Welcome to episode #081 of the Super Data Science Podcast. Here we go!
Today's guest is Data Visualisation Designer Nadieh Bremer
If you are interested in the more creative aspects of data science, while still maintaining a strong focus on the technical challenges, data visualisation might be the path to follow. Nadieh Bremer shares her own journey with us today from data analyst within a top consulting firm to building her portfolio and having her own freelance clients, including the opportunity to work with Google, who approached her directly!
You will hear us discuss our experiences with data visualisation software d3.js, working as a consultant, and starting a business. Nadieh will share her top data visualisation tips, as well as her tips on presentating a visualisation.
Jump right into this value-packed episode!
In this episode you will learn:
- The Relationship Between Algorithm and Visualisation (12:30)
- The Benefits of Consultancy Work (14:50)
- Starting out as a Freelancer (23:07)
- Top Data Visualisation Tips (30:50)
- What is d3 and How Long Will it Take to Learn? (42:14)
- Creating Presentable Visualisations and then Presenting Them (49:19)
Items mentioned in this podcast:
Kirill: This is episode number 81 with Data Visualisation Designer Nadieh Bremer.
(background music plays)
Welcome to the SuperDataScience podcast. My name is Kirill Eremenko, data science coach and lifestyle entrepreneur. And each week we bring you inspiring people and ideas to help you build your successful career in data science. Thanks for being here today and now let’s make the complex simple.
(background music plays)
Welcome back, everybody, to the SuperDataScience podcast. I've got a very, very exciting guest on the show, Nadieh Bremer is a person who I've been a fan of for a long time. She is the creator of the website visualcinnamon.com. If you haven't seen it, you have to see it. It's got some of the most beautiful visualisations there. And before we continue, I wanted to quickly say that this podcast is one of those rare podcasts where it's also available in video, so if you are around your computer, then I suggest you check it out, check out the video recording for this podcast at www.superdatascience.com/81, and there you'll Nadieh and you'll get inspired by her and you'll see how we're talking about all of these things, the same things as in the podcast, exactly the same, but with video. At the same time, if you're listening in your car, or on a run, or a jog, I totally understand. That's totally cool. Keep listening. You'll get exactly the same insights and then maybe some day you can check out the video as well, or just have a look at parts of our conversations there.
And so, back to the podcast. What are we going to talk about? We're going to talk about visualisation. How visualisation is super important in today's world, how it tells stories in data science, and what tips and hacks you can apply to your data visualisations.
So stay tuned. You will get lots of tips and insights. You'll also get some insights on freelance work and shifting careers. It's a very exciting podcast. I can't wait for you to check it out. And without further ado, I bring to you Nadieh Bremer, a freelance data visualisation designer.
(background music plays)
Welcome everybody to the SuperDataScience podcast. I've got a very exciting guest, of whom I've been a long term fan, Nadieh Bremer, on the show. Nadieh, welcome.
Nadieh: Thank you. Thank you for having me.
Kirill: So where are you calling in from?
Nadieh: I'm calling in from a really tiny town outside of Amsterdam in the Netherlands.
Kirill: Ok. Wow. I've been to Amsterdam, and as we discussed, you have a lot of pigeons in that city. Do you have a lot in that little town as well?
Nadieh: No, no, no. Our town is in the green. So I look out and I see cows and horses and sheep. But I hardly ever see pigeons. Which I am thankful for!
Kirill: Cool. But yeah, thanks for coming on the show, and probably nobody knows the story except for me and Nadieh, because I have just told her about it, that the first time I encountered her work was when I was looking for a self-organising map implementation in an open source software, and you have a fantastic implementation in R which you showcase on your website. Unfortunately, you couldn't share the code because it was while you were working at Deloitte in the Netherlands, but nevertheless, that was very inspiring to me. And ever since I've followed your work. So tell us what exactly is it that you do?
Nadieh: I visualise data. So I make insights become visible to people in a way that's like converting numbers to actual visual encodings. And I do that either to inform people – it's always to inform people. But I also try to make it visually engaging, so that you are also drawn into it and want to investigate actually insights that can be found.
Kirill: Ok, ok. That's really cool. And like I would probably take it to the next level. You come up with very innovative and creative ways -- like I had a look at your website, and just some of your recent ones, like the one you did for Google Labs, or the one on the butterflies? What was it called? The Marble Butterflies?
Nadieh: Yeah, exactly.
Kirill: Fantastic. It doesn't even look like real data. Is that real data you used for The Marble Butterflies?
Nadieh: It's real data. I manually gathered data on butterfly species, and colours and wingspans, and that is what's guiding this visualisation of marble butterflies is I'm relying on a dataset of about 90 different butterfly species.
Kirill: Wow. And how does the visualisation move across the canvas, is it random movement?
Nadieh: It is random, yes. So the idea is that every animation frame in the browser, a butterfly is created and it sort of flitters across the screen, and that is sort of randomly generated, which way they are going to go. So the colour, the size, and the look of the path is defined by data. But then the actual path across the screen is more just random, and at some point it tries to go up, and at some point it might go into a turn, or a different turn. And then lots and lots of those butterflies over on top of each other create this sort of filled canvas with lines and such.
Kirill: That’s so cool. And how long did that take to create?
Nadieh: That was actually pretty easy because I wasn’t really trying to make data insightful. It was just “We have data. Let’s make it look in a way that’s mesmerizing to watch.” That one took me about 20 hours from start to finish, but that was actually pretty fast for my typical personal projects.
Kirill: Okay. But that was mesmerizing. I’m just looking at it, it’s hypnotizing and maybe even calming with some people. What about the Breathing Earth? I think that’s one of the latest ones that is featured on your website, even though you created it last year. How long did that take?
Nadieh: It was actually created in April of this year. It took about 80 hours to create, and that was mostly because I wanted to learn some new tools to visualize things online. I was typically using d3, but that one has a limit on the number of objects that you can visualize and this time I had 50,000 circles representing the Earth throughout the year in terms of how green different patches of the Earth are. I had to learn Pixi and WebGL and these kinds of other tools and that took a lot of time, figuring that out. That was the bulk of that. And learning how to deal with satellite data was something that was also new to me, so figuring out how to use that in R and get out the data that I needed for my visualization.
Kirill: Okay. That’s cool. So that one was created with R?
Nadieh: The data was prepared with R, but the visualization was eventually created with a combination of a little bit of d3 and a lot of Pixi and regl.
Kirill: That’s fantastic. I love how we just went from you’re an artist, and you’re very creative, and you work with butterflies to you talking about GPUs and how technical you can get in all these things. That’s so cool. So there we go, real data science in action. Let’s take a step back. Let’s see how you came to this place, because this is a podcast about careers and how to inspire, inspiring people to take on different journeys in the space of data science. How did you start? What did you study at university and where did you go from there?
Nadieh: I studied astronomy. I really loved it. I’m still fascinated by everything astronomical and all those things, but I knew at the end of my Master’s that I didn’t want to continue on to do a PhD because I really hated writing papers. So there’s no standard way to go from astronomy into something that isn’t astronomy. So I looked at a lot of things, and eventually ended up at a consultancy branch of Deloitte. They were just starting up an analytics department within their business intelligence department and that really intrigued me, the idea of still trying to dig through data to find insights, but no longer about stars and simulations and about new galaxies, but about what people buy. It still intrigued me to find the insights.
So that’s how I started there. I was working as a data scientist, but the first two or three months, it was mostly learning, machine learning, because it was so new in the Netherlands then, that I didn’t have it officially during my astronomy years. Eventually, I came to understand that some of the things that I was using had a proper name in machine learning, but I didn’t know it at the time, so I was really going through all kinds of papers learning about support vector machines and linear regressions and these kinds of things.
I was doing that in the next three or four years, just mostly data science, but as consultants, we were always external at our clients, which means we had often had to present back to our clients what we were finding with PowerPoint, and that often implied making some sort of visualization of our data and our findings. And then slowly through that, I discovered at some point that I was spending more hours creating a good and understanding visualization than I was having fun in getting 1% extra accuracy in my predictive model. In the last year before I started my third year at Deloitte, I completely switched focus from data science to learning data visualization more and more and becoming the expert within our department on that.
Kirill: Interesting. Was it the case that others, after they did their insights, they’d come to you to create the visualizations?
Naideh: Well, in the final year it was a bit of a blend. I was still doing a lot of data science because it really depends on what the managers can sell, and I had 40 hours to fill, so I still did data science, but I was giving advice to other people on how they might visualize something and then some of the bigger visualization jobs I would take on, but I would still be involved in the data munging and the analysis in some part. In our team we would have a division, but I still feel that to make a really good visualization, you need to understand your data, and the best way to understand your data is to have some sort of feeling with it that you might get through cleaning it and analysing it.
Kirill: Okay. That’s a very interesting link. You mentioned that you enjoy personally more the visualization part than the modelling, and also that it’s important to do both in a way, at least to some extent, to understand the data better when you get to the visualization. What would you say is most important from the perspective of the client? Or it doesn’t have to be a client. If you’re working in an organization from the perspective of your boss, or the manager, or the CEO, or the company even, how would you rank the modelling versus the visualization? Which one is more important or less important? How do they go hand in hand?
Nadieh: That’s a tough one. I mean, they’re so dependent on each other. I have found, though, sadly in a way, that people could have spent 3 months on creating a fantastic algorithm that does wonderful things, but if they cannot get the output across to their audience, usually to our less technical clients, then that was all for nothing, because they wouldn’t use it, they wouldn’t implement it, and it would just sort of fall by the wayside. And then you might make a simple but very effective visualization of that in just a few hours and that could sort of switch the tides around it completely because suddenly a client could understand “Oh, that’s what you’re doing. That’s what I’m seeing and that’s what it’s used for.” I’ve seen that happen, but I feel that a visualization needs the fundament of the data and the insights and the algorithm to create something in the first place. I mean, an algorithm can exist without a visualization, but a visualization might not exist without data preparation or the algorithm. I think an algorithm can really benefit from a good understanding of how to visualize it.
Kirill: Okay, gotcha. But at the same time, I would probably say that there’s a limit to how much you should spend on an algorithm, right? You know the 80/20 rule. You make an algorithm with 90% accuracy, or you could get it to 91% accuracy and spend extra week on it, or you can take that week and spend it on the visualization, and I think that will be more beneficial. So it’s that kind trade off.
Nadieh: Yes, I completely agree with you.
Kirill: Yeah, in terms of time. Okay, cool. So, you worked at Deloitte, and I also worked at Deloitte, and it’s really cool because I found out about your visualization just after I left Deloitte, about your sums algorithm and about your website. By the way, those who don’t know, the website is visualcinnamon.com and we’ll mention the links at the end, but if somebody is watching on their computer and wants to look around, they can find it there. Let’s talk about Deloitte for a second. What were the best things that you took away from Deloitte? Did you enjoy your time at Deloitte, and if you did, why?
Nadieh: Oh, yeah, I fully enjoyed my time at Deloitte. I very much enjoyed the clients that we were working for, the different kinds of challenges that we were facing. It was always challenging in a way that we were doing something new, or had to figure out our own way, or dive into maybe some new algorithm that we hadn’t used before but seemed interesting to use here. And I also very much enjoyed my colleagues, who were just as crazy about analytics as I was. And just the fact that I also very much enjoyed the diversity of being a consultant, so that you actually see a completely different client every few weeks, or every few months and you have to become an expert in their field of work in no time at all. And then you only have a certain number of weeks or months to do something, so you’re working towards something that has — the deadline is always looming in a way, which makes you — as a team, I felt you come together more quickly.
The teams change, every new project I was with different people, so I really liked all of that diversity and just the things that we were working on. It was from retail companies to fast food companies to insurance companies, it was all over the place. I guess I also enjoyed the fact that when I said I wanted to do data visualization, that they were supportive of that and they were trying to find and sell visualization-oriented data analytics kind of projects.
Kirill: I totally agree. And I also like that you learn a lot through all these engagements. You have to go through different tools, you have to go through different methodologies and explore new things that you can’t even find anywhere else and come up with your own new ways. It really develops that side of your creative thinking and ability to think on your feet and get these answers quickly. And if it doesn’t work, move on, get the next one, move on and then finally find some way to quickly get this thing up.
Kirill: Yeah. It was fun time. For me it was like going to university for the third time. (Laughs) That was cool. Okay, so you spent 4 years at Deloitte. That’s quite a lot. I heard that the maximum churn that they experience, Deloitte and other big consulting firms, they experience like 5 years and people usually leave after that. So did you leave because you burnt out, or did you leave because you had other things that interested you?
Nadieh: It was a combination of things. I was having a good time, but also the typical path upwards within a consultancy is the expectation that you want to become a partner, you want to rise quickly, become a manager of sales stuff. That wasn’t driving me. I just wanted to have fun at my job, and I wanted to become more of a specialist and a generalist. I wanted to help people, but I didn’t want to be a people manager, I wanted to be teaching others what I had learned on a technical aspect. And I definitely didn’t want to do sales.
So at some point I was at the level right below manager and I was coming up to become manager and I felt that was probably not where I wanted to go, but I also didn’t want to stay at the level below manager in that sense that I would be earning to go up, but because I was having a good time at the place where I was and I wanted to stay there. But there was no other way to show the appreciation that — it’s quite difficult to explain it in a way. I felt like I would have—
Kirill: Their only way to show their appreciation for your work was to promote you, but that’s contrary to what you want. You wanted to stay in a technical—
Nadieh: Yeah, in a way. But I didn’t want people that did want to become manager pass me by in a way even though we were both equally appreciated. It was on my mind. But then I was contacted by Adyen, they were looking for a data analyst and I told them I wasn’t interested in that, but I was interested if they needed a full-time data visualization designer. They thought about it and they were like, “Yeah, actually we could really use that as well.” So then I came there and there was a great vibe and I was also interested, after 4 years of Deloitte, how would it be actually working at one place for a long period of time that you could really dig into that data. This felt like a really good place to try that out and after 4 years it felt like, “Yeah, maybe I should move on and try something else.”
Kirill: Okay, fantastic. And what does this company Adyen do?
Nadieh: Adyen is a fintech company, financial technology. They handle payments. So if you take a ride through Uber, or buy something from the App Store, or get your clothes at Nike, and you pay for that, it’s actually Adyen that handles the payments. They make sure that they sort of get your data and they process it in the right way, they check for all kinds of risk and fraud settings. If that bank’s computers are not working, they try a different kind of route to make sure that it eventually ends up with the bank and ends up at other side at Nike or Uber. So they process payments in that sense so that the companies who sell products or services don’t have to worry about that.
Kirill: Okay. So kind of like Stripe for credit cards?
Nadieh: It’s exactly like Stripe in that sense, yeah. Adyen does more payment methods though, I think.
Kirill: Okay, that’s interesting. So what was your role at Adyen?
Nadieh: Anything that had to do with data visualization. I’ve made a lot of dashboards really from start, trying to figure out what we need to visualize, to designing it, to testing it, building it, to testing it some more, to making better visuals for marketing slides, to making internal tools such as an organizational chart, or we had these big screen where you could sort of see how things are going throughout the day, things for that. So it was anything where data visualization was involved, I could be helping that out or I could be leading that.
Kirill: Okay, interesting. Can you share some of the tools that you used predominantly in your work at Adyen?
Nadieh: Yeah. For sketching, pen and paper. (Laughs) I always start with just pen and paper because it’s so easy to just draw whatever you want and you don’t have to be an artist for data visualization – it’s just squares and curves and circles. But afterwards, I would typically go to Sketch or Illustrator to make some mock-ups to show it to other people. And then often to d3 — the dashboards were built in d3 and lots of other things as well. For the marketing slides, though, it was often a combination of R combined with Illustrator for final touch-ups.
Kirill: Okay, very interesting, so R and d3. We’ll get to the tools in a second. How long did you spend at Adyen, how many years?
Nadieh: I was there full time for a year, and then part time for another 6 months. So a year and a half.
Kirill: And now just recently you moved to doing your own thing, completely independent, by yourself. Congratulations on that.
Nadieh: Thank you.
Kirill: How do you feel?
Nadieh: I feel super happy being able to do this because it’s exactly what I was looking for. Now I can choose my own jobs that seem interesting and work with clients that I’ve never dreamed of working with, like Google, for example. It’s been fantastic already.
Kirill: That’s so cool. A lot of our listeners would love to be independent freelancers just like you, maybe not just in the space of data visualization, but more inside of analytics or modelling or AI or anything. Can you tell us a bit more about that step? Was it a hard step to make, and how did you prepare for it? You mentioned that you were part-time for six months. How did you prepare yourself on one hand from a professional side that you will have work to do and you will have clients to work with, and on another hand from an emotional side, to take this leap and to go on your own?
Nadieh: It’s a little bit twofold in that way. I’m very risk averse. I’ve been thinking about freelancing for a long period, but I always felt like I can’t handle the fact that I don’t have exactly that amount on my bank statement at a certain point in time. The first thing came when I was on vacation last November, and Shirley, a dear friend of mine who’s also data viz designer, we were both approached by Google to do a project for them. I was full-time employed, so I couldn’t, but I felt like I just don’t want to miss that opportunity and work for Google.
So I just decided that I want to try and start the freelancing thing. And what I had in mind was to do it part-time, so I talked a lot with Adyen, and eventually came to a decision where I would stay on three days in the week with Adyen, and then two days I would do my own thing. That would give me enough time to pick up Google for the first three months and then continue doing Adyen, and that would still give some income, enough income that I could pay my rent and get through it and such, even if after Google nothing came anymore.
That at least for me financially gave that enough time because then I could sort of in those three months that I was doing a thing for Google, I could also mention and promote myself as a freelancer and see if I could get other jobs for the months afterwards. That was actually going so well that five or six months later — I wasn’t doing administration well enough, so I was looking at how many hours have I sold and I saw that if I was going to do everything, I would be completely full-time from July onwards and I’m like, “Yeah, that’s not possible. I only have two days, so…” And then I thought, “Well, now I should be able to handle full-time,” so I quit at Adyen and I started full-time because I had that sort of insurance that at least for the first two or three months after July I would be able to fill enough time to get enough cash flow, to get that financial security feeling back.
But on the other hand, something that also had to do with it is that I — when I had this sort of stroke of lightning, I was like, “Oh, maybe I’m more visualization than data scientist,” I started a website where I shared all of my personal projects and tutorials like the ones that you read on R and sum because I really wanted to teach other people the things that I had learned, but I also wanted to share what I created to get input back, like “Does this work?” or “Is it a bad visualization?” And I’ve been doing that for the past two or three years just because I was so enthusiastic about visualization, and that has helped me — I didn’t know that at the start, but it has helped me to get a presence in the field of data visualization, and that helps me get clients now.
So I didn’t start it with the idea that I would become a freelancer three years later, but right now, I’m so benefitting from the fact that I started that blog three years ago. Also because I used to get e-mails now and then from people asking me if they could hire me and I would have to say no, but now I can say yes.
Kirill: Yeah, that’s totally cool. I was just about to ask that. How did Google find out about you? Probably through your blog, is that right?
Nadieh: Yes. Both through my blog and the fact that I’m doing a personal project with that friend, Shirley Wu, called DataSketches, and they picked up on that as well, that combination.
Kirill: That’s a very important step for those out there wondering, “Oh, yes, if I get contacted by Google, then of course I’ll become a freelancer.” Well, you know, to be contacted by Google, or by any other company, you need to be visible and you need to have a name for yourself out there. So if you don’t have a blog, first step is start a blog. Have a LinkedIn, start a blog, we’ve talked about this on other podcasts and you can listen to those, but definitely having a presence and showing not only your skills, but also as you just mentioned, your passion for a certain area, even without the thought of how this is going to benefit you in the future, how you’re going to monetize this. Because a lot of people start blogging and think, “How am I going to put advertising on it? How am I going to make money off of it?” No, you just do it because you’re passionate about it.
And I really appreciate what you said, that you wanted to share and educate, you wanted to share with the world and educate people so that they could do the same things that you’re doing and empower them. That’s very powerful to showcase to others, so whether you want to get a full-time job with a great company, or you want to do freelancing, that’s always going to tell people about your strengths and what exactly you’re interested in, what you’re good at, and that will always be a good case for you when you’re bidding for these projects or looking for jobs. Yeah, that’s a great way of doing it and it really worked out well for you, so that’s very exciting.
Kirill: Okay. So, you’ve started this freelance career. How does it feel? Does it feel different? Does it feel more stressful or less stressful than working inside a company? From my own experience, on one hand, you get the illusion of freedom of time, but on the other hand you tend to take so much on, you’re working so much. But again, on the other hand, it doesn’t really feel like work because you’re having so much fun. How is it for you?
Nadieh: It’s also a combination of things. I read some books when I started freelancing on how many hours should I sell and I try to stick to that, but sometimes something really amazing comes by and I go, “Okay, I’ll take that on as well.” But it’s true, the fact that you can choose or at least can have a say over the projects that you take on, it’s typically a more fun project than you might get if your manager tells you to do something. And sometimes, if I sit and think about it, I might be daunted by a big project that’s coming up now and I have to lead a few people and I’m like, “Oh, my God, that’s so much responsibility.” But then I just stop thinking about that and write down my to-do’s and follow them. It’s no use to linger on these things but just focus on the tasks and the fun parts.
Until now it’s been going okay. I have no difficulty with picking up some hours in the evening or over the weekend. And if it’s sunny outside I might take a long lunch, over an hour and a half, instead of just half an hour. But I pick that up at some other points. I have to admit, though, I don’t have super-much experience because I’ve only been full-time for like a month and a few days, but at least until now, I’ve been able to manage it.
Kirill: Yeah, that’s fantastic. All right, how about we talk about some visualization? Obviously you have all this experience about visualization. What are some of the advice that you can give to our listeners on visualization? Let’s start somewhere and then we’ll go from there and see what tips and hacks we can go into and things like that.
Nadieh: I think the most important thing seems so obvious, but is forgotten more often than you may think and that is to know your goal at the start. So even if you have data, you need to have a question that this data or the visualization needs to answer. What should people learn when they are looking at a visualization of this data? Make sure that you have that clear, actually write it down. I mean, you may think you know it in your mind, but if you can’t make a full sentence on the paper, then you don’t quite have it yet. Or maybe there’s a goal. Maybe they should be convinced or maybe they should take action, but make sure that you know what you want to get out of the visualization beforehand.
And then another tip is that maybe you should just try to broaden your mind in that sense. There’s more than just lines and pies and bars out there. There are so many different ways that you can visualize data these days, just look up Sankey diagram, which is an amazing visualization to show flows. And there are so many more visuals. I mean, you can’t put data into a bar chart or a line chart and think that it will solve everything and show all kinds of insights. Each kind of dataset combined with a question has visual forms that work better and that work less well.
And it’s good if you just take some time to think about if your visual form that you have in mind is actually going to show the insights or answer the question that you have in mind. Don’t just put it into the standard options that your tool gives them, but try and see what the broad scope of options is and think about all of them and choose which one of those will work best in that sense.
And the rest is more on the design front, the typical things, like try never to use pie charts, don’t make the non-data parts stand out, no black axes, very dimmed down, gridlines, these kinds of things, make the data stand out. Use annotations; if something is interesting about data, call it out, make a circle and say, “Oh, my God, look what’s happening here.” And don’t have a title that says ‘X versus Y,’ but say the conclusion, “The stock prices have been going up until blah-blah happened.” Don’t underestimate the use of text to help people understand the visualization.
Kirill: Okay. That’s some great tips. So just to recap: have the question or the goal already, broaden your mind, so use other diagrams, like Sankey diagram, for instance, and then some design tips like avoid pie charts, dim the axis, add annotations. What would you say about colours? I’ve noticed you’ve used lots of different combinations of colours depending on the visualization and the mood you want to convey. Would you say colours are important? And how do you pick colours for your visualization?
Nadieh: Yeah, colours in data visualization are very important because often they can actually convey data as well. Because if you want to distinguish different categories, different countries, it’s best to give them different colours. Or if you want to say something goes from low to high, you can say that’s like light blue from dark blue. So there are many ways you can make that, because people can see that easily, they can interpret that easily by using colours where light is low and darker is higher or more.
But they also have a lot of challenges, because you might have 220 countries and we can’t distinguish 220 different colours, it’s more like 8 different colours – that’s typically the absolute max that you can have in different colours. So there are always these challenges with colours as well, even though they are very powerful, and they can also help the mood indeed, should it be playful or more subdued, for the data itself. Or this visualization I made about Earth – green, of course, made perfect sense; whereas for a visualization I made about translations and words, it felt more like a grey “books” kind of scheme might work, and other times it could be more in a tasteful rainbow kind of scheme.
Kirill: Okay, so colour is quite important. We’ve talked about some design elements. What are some of your favourite visualization charts or types of visualizations?
Nadieh: Ooh, that’s difficult. I especially like the visualizations where the chart is in a way unique, where the chart has been custom-built to show that particular dataset. I have a lot of Pinterest boards where I save all of my favourite visualizations that I often refer to.
Kirill: Will you be able to share them at the end of the podcast?
Nadieh: Yeah, definitely.
Nadieh: Anybody can see them. There’s one called ‘Poppy Fields,’ it’s a visualization about all the conflicts that have happened across the Earth over the past 100-ish years, when it started, when it ended, how long it lasted, the number of people that perished, in which continent it took place, and it looks like a field of poppies. So it’s very visually appealing and there is so much information that you can gather from it by actually looking at it for any certain amount of time, but it’s also easy to interpret. It’s that wonderful combination of many layers of detail combined with intuitive visualization. That’s always difficult.
Kirill: Okay. This is actually the next thing I wanted to ask you. How do you go about the trade-off between having a lot of information but then too much in your visualization, versus having a smaller visualization, or a nicer looking visualization, but less information on it? How do you go about that?
Nadieh: I always find that one difficult. I don’t have a formula for that, though it has to do with looking at each case and deciding if adding an extra layer of detail would really add to these stories that people would be able to find, or if it’s just an extra level but it can’t really be connected to any of the other variables that are already present, so it wouldn’t be a 1+1=3 kind of thing, but it’s just something extra and it sort of stands alone. If it really lets people see the complexity of the underlying data more, then I’m all for including that.
But it is also thinking about how busy is my visualization already and do I think I can add something more? That’s really just a personal choice that I make and then try and ask other people if that is working or not and if I should take it out. Sometimes I succeed, and also sometimes I fail, and it doesn’t work and it’s too much and people are like, “I have no idea what I’m supposed to get from this.”
Kirill: Okay. So you do ask other people’s opinion before you deliver the final result?
Nadieh: Yeah, I often share the final result secretly with some people and they give me some feedback and I incorporate it if I can before I share it out in the world.
Kirill: Yeah, I think that’s important. When I was creating the course on R-programming, I encountered ggplot2, and the grammar of graphics, and I actually studied the grammar of graphics quite a bit and I was very fascinated by the layered approach that you have in the data, then you have some geometries, you have some aesthetics and so on. What are your thoughts on that? Do you follow that type of grammar of graphics approach, or do you have some other view on how to create your visualization?
Nadieh: I really use ggplot a lot and I quite enjoy it, although I have to admit I’ve never read the grammar of graphics. I know what it’s about, I like parts of it, but still I feel that for the sort of more creative approach that I try to take, I’m not too constrained by it. I feel that it has a good set of things to handle. It’s something if you don’t otherwise have your own thing, then it’s a good way to follow. But if you at some point want to break out, you know, know the rules and know when to break them, in a way, that’s sort of how I feel with those. It’s a good handhold, but I don’t want to be constrained by it.
Kirill: Yeah, totally. And just to be clear, I haven’t read grammar of graphics either. I’m just like you, I just studied little bits and pieces. I think so as well, that for certain applications, more kind of standardized approaches can be useful, but for more creative stuff that you do, maybe there’s other ways.
Okay, how about we talk a little bit about the tools now? We’ve mentioned a couple, we’ve talked about d3 and actually, when I was at Deloitte, funnily enough, I played around with d3, we even had a little competition inside our data analytics division on creating d3 visuals, which is really cool. You mentioned d3 and R. Before we move onto those, have you ever used Tableau?
Nadieh: Yes, I’ve used Tableau and I’ve used QlikView as well.
Kirill: Okay. What did you think about Tableau and QlikView? d3 and R are more advanced ways that you get into visualization, as I can imagine. Do you think Tableau is a good visualization tool?
Nadieh: I think it is a good visualization tool. I have to admit that I don’t use them anymore, but I can explain why. I feel that especially for exploratory visualizations, Tableau works really well because it’s so easy to just drag things through everywhere and get different views of the data. The reason why I’m not using it anymore is because your options are limited. I mean, I’ve seen people hack wonderful things together, but in essence you have a collection of x visualizations that you can use, and I typically want something that is slightly different or completely different from the default and to make those — even the small adjustments in Tableau, or especially QlikView, can be quite a chore and that’s what eventually drove me to d3, which is very low level, but it gives you — you’re more constrained by your imagination and the browsers themselves than the tool.
Kirill: Yeah. And your blog is a testament to that. When people check out Visual Cinnamon, they will see that you’ve created some incredible things. I wouldn’t even imagine that you’re going to have something like that in a browser just working — it’s crazy. It sometimes feels like it’s drawn on a paper, but then it starts moving and you’re like, “Whoa! It’s really cool.” So, d3 – what is d3? Why is it a visualization tool? Why is it popular?
Anyway, d3 is a way to bind data to these elements. So, I could say that I want to create 10 rectangles because my data is 10 elements long and I want to have the width of my rectangles depend on the data that is bound to them, so the first rectangle will get the data of the first row and the second rectangle will get the data of the second row and so on. So really, that is what d3 in its base does. It has a lot of things to make scaling easy, to make colours or lots of other things that are built around it that make it easier to sort of handle and draw these shapes eventually.
Kirill: That’s really cool. And how long would you say it would take somebody to pick up d3 on a very basic level?
And after that, it still took me about a year on and off, sometimes doing something with d3 in my free time before I got to the point where I could use it during my job as well, which then sort of accelerated afterwards and I could do it more on my job, learn more, do more advanced stuff at my job, learn even more and so on. I would say it’s a steep curve, but for me it was definitely worth it.
Kirill: Yeah, that’s really cool. I can attest to that. When I was learning d3, also just for fun, it took me I think 2 weeks before I had quite an interesting visualization of — I know you said don’t use pie charts, but I had pie charts. (Laughs) It was clickable and it was a visualization of — what I did was I took everybody in department at Deloitte and visualized what projects they worked on and what their skills are and you could click and come up with information.
Of course, to get to some very cool visualizations, very creative visualizations like Nadieh is doing, you need some time and you need a lot of practice. But to get some very basic ones like pie charts, scatterplots and so, just a couple of weeks of dabbling around in d3 can get you there.
Nadieh: That is definitely true. And yeah, you said that Blocks is d3’s version of CodePen, so that’s bl.ocks.org, but there’s also blockbuilder.org, where you can sort of code within the browser, which is really interesting. But yeah, it’s quite easy to just pick an example from somebody else, put in your own data and try and figure out how to make it work with your data and then make some adjustments to that.
Kirill: Yeah. And then you find this one thing where the colour goes transparent and you’re like “What is going on?” I’ve had this so many times. One part of the visualization for some reason is just not working like everything else — it’s different to an algorithm because you can actually see when something is wrong. If everything is right, then everything will look right. It’s good in a way, but it’s also so frustrating when you can’t find that one little problem in your code.
Nadieh: Yeah. I’ve had it so many times where at some point I was convinced my computer was being irrational and it wasn’t me. Of course, it was somewhere in my code.
Kirill: Yeah, that’s always the case. Okay, thank you for that overview of d3, I think that’s powerful. And what I wanted to talk about now is you’re also a speaker, this is your fifth or sixth podcast or something like that, and it’s not only podcasts, you go to conferences and you speak at conferences. Tell us a bit more about that because that ties in with — I’ve seen one of your speeches on YouTube where you go back to astronomy, and then you go into visualization and then you combine the two, astronomy and visualization, in that presentation.
Just about presentation – it ties in with visualization, because once you have the visualization, you need to present it. What would you say about that? How do you prepare visualizations in a way that when you’re actually presenting them to a client, that not only can they make sense on their own, but that you can add value through speaking about the visualization, actually being there and talking about it?
Nadieh: I feel that I do typically, if I create something for a client, I create something where they wouldn’t need me to have to look at it or use it or get insights from it. So, typically, if I do the design part of a visualization for a client, I have these screenshots of all my designs and the inspiration that led me to those designs, and I really talk through it like a PowerPoint slide in that sense. And then we talk about the different designs and what might work and which won’t work and where do we actually want to build some of this out into an actual working data viz.
But for a final data visualization, I’ve often had touch points with the client before that, before the final part, because you want to make sure that they are still happy with it, the way it’s going halfway as well, you know, is this going in the right direction, so it’s not the first time they’ve seen it and they sort of know how to understand it as well. So the final point where I’m presenting it back, face to face to the client, is typically more of an official thing, but unofficially it’s not the most important part of it in a way, because of these touch points throughout the weeks leading up to that.
But in my presentations that I do for conferences, it’s slightly different because then the people haven’t seen it before. And there what I do is I make my presentations in reveal.js, so my presentations are a collection of websites, HTMLs, and that makes me actually incorporate the interactive visualization on a slide in a way, because a lot of my visualizations are interactive, and I can make screenshots of that, I can make movies of that, but it’s not the same as actually interacting with it while I’m speaking about it.
So, during a presentation, I will make sure that I have access to the interactive visualization where I make stuff happen, so I can make steps happen where if I click on ‘next,’ some part of visualization might change or highlight certain parts that I’m speaking about at that point or have an animation running so that when I’m talking about it, people are seeing it happening on the screen. So for the final portions, it takes a lot of time to build my presentations, typically 40 minutes of presentation takes me about a month and a half of evenings and weekends, too, to code out, but it’s been worth it in the end because I see that it works a lot better than just PowerPoint slides with screenshots and movies, to have these fully functioning visuals that do stuff during a presentation.
Kirill: Wow, that’s really cool. Yeah, it would be very interesting to see one of your presentations. I think there’s a few of them floating around online. And I’d like to touch on what you mentioned for the corporate presentations. That is a very good technique when, you know, sometimes I like to leave it as a surprise, and then at the very end show the client and then you’re like, “Oh, they hate it, something is wrong,” so it’s a very good technique to work with them and show them the intermediate steps and work with them on the designs and get their confirmation throughout the way so that when the final day comes, you present and it’s actually something that they have actually been working towards. So if they don’t like it, then they’re kind of saying that they don’t like what they’ve been working on and they’re not going to say that, right?
Kirill: They’re obviously going to like that. So it’s a great way. Thanks for sharing that. And wrapping up our podcast, because this has been a very quick hour, it just flew, what are some of your final tips for those looking to get into the space of data science and be like you, do what you do?
Nadieh: I’m definitely not the first person to say this, but I feel like the best way is to be driven by passion, but also to do a lot. So the best way to learn is to either apply your own algorithms or create your own visualizations, and every time you do that, you learn something new that you can take on to your next project and then get it more and more advanced. But also share your enthusiasm to people so that they notice that. Even if you can do something very well, but you don’t like it, try not to mention it, so that people won’t actually know they can use you for that.
I used to be able to do SAS, but I hated it, so I never mentioned it to anyone. In that sense, people knew where to find me for the things I was promoting myself for. But you need to be able to do it. It might take more time. Probably, if you really want to get to it, you might have to put in more time that you can do at your work and you have to do personal projects as well to get at the front of data science to work with those algorithms that you might not be able to work with at work. You need to get a handle on it with personal projects in that sense. So that would be my advice.
Kirill: Fantastic. Thanks for sharing. So, guys, let your passion drive you and show what you want to work on and don’t be afraid to spend your personal time on these things that you love to get even better. Thanks a lot of coming on the show, Nadieh. What’s the best way for our listeners to contact you and follow your career and see some of your amazing visualizations?
Nadieh: I’m very active on Twitter, that’s just my first and last name together, @NadiehBremer, because the data viz scene really lives on Twitter. So if you want to know what people are working on, work in progress or final results, you can follow me there. And I have a website, visualcinnamon.com. Again, because I don’t expect people to spell my name correctly, so I thought at least my website should be an easy name. That’s where I share all of my finished projects and all of my blogs and tutorials and new presentations. I’m scattered around the web, but those two are definitely the most important ones.
Kirill: Okay. Are you active on LinkedIn?
Nadieh: Yes, as nbremer. I’m also on Twitter @NadiehBremer and also on Instagram @NadiehBremer.
Kirill: Gotcha. Okay, we’ll share all of those in the show notes, definitely. And one last question for you: What is a book that you can recommend to our listeners to help them become better at data science and visualization?
Nadieh: Let me grab it. I have a whole collection of data viz books, but I think anybody starting should start with this book, “The Functional Art” by Alberto Cairo. I like it a lot. It’s one of the first books I read as well because it is so diverse. It looks at both the psychology behind seeing, which I find super fascinating, to figuring out this question that you need to have and how that can affect your eventual visual state. He shows a few horrible ones that he turns into good ones by looking at them in different ways. He has been a graphics editor at several newspapers, so he shows a lot of results.
And finally he has interviews with people in the field that I admire a lot, so it’s very interesting to read about how each of them, in their own ways, handles the visualization of data. I really like it because it’s very introductory, but also very diverse and complete in that way.
Kirill: Fantastic. Thank you a lot for sharing that – “The Functional Art” by Alberto Cairo. So there you go, guys, that was Nadieh Bremer. Make sure to follow her on Twitter, we’ll include the links in the show notes. And once again, thank you so much, Nadieh, for coming on the show.
Nadieh: No problem. Thank you again for having me.
Kirill: So there we go. Thanks, guys, for listening to this podcast. I’m so excited to have shared Nadieh’s story with you. I hope you enjoyed the insights that she shared. Personally, my favourite was just the general sentiment, the general energy that Nadieh conveys about her work, about visualization, how excited she is. If you were watching the video version of this podcast, you would see how her eyes light up when she’s talking about data visualization and some aspects of her work.
So it’s very inspiring, very encouraging to see people who are so passionate about the work they do. And, of course, Nadieh’s whole story is very inspiring, how she didn’t give up on her passion, how she decided what is her passion, what she’s good at, what she wants to do. And even when a job opportunity came along for a data analyst, she said that she would prefer to be a data visualization designer and she created that job opportunity for herself.
So that’s a story to get inspired by and, of course, reach out to Nadieh, follow her on Twitter, make sure to find her and follow all of her new updated visualizations, also check out visualcinnamon.com, there are some amazing, fantastic visualizations there, and you can get all the links to Nadieh’s profiles on LinkedIn, Twitter and her website and more at www.superdatascience.com/81. There you can also get the show notes and the video for this episode.
And if you enjoyed this podcast, we’d really appreciate if you could leave a rating or review on iTunes, it would really help us spread the word into the world and get more people inspired about data science. And on that note, I look forward to seeing you next time. Until then, happy analyzing.