TELLING STORIES OF VALUE WITH IOT DATA VISUALIZATION BAREND BOTHA VIDEO TRANSCRIPT Tell me a little bit about yourself and your background in IoT. I came from a web development and design background and I moved into flash development. Quite a few years of doing that, I moved towards mobile development. In the last couple of years, I ve followed the Internet of Things quite a bit and when I saw data v isualization starting to become quite a big thing, because of my background in flash I thought that s a great entry point for me. That s kind of how I got to get involved in IoT. I live in London and I recently started a website called www.iotdataviz.com. That is more just to be kind of an exploration tools that are available but also to release articles and possibly interviews about data visualization and the platforms and tools, etc. Nice, that s a good resource. Yes, it s still growing. Why don t we start from the basics? Why do data visualization at all? I think at the moment a lot of consumers, business people, and government are starting to realize the big potential that is hidden within the data. But I think we need to look at what is currently happening in the world. The whole Internet of Things exploding at the moment, even when I look at social media and people seem to action with that Likes, uploading content, sharing things they re all getting aware of how their actions are represented in statistics even though they re not from that background, they re starting to get a bit of an idea of that. Now if you take it a step further, if we introduce wearables, then you see people are very much engaged in that sort of things. Because it s their health, they re monitoring themselves and that kind of thing. Zoom out a bit further and we re looking at the city. That s quite a jump but that is true. All the problem areas within cities that are IoT has got the potential to address a lot of these make it run smoother, saving costs, that kind of thing. To answer the question, data visualization allows us to look at the actual data and explore it in a way that that is understandable to most people. What it could do is for example, it could make things that are non-obvious, quite clear, aggregate different data sets together and show you visually how that work together. This is part of the toolsets in order to gain the insights that people then use to mak e business decisions or health decisions, etc. Paint a picture for us. You re looking at a screen, is it a 3D model that you re rotating? Is it animated? What is the process actually? It very much depends on the domain because IoT is broad and I felt when I first went into it, it is daunting because you ve got all these verticals and all these different problem domains, and the solutions kind of reflect that. The traditional way to do data visualization is you would first look at IOT-INC. 1 AUGUST 16, 2016
who s your audience, as in design, who s your target audience? That is important because if you re doing some visualization within a company, that s different from doing visualization for public for example. I would say that s the first step. And then after that is what is it that you re trying to say? What s the function? Those two are quite important. Things like budget constraints of course are important as well. That s the initial, before you actually start doing anything. Then you need to look up the data itself and see the data you ve got to work with, is it correct, is there anything missing? That s even before you re visualizing? Exactly, that s just looking at the data itself and trying to see the interesting properties that you could use for the specific problem that you want to address. Once you ve done that then you can actually start to explore the data visually. Now this will not be the final product, this is just you re analysing it in a visual way because a lot of the value that comes from visu alization is the fact that that it s not obvious, sometimes it s just data so you wouldn t know. Once you start to do visualization you ll see there s an outlier over there or things like that. After you ve done that you will get a good idea of, in your dataset, what s important or what s not then you actually start to say, okay, this is perfect for the function, the purpose, now let s choose the correct visualization type to use. And that will then also be iterative process because it s a design process so you ll go through multiple permutations until you get to the right visualization that tells the story the best way. It s not just about stories, I know story-telling is a big part of data visualization but it could also be for exploratory purposes or that kind of thing. At this phase, you really want to look at what the visualization comprises of, things like colour, more design elements, annotations, what are you using all those things are important and trying to make it as clear to the purpose as possible. That s kind of brief outline. What I m hearing you say is you need to first look at the problem you re tryi ng to solve, make sure the data is up to the task, it s complete and so forth, and then you re choosing different visualization templates (I don t know if that s a good way of saying it) or different types of visualization to put your data through, then you look at it. I want you to give u s an example and then maybe the visualizations I d like to understand them a little bit the different types of visualizations. This is maybe quite a large example but we would deal with traffic and weather data within the city for example, also some of the camera data and that kind of thing. You overlay that on a 3-dimensional model of the city. By the way, this is a commercial product that is available and is being used and deployed at the moment. This is quite on a big scale but you would see if the data is coming in, is it correct, does it relate to the problem? Then you would aggregate that in some way. What does that mean? You would put it together, overlay it over this 3D representation but also, I think this is where it gets interesting because it s quite complicated, different streams of information. Once you put in a dashboard of some kind, you then start to see, oh, there is something interesting, I never thought that that would influence the traffic for example. IOT-INC. 2 AUGUST 16, 2016
What I m wondering is, is it always a geometric mapping on to something? Is it the data itself could be the geometry in itself? What are we looking at usually? Because that s a great visualization, a 3-dimensional model, you map on top of a 3-dimensional model It could be very, very simple; it could be streaming data that s coming in from some device. So just graphing it then? Yeah. It could be as simple as that. It all depends on what is the problem? You always have to take it back. And also, what is the level of your audience. So that s why it s important because if you re presenting something to a group of data scientists, you could go quite low. If it s general public, they don t have any idea. That is what makes it all so exciting because it s about representing some of these abstract concepts as well. Exactly, that is what you re doing, you re representing these abstract ions into some visualization. What are the steps then? Take us through the mechanics, you ve kind of talked about a little bit already but maybe just walk us through the mechanics of what you re doing here. As I said, you need to make sure that you know who it is for and what the purpose is. It s going to be like the fidelity of the data, how much detail you want to put in there or how low level. Exactly, and that s a really good point is the resolution of the data. How much are you dealing with here? Do you need to reduce it? And then you get into this phase where you re actually looking at the data itself to make sure that it s got everything that you need. And also, possibly we ll have more. You wouldn t use old information, you would then go and find the bits that are relevant and extract that, and then build upon that. After that step, you go into this phase where you are actually exploring it to see, have I got it right? Are these the most important elements or properties? and then after that, you say, okay, we ve got it, this fits the purpose, now let s find the clearest representation for our purpose. You mentioned telling a story, which I found was intriguing. What does that mean in the context of data visualization? After this analysis phase, you might find some pieces of the puzzle or some realization that the visualization gives to you. Especially if you re dealing with time series data, you ve got things changing over time and this is especially important for IoT because a lot of it is location-based, time-based data. With things changing and that s where the value lies. Things like population movement, weather patterns, that kind of thing, you can represent that visually and it paints quite a strong picture and it allows you to tell the story. But you ve got to be careful with that as well because you can beat the data until it tells a story you want. You have to be kind of unbiased and hopefully take a good journalistic approach toward what you re doing. So the story then is generally a temporal based story, kind of what s happening over time? IOT-INC. 3 AUGUST 16, 2016
Sure. There are quite a few good examples. Is it put in form of an animation then? Yes. So you would use animation to tell your story, to indicate the change. That can be quite effective. Again, you need to be careful because you don t want to create the wrong response. And I think that s where a lot of the good or bad visualization comes out. In the interpretation of it? Or is it excluding certain data and emphasizing another? Exactly. You could bend the purpose towards what you want to say. I don t think that s anything new but maybe when you see it visually, maybe it s more powerful so maybe it s reinforcing the fib to a certain extent or stronger way, I don t know. Why don t we talk about tools? What are some technologies and some tools that are used for doing the visualization? As I ve said before, for IoT the problem is kind of domain-specific. The example of the 3D city I mentioned, that s a real platform, a real product out there. There s actually more than one and some deal with urban planning for example. Related to that, you ve got geo and map-based tools. There s quite a few out there, including ones that allow you to work with satellite imagery and that kind of thing. It s really outstanding that you can access that kind of information. A lot of these you have to pay but still you ve got the ability to access that. With the map-based visualizations, that s where sensor location-aware data becomes really, really valuable. It s surprising how many different tools there are available. For doing geo-spatial visualization? For doing that kind of thing and also open source tools. You do have the tradition al data visualization tools that deal more with the financial side of things. More like BI stuff, right? Which is a little bit less sophisticated. Exactly. I think it s not necessarily the customize ability of it. Also, the learning curve for some of these tools, what you can or cannot do, what kind of interactivity does it allow or does not allow is it webbased, is it a stand-alone platform, is it free or you have to pay. All of these are considerations to take. Then you ve got your languages like Python R and also JavaScript, there are loads of libraries out there and rightfully so because of the website of things being so popular although Flash died. That is exciting for me because the things that people can build for free now and if you combine that with the ability to buy hardware for cheap and program like Arduino. I get very excited about that side of things because I think what kids are growing up with today building robots, building satellites. I think we underestimate how they perceive the world, growing up with ipads, iphones, small frames, all of that stuff, apps. I think that s an exciting intersection and I m not sure if people realize yet how big it could be from that perspective. How many designers you ve got out there, whatever type, with data science getting such popularity now and rightfully so, there s a big need I think. IOT-INC. 4 AUGUST 16, 2016
It s interesting because I didn t really think of it that way. What you re saying is applying the design sensibility to science in a sense and that s kind of where you came from in a way. For me personally, if you work on Flash projects, mostly that would be advertisi ng. A lot of times it would be revolving around selling some product so I think this now you ve got meaning behind and that s really, really exciting. That s really exciting. Where can our viewers find out more about you and what you ve been doing? You can go and visit www.iotdataviz.com and you can look me up on Twitter, it s @barbotha. I d love to hear from people especially on IoT data visualization. As I m saying, it s a growing platform so please feel free to connect with me. IOT-INC. 5 AUGUST 16, 2016