career move and the financial climate at the time wasn’t brilliant. So I took on a PhD then. And then in terms of actually getting into, into my first data science position was, as I said, just after I finished my PhD, I had been working about six months, eight months as a postdoc, and then a recruiter just described a role that was available at Tesco at the time. And it sounded a lot of what I was doing in my current postdoc role at the time – making predictions based on data and exactly the same techniques – sounded really interesting. And it must have been the way the recruiter sold it at the time as well then, because it’s something I was really keen to take on and then made my move off the back of that then. So yeah, kind of moved into it a little bit, I guess, semi deliberately from taking a PhD on first, but always with the view of moving over to a business role at some point after that.
Brian Tarran
But it wasn’t like you started out your further education thinking, “I want to be a data scientist, what do I need to do to kind of get there? What are the subjects I need to focus on? What are the topics I need to research?
Niclas Thomas
Yeah. Oh, absolutely. Yeah, it certainly wasn’t by design at the very start of my journey. I fell in love with math, really, and just fell into data science because of that, really, I loved numbers and loved solving maths problems. So that’s why I did a degree in it first of all, then and certainly, you know, even midway through my degree, then I wasn’t really sure what I wanted to do. It was more, as you say, just by chance, then, that there were a few opportune moments that came around then, that opportunities came around at the right time to fall into that career.
Brian Tarran
Doing a PhD in machine learning as you did, that was quite a – in hindsight – a smart choice of PhD to pursue, I think, right?
Niclas Thomas
I think so. Yeah, I suppose it was– still even at that stage it wasn’t necessarily, again, the terminology ‘data science’ wasn’t really around. Certainly, when I started my PhD in 2009 2010. It wasn’t really terminology, at least it may have been in usage a little bit in terms of being on, you know, if you look for jobs on LinkedIn or Indeed, but it certainly wasn’t terminology that that I would have been particularly familiar with.
Brian Tarran
Your first job in data science was at Tesco. You mentioned that you were you were kind of recruited to that role there. How does it compare to your current role? So I guess, you know, what’s the difference between being a data scientist versus head of data science as you are now?
Niclas Thomas
Yeah, I think there are probably more similarities than differences, I would say. We were quite lucky in the setup in Tesco that the recruitment strategy seemed to be more focused around people who already had some experience in, generally, either already had business experience or a PhD. So we were fairly independent in solving our own– the project that we were working on and working on that. Not necessarily with the head of data science guiding us, you know, day by day, in terms of the actual nitty gritty and the technical detail, which is great, then. So it did mean that we had responsibility and ownership for our product quite early on. So yeah, I really enjoyed that. I suppose I was writing a lot more code in those days than I do now. I rarely, if ever write code at the moment. So I think that’s probably true for the last maybe three or four years, I think, only occasionally getting my hands dirty. And even when it is, it’s not really to build an algorithm, it’s more to inquire about what data we have to solve the algorithm then. So even when I do get my hands dirty, it’s more in the very early stages of the whole algorithm development lifecycle. So I think that’s probably the biggest difference is just the actual ownership of development there – probably expected, I would say, but it’s– I think that’s one of the beauties of being in your first job or two in data science. I think the– I think in most places I’ve seen, I think you’ll get ownership of, of the work, the stuff that you work on, on a day to day basis, quite early on. And you’ll be expected to contribute code and ideas for that as well, which I think most people would love. I certainly loved it at the time.
Brian Tarran
What was the most important thing you learned in your first year in that job?
Niclas Thomas
I think, again, it’s probably a lot around the ways of working, I would say – of the various ways you can [work], which I never really thought about it before. Working in academia, it was quite isolated, I suppose. You work on your own project, you work on your own work and don’t really– or at least, I found I didn’t really work with anyone else that much. Maybe that was the nature of my work as well, we’d obviously be dependent on people working in a lab to get data. But I think the day to day work, I was working quite in isolation, whereas the team aspect of working, I think, was a steep learning curve then – so agile methodology, and everything around that, which was very, very new to me. And the various ways you can do that. I’m generally not someone for overly putting processes in place in a team, only where necessary. But I think there’s some great learnings from that as well. It certainly started to shape how I think I would want to run a team if and when I got to that position.
Brian Tarran
So, Nick, what have been your career highlights so far?
Niclas Thomas
I think in terms of– there was one product we built in Sainsbury’s in particular. So in terms of, on a product level of replenishment. So how do you most efficiently get products from the back of the store onto the shelves of an individual store? And what’s the most optimal strategy to do that, which I love for a variety of reasons. A, it was one of the first full data science products that we had deployed and worked on as a team in Sainsbury’s. So there was that kind of milestone about it. I think it also stood out as a really nice move away from classic machine learning – i.e., making a prediction, a classification model – to something that was a bit more operations research based and more based on optimization. So using graph theory, making a graph network of a store. And using that to solve the problem of taking a route through the store, for example, a bit like a Google Maps for a store basically, was how we always pitched it to our stakeholders, and how can you choose the best route and again, moving more into a bit more of a vehicle routing problem, then: if you’ve got two different trolleys, how do you decide what items to put on trolley one versus trolley two? So there’s loads of interesting stuff on the technical side of things and it was, again, I felt it was probably one of the highlights – as well as the end product, it was also the one I worked on at the very start. So actually, the understanding whether it would be possible to do that, what kind of technical approach. So I think certainly from a product perspective, that’s probably stuck in my mind. Aside from that, on a more personal level, I guess, I did decide to write a book off the back of my PhD. Just mainly on my experiences from my PhD and postdoc. I mean, it’s not like a confessional. But more on the– just working with non data scientists and making it more accessible was really what I really focused on there. So having worked with clinicians, immunologists and others as part of the medical research that I did, I felt that data can be accessible if you pitch it in a way and make it easy to use. And so that was the purpose of what was largely an educational textbook.
Brian Tarran
Do you want to give a short plug for the book, what it’s called and where people can find it?
Niclas Thomas
Yeah, so it’s, Data Science for Immunologists is the name of the book. It’s available on Amazon. I’m one of the two co-authors on that then. And we do have a website, datascienceforimmunologists.com, as well then if you did want to visit and you can either buy the book, there’s a link on that website or just go straight to Amazon and it’s available there.
Brian Tarran
This next question, we’ve gone from highlights to lowlights. Have there been any mistakes or regrets that you’ve had along the way in your data science career so far?
Niclas Thomas
The main mistakes I think I’ve made before is not valuing, A, communication or soft skills, but B, the leadership and management as well then. And I think especially it’s something, when working at Gousto as well that was something that was a big focus of the team and something that I really took from my time there as well was the, I guess, the art of good management and good leadership, you know, what the difference is between the two. So I wouldn’t say there’s any one bang event that’s a mistake or regret, but it’s probably, as ever, it’s probably I would have put more emphasis on it sooner had I known that how important those skills would be.
Brian Tarran
Yeah, but I think that’s understandable to a certain extent. If you’re coming from, I guess, a role that’s very hands on, doing things yourself, getting into the messy details of a project, it can sometimes be hard to kind of take a step back and adopt more of a kind of leadership, management position, can’t it?
Niclas Thomas
Yeah, definitely. Yeah, definitely. I would agree with that. And I think it’s also, I’d probably say for a lot of people starting out, and certainly it was for me, that the technical– the technical aspect is probably why you get into a role in data science in the first place, that you just love solving problems, basically, whether that’s with code or with pen and paper. And so that’s, that’s what you want to do. And getting your mind focused elsewhere away from that is probably not viewed as the most fun thing to do, I probably wouldn’t have, when I was starting out in 2014, 2015, I probably wouldn’t have thought it was as fun or as interesting to do that as I do now, maybe. So I think that’s the other reason why it probably doesn’t get as much focus earlier on in my career anyway, at least, as it probably deserved.
Brian Tarran
How do you think your– how do you see your role, I guess, evolving over the rest of your career in data science?
Niclas Thomas
I suppose on a personal level, for me it’s, I’m always thinking of what, 10 years down the line, do I still want to be focused just on data science? Or do I want to be focused on a data role, more broadly? I suppose that’s always the main question to ask. And so by that I mean, looking at data engineering as well, data analytics, and being responsible for a wider group. I think the way the field is going anyway, I think a lot more companies seem to move to vertical management rather than horizontal. So by that, I mean having heads of data in different areas of the business. So rather than having a head of data and a head of analytics, you might have a head of data for certain aspects of the business and another head of data then that’s responsible for both in other areas of the business, then. So either way, I think that the broadening of responsibilities and not just being responsible for data science is probably one way I would see my career potentially moving. At the moment, I love just focusing just on the data science, I’m really happy doing that now. But I think that could be one way that my focus changes in the future.
Brian Tarran
What personal or professional advice would you give for anyone wanting to be a data scientist?
Niclas Thomas
Yeah, so first of all, the balance between the soft and hard skills. I think I’ve alluded to it before, but the– don’t put too much– I mean, still emphasise on the technical skills are really important, but don’t feel like it’s the be all end all. I think just understanding the softer side of how you communicate, how you tell a story, for example, and storytelling with data, I think is really important. So I’d say that’s probably one focus area. I think that the second would probably, and maybe it’s a harder one to act on, but being passionate, I think, because whenever I’m looking to recruit anyone new into my team, I think it’s as much about understanding what the potential of that person is as is what is their current performance or where their current capability is – how good they could be in the future is arguably more important. And I think a lot of that comes to ultimately someone’s– whether they have a fixed or growth mindset. So by that, I mean, ultimately, do they want to learn or not, and if they really want to learn, as a lot of data scientists do, but if they have a huge passion for or about data science, and wanting to learn about just how to get better – whether that’s a better coder, better at maths, anything around that – then if you have that attitude, I think then it’s, A, you can have a great impact on our team, but B, I think it’s a sign of someone who can be a great performer in the future.
Brian Tarran
So what do you think will be the main challenges facing data science as a field over the next few years?
Niclas Thomas
I think probably, certainly, currently maybe living up to the hype, I suppose. And matching I suppose the classic Gartner Hype Cycle of, it feels like we’re probably at the stage where there’s a lot of– the hype has been around for a few years of data science now and I think making sure we tackle the right problems, I suppose, is one of the – and by ‘we’ I mean, Next as a business or whatever business we’re working in at the time – I think it’s making sure we’re working on the right things. Because I think a lot of people will be keen to have data scientists as part of their work and the product they’re trying to build. What is the best place to spend our time, and what projects we should be working on most I think is– becomes important then because, as I say, there’s a huge demand for data scientists time, I think, in every company. And so choosing where we spend that time wisely, I think, becomes the key challenge and the important decisions for, especially for a head of data science like myself to make then, to make sure we’re best using the team’s capacity, then.
- Copyright and licence
- © 2023 Royal Statistical Society
This interview is licensed under a Creative Commons Attribution 4.0 (CC BY 4.0) International licence.
- How to cite
- Tarran, Brian. 2023. “‘I fell in love with math, really, and fell into data science because of that.’” Real World Data Science, October 4, 2023. URL