Storytelling with Data

According to SAS Institute, “Data visualization is the presentation of data in a pictorial or graphical format. It enables decision makers to see analytics presented visually, so they can grasp difficult concepts or identify new patterns.” It is one of the most important parts of Data Science. Every ‘piece to the puzzle’ when it comes to a data story is vital, but many would argue that one of the last steps, data visualization, is the key to a successful story.

________________________

"Today, it's all about storytelling and infographics, and visuals are a key part of that. All researchers want to connect the reader with the data, and that's what data visualizations and storytelling can do."

________________________

I had the chance to speak with a data visualization expert this week, Juan Ca?ada, whom also recently gave a visualization lecture for my Master class. Juan is currently the ‘Head of Maxwell Render at Next Limit Technologies’ where he has worked on projects ‘in the fields of simulation, computer graphics and big data’. He shared with me some interesting notes about Data Visualization…

https://www.dhirubhai.net/in/juancanadagarcia

Why is data visualization important for data science?

Our brains are supercomputers. While current machines are much faster than us in many areas, the opposite is true as well in a large number of fields. A very remarkable part of our brain is devoted to processing quintillions of photons that arrive into our retinas every second and making meaning from what we see. Our visual skills to detect clusters, patterns, outliers are superior to any other device to date. Data visualization is not just about making nice charts as some people believe, but about using the most powerful and the cheapest tool you have at hand to do analytics: your brain.

You have an educational background in Mechanical Engineering. So what got you interested in data visualization?

My passion for CAD and 3D visualization started even before going to university; I have always been interested in writing software to simulate the real word. When you simulate things using a computer, many times you have to display what you are simulating. In other words, you have to visualize information, and that typically involves an enormous amount of data. Therefore, the movement to other fields of data visualization such as BI was very natural.

Can you give an example of a real-life scenario in which visualization plays an important role?

It is difficult to pick just one given the amount of examples that come to my mind, but perhaps my favorite ones are related to education and learning using our visual thinking system. When kids study the Pythagoras theorem, most of the time they are forced to memorize things like “the square of the hypotenuse is equal to the sum of the squares of the other two sides.” That sentence that kids obviously find weird can be replaced by a simple drawing that anyone understands. The same idea can be applied to much more complex areas, from Fourier transforms to Cantor’s theory of infinite sets, and most of the time difficult concepts are easier to understand through good visualizations.

That extends to any other area, from business to politics. When humans have to deal with complex issues, there are few things as powerful as visual thinking, especially when it is combined with state of the art computer-based analytic techniques.

During the MBD lecture you gave, you showed us how the current company you are working at, Next Limit Technologies, uses advanced data analysis and visualization to produce realistic special effects for movies. This was very interesting for us. Explain to us briefly how exactly data plays a role in special effects?

People would not believe how much data is generated for making movies that contain computer-generated effects. An average frame of a complex visual effect shot usually requires large clusters that do computations for a long time. High-end movies typically require saving data in the order of some gigabytes per frame, sometimes even more. You need to render 30 frames per second for around 90-120 minutes, so just in terms of data storage this is a serious big data problem (and I am omitting aspects such as stereoscopy and virtual reality which multiplies these numbers at least by a factor of two).

Storage is not the only aspect that needs to be addressed here but in simulation software -while we work on the visual effects for movies our core business is in the engineering/design side that is at least equally challenging- computing times are critical. That forced us more than a decade ago to spend a lot of effort on developing techniques to distribute calculations across network environments, doing things very similar to the map-reduce paradigms that were formalized later.

Which visualization tools do you like the most?

I think most of the last generation tools (Tableau, Qlik, SAS VA, etc.) are very good and it is possible to create fantastic data visualizations with all of them… and the opposite is true: without enough expertise, any of these tools can generate visualizations that do not map data properly, therefore forcing the audience to focus on encoding rather than on data, and eventually can lead to bad decisions. Rather than the tool, what is critical for any data expert is to understand how people process visual information, and how data should be encoded in each case, depending on the audience, context, variables, etc.  The ultimate goal of a good visualization is to help the audience convert data into knowledge with maximum efficiency. That depends much more on the skills and experience than on the tool.

Tools evolve quickly, today perhaps there is one slightly better than others, but tomorrow a competitor releases a new version and the situation changes. Or a 3rd player acquires both of them and merges them into a single product. The days when learning one tool meant having a good job for a long time are gone, that is why it is so important to focus on the principles instead. 

With that said, my favorite data visualization tool is D3, which is more a visualization framework (requires some basic knowledge of standard web development tools such as HTML, CSS and Javascript) than an out-of-the-box tool. I teach D3 and I love to see how students get super excited once they start to get familiar with the framework and understand its power. D3 allows you to create your own visualization systems instead of just using standard charts. This is especially important for exploratory visualization (in contrast with descriptive visualization where standard paradigms are usually enough), where each problem has its own particularities and context; in those cases creating specific visualization paradigms could lead to better insights. 

What is your favorite data visualization project that you have been a part of thus far?

I have spent -together with a fantastic team- more than a decade at Next Limit creating a visualization technology called Maxwell Render, which is used for visualizing 3D prototypes (houses, vehicles, etc.) before they are actually fabricated. We are reusing part of that technology to bring virtual reality to data visualization, and while there is still a long way to go, things are looking great so far. That is definitely my favorite project, even if I cannot share too much yet.  

Thank you Juan!

[https://www.sas.com/en_sg/insights/big-data/data-visualization.html]