Analytically Amazing, Visually Stunning

By Ross Perez January 31, 2012

After some engaging conversations with our (very passionate!) customers on the Roadshow, I want to address a topic that frequently came up.

Can visualizations be visually stunning like a classic infographic, but remain steadfastly true to the principles of analytic visualization laid down so long ago by Minard, and codified by Tufte and Few? I could not be more adamant in believing this is possible, and I have chosen a viz to prove it with!

First of all, I want to note that this discussion is not in any way limited to a couple of discussions I have had with customers. It has also been an incredible lightening rod on the web, most notably on Stephen Few's blog. It seems the discussion is just as heated as any political debate... possibly more!

There are two main "religions" (and many small sects within them). Infographic supporters seem to generally believe the visual effect and beauty is the most important aspect of a visualization and will lead to increased interaction, exposure and usefulness for the viz. Practitioners of visual best practices (Tufte, Few, et al), seem to generally believe that the visual effect is only a side effect of an analytically useful visualization, and should not be sought on its own. Essentially, the analytical benefit of a viz is its reason for being and therefore its authors only goal is to empower analysis.

As a veteran Tableau Public user and devoted fan of Stephen Few, one might think I am firmly in the "best practices" camp. To be honest, I do lean that direction, but I also firmly believe in the power of a beautiful visualization and I enjoy reading David McCandless's blog as much as critiques of McCandless in our customer Andy Kriebel's blog. In short, I believe one cannot reach their potential with a publicly facing viz unless it is both analytically useful and visually stunning. No emphasis on either one. A viz needs to be beautiful to draw people in and engage them, but it also needs to be analytically valuable in order to serve a higher purpose than just beauty.

Let me show you a viz I believe proves this is possible and desirable. It was created by Adrian Abarca, who is a business and workforce data analyst and Tableau customer. However, like many of us he also has a passion for visualization and moonlights as an infographic/visualization creator. Take a look at his viz of provincial and municipal budgets in Spain below.

To ground this viz a little bit, Presupuesto is Spanish for budget, so the viz at the top left shows the percent of total budget for each autonomous community (cognates are helping me here I admit). Immediately it is clear that there is an enormous amount of the total budget going to Pais Vasco in the North. The highlight table on the right shows the same data, but for individual municipalities. On the bottom, you can see the initial forecast total revenue, total initial credits, and also total costs by division, which can be filtered by the other sheets.

I know at this point most people will be saying "But it's in Spanish!!". Indeed it is. That is actually a little bit of a shot over the bow towards the practitioners and believers in the visual best practices camp (of which I am one), who I think may sometimes resign beauty to the back seat. This viz may be in Spanish but even if you cannot read it, you are intrigued by it and desperately want to learn what it has to say. I went out and Google Translated this whole viz because its visual beauty drew me in and made me want to understand it. My point: analysis is incredibly important but if no one is drawn to the viz in the first place, who cares? People share and spread knowledge when they are engaged by it.

Of course, without the analysis there is no point for it to exist in the first place. Here are some reasons why this is an analytically useful viz:

  1. There are easy to read and well defined user-instructions to help with the interactivity
  2. The sums of the most important statistics are called out above the main visualization
  3. The outliers are clear in all the different sheets
  4. The highlight table is easy to read and more compact than a bar would have been
  5. The tooltips on the highlight table. Hover over anything for a clean and beautiful description
  6. The ? in the top right hand corner describes the viz in detail without cluttering the viz

There are things I would change as well, like the color of the title bars, making the titles a little bit larger, and grounding the color legend with numerical ramp endpoints. I might also want to realign the bottom viz horizontally to make the descriptions easier to read. As we all know, we can pick apart any viz because perfection is unattainable.

However, I think this viz proves the point that anyone with an interest can create an analytically useful and visually stunning viz. In fact, I believe they must in order to fully utilize the ability visualizations have to communicate with others. I am always a little disappointed when a viz fails to engage my emotional brain (beauty) and my analytic brain (viz best practices) at the same time... why should I have to go without either one? Do you agree? Need more evidence to make up your mind? Let me know what you think below.

Addendum: Perhaps this viz isn't your favorite? Check out the following analytically useful and visually stunning vizzes that corroborate the points above in a different form:

  1. What each job gets paid (Guardian Datablog)
  2. North Atlantic Hurricanes (NewScientist)
  3. Tale of 100 (Wall Street Journal)
  4. Oil Exports, Imports and Consumption (Data Driven Consulting)

Comments

Submitted by TJP (not verified) on

If the visual representation is misleading, it is not useful no matter how engaging it might be. Visualization is subject to the same criteria as writing. If someone wrote a misleading but "stunning" article, it would still be misleading and do a disservice to the audience no matter that the "beauty" reached a wider audience.

The issue is larger than visualization and goes to the need for better numeracy and statistical think in the the larger population if the larger population is expected to use data to make informed discussions.

Submitted by Joe M. on

I read your question:
"Can visualizations be visually stunning like a classic infographic, but remain steadfastly true to the principles of analytic visualization?"
and your answer:
"A viz needs to be beautiful to draw people in and engage them, but it also needs to be analytically valuable in order to serve a higher purpose than just beauty."

Combined with the rest of your post, I would like to ensure that I clearly understand your stance, are you saying:
1. Visual best practices do not lead to a compelling visualization
2. That it is okay to visually misrepresent data in regards to human perception to tell an emotional story?

If so, I take issue, and disagree with these statements.

I am very passionate about the method of visual analysis Tableau's software enables, and I make the choice daily to use Tableau because it enables this cycle of visual analysis. The video at http://www.tableausoftware.com/videos/zen is a wonderful example, and why I love to use your software.

I am still very much a student of the science of visual analytics.

In my opinion and daily work:
1. Using visual best practices enables accurate visual comparisons in regards to human perception and communicates information.
2. Making comparisons accurately and quickly is my goal.
3. Embellishing for the sake of making it more attractive or using chart types that are visually deceptive is not productive towards the goal of enabling comparisons.

We would do well to apply Steve Job's statement on design, "It's not just what it looks like and feels like. Design is how it works" to the practice of visual analytics.

Submitted by Ross P. on

Hi Joe,

I am definitely not saying either of these:
1. Visual best practices do not lead to a compelling visualization
2. That it is okay to visually misrepresent data in regards to human perception to tell an emotional story?

Let me try to clarify!

I think we agree about best practices- they cannot ever be forsaken for beauty. We cannot give up on them or attempt to put them to the side. However, I also believe that you can respect all visual best practices and still be left with a viz that can be improved. Viz best practices might get the beauty of a viz 90% there, but one can continue to enhance and perfect it to achieve the full 100%. That is what I meant when I said:

"I am always a little disappointed when a viz fails to engage my emotional brain (beauty) and my analytic brain (viz best practices) at the same time... why should I have to go without either one?"

You brought up Steve Jobs and Apple. I agree with his statement fully, but I also believe it applies backwards: "It's not just how it works. Design is also about what it looks and feels like". Imagine if they had created the iPod, made it perfectly useful and functional, and then did nothing else and stopped improving from there. My first MP3 player was very useful and functional (maybe even slightly more intuitive than an iPod Shuffle), but now I cannot even remember who made it or what I ended up doing with it. I certainly did not buy another. It had no beauty, form, or attractive qualities besides its ability to play music (which it did fantastically). The iPod has amazing beauty and excellent function without marginalizing either. I have only had two iPods, but I have definitely been back to the Mac store on account of their reverence for both form and function.

That is the heart of my statement: you can have both and sacrifice neither. Until you have both a functional AND beautiful viz it is not complete.

I hope that clears up my thoughts! Thanks for the comment.

Submitted by Ross P. on

Hi TJP,

Totally agree- I am not advocating misrepresentation! I want everyone to adhere to viz best practices and also continue to think about and enhance the beauty of their visualization (without sacrificing the function and accuracy of the viz).

Submitted by Stephen Few (not verified) on

Ross,

Why must a data visualization be beautiful? Is beauty necessary for a visualization to fully inform? If so, how does beauty make this happen? To be informed by a speaker, is it necessary that the speaker be physically beautiful? If not, why do visualizations need to be beautiful but speakers do not?

Is it perhaps true that beauty can at times, when properly implemented, support the goals of a data visualization, but that it is only sometimes useful and can also sometimes be harmful?

Submitted by Dimitri B. on

I would like to borrow Stephen's speaker example to offer my view.
I don't think the speaker needs to be physically beautiful to engage the audience.
But let me share an example from my uni days - I had two (very) different lecturers, both alternating in running the same naturally boring engineering course. Both were PHDs and experts, both were informing, i.e. saying things that we, students, needed to know. As far as I can judge (both lecturers and I are male), they were much for muchness in terms of looks.
One delivered lectures in a monotonous, droning, voice. He delivered accurate and complete information on the subject. Most students were asleep or doing their own things on his lectures.
The other was a terrific speaker, like most presenters on Ted Talk. He had a pleasant, projecting voice. He entertained us with lively examples from real life, snippets of history relating to the topic, occasional engineering jokes, etc. No-one was sleeping, your couldn't hear a a whisper on his lectures.
In both cases we were informed, but guess which "informing" we liked more? I can assure you that, although no-one understood why we needed that esoteric and purely academic (in our case) subject, the other lecturer made it interesting enough for us to learn something.

Maybe a data viz doesn't need to look good to be informative. If it is my job to to understand it, if I am being paid for it - I won't mind if it is not beautiful. But if I am reading a science book for pleasure, I would like data viz in the book to look good.
To offer a more extreme example, I can be informed of what is happening by a sketch drawing of the "Last Supper" painting - all the characters are there in the right pose, but I will also enjoy the view if I see the real thing.
There are times when beauty is not important, and there are times when the experience seizes to make sense without it.

Submitted by Adrian Abarca (not verified) on

I believe this visualization is aesthetically pleasing , legible and conveys the message effectively. Its beauty does not distract on the contrary it guided you towards the areas which require attention accomplishing the intended objective. As Ross has pointed out, Beauty is relevant in data visualizations but should not overshadow the intended purpose.

Submitted by Guest (not verified) on

Ross got punked - Plain & simple !!

Submitted by Forrest A. on

When statistical data is "dumbed down" to make a point or draw a person in, it is inherently inaccurate because it has taken the data into a direction that "leads the witness".

The "holy grail" is to represent data in such a way that it "draws people" to interact with it but allows them to drill down into that data to formulate a broader understanding of the data that drives the model, how it was gathered, the statistical significance of variances, blah blah blah.

I hate to use this analogy because it may prove a broader point, but the usability of Wikopedia is what brings people into the fold, add content and make the user learn and broaden their "statistical / content think".

Submitted by Stephen Few (not verified) on

Dimitri B,

The one professor that you described managed to get through to you effectively because he was a good communicator. Your example perfectly illustrates the point that I’m making. What matters is that we craft data visualizations to communicate effectively. On those occasions when visual beauty is required for a visualization to communicate, then beauty should be incorporated to the degree that’s useful. Visual beauty in and of itself is not the goal. It is important, however, that visualizations exhibit a basic aesthetic that is pleasing to the eye, because an ugly visualization will not invite the viewer’s interest.

Adrian,

Your visualization is aesthetically pleasing in a way that supports the data effectively. It follows the basic aesthetic design principles that are required to make a visualization pleasing to the eye. This is important. What’s not important nor even desirable in most cases is an attempt to present the data in a “visually stunning” way (e.g., what McCandless attempts to do). In my mind there is a big difference between aesthetically pleasing, which I support, and visually stunning, which is rarely useful.

Submitted by Ross P. on

Hi Stephen,

Thanks for responding! Let me explain my point from another direction, because I think perhaps you believe I am disagreeing with the heart of your work, and I am actually a huge admirer/believer in the principles you work to promote. Indeed we all are at Tableau!

So... would you agree with this statement?

"A visualization can be beautiful and effective at the same time, but if it sacrifices its informational integrity (or analytic purity) than its beauty is destructive and the visualization is not useful."

From your other comment it seems we both agree that Adrian's visualization is a good (not perfect... what is?) example of a visualization that is both beautiful and analytically informative. That is why I chose it to discuss these, because I feel it represents the argument that beauty and excellent analysis can go hand in hand and do not have to be diametrically opposed. Also, I will disagree with you that this viz is visually stunning (I think we may have a different definition of that word). Other adjectives I would use in addition to stunning or in replacement of it: (visually) pleasing, beautiful, proportional, engaging, etc.

We can all agree McCandless does not adhere to viz best practices and often goes out of his way to not use them. Yikes! I enjoy his work from a visual perspective only... not an analytic one. Actually I enjoy reading his blog in another way, because I will end up asking myself how I would go about the visualizations he makes differently.

However, I also believe (and I think this is where we differ?) that beauty can enhance the emotional connection viewers have with a viz and encourage them to learn more from it than they would otherwise. If Adrian had not spent the time to think about his fonts, placement, colors, imagery- I believe this viz would be significantly less interesting to many people.

Perhaps my definition of beauty is what you would describe as the proportion, sizing, placement, color and font side of visual analysis?

Would love to hear your thoughts on all of that. Thanks again for responding.

And here I thought I was sitting on the fence...

Submitted by Ross P. on

Dmitri,

Thanks for your response! That is an interesting way to think about it that I had not previously considered.

Submitted by Stephen Few (not verified) on

Hi Ross,

I didn’t actually assume that we disagree—quite the opposite—but I was concerned that your words as written could be misconstrued to justify a perspective that neither of us support. You wrote: “I believe one cannot reach their potential with a publicly facing viz unless it is both analytically useful and visually stunning. No emphasis on either one. A viz needs to be beautiful to draw people in and engage them, but it also needs to be analytically valuable in order to serve a higher purpose than just beauty.” You then went on to show Adrian Abarca’s visualization to prove your point.

You said that making a public-facing visualization “visually stunning like a classic infographic” (as in the work of McCandless, I assume, based on your later comments) is as important as making it analytically valuable (accurate, informative, meaningful, clear, true, etc.). I think it’s a mistake to give equal weight to qualities that make a data visualization “visually stunning” and those that make it effectively informative, meaningful, and thought-provoking. Some of those in the infographic community have caused people who rely on data visualization in their work to care more about aesthetics than the information contained in the data. This emphasis has caused great harm. It holds us back from achieving the great potential of data visualization. As you know, I care about aesthetics to the degree that they support the usability of data visualization. Beauty for its own sake, however, should not be a goal of data visualization.

Adrian’s visualization does not demonstrate your point—at least not as it was stated. Instead, it illustrates the application of good visual design practices to make a visualization aesthetically pleasing in a way that supports its usefulness, not as a way to make it “visually stunning.”

I’ve learned the hard way to be very careful in choosing my words when writing about a controversial topic like this. Once your words are out there, it’s difficult to take them back or revise them.

Submitted by Ross P. on

Hi Stephen,

Good point- phrasing is indeed important. To be honest, I think my phrasing was fine but I would not have used "classic infographic" as an example of "visually stunning" if I had it to do over again (so one phrase). I guess I was trying to be a little antagonistic and start a discussion... but I probably should have reconsidered that example regardless. I am glad we are all talking about this though.

Despite that example, I stand by my statement that a great (public facing) viz should be visually stunning (which to me means beautiful, proportional and engaging) and analytically useful. No emphasis on either. Neither being marginalized. This means, as you have pointed out, absolutely not marginalizing the visual best practices we both believe in. Beauty does not trump analysis. But, as far as it does not impede the analysis, I am a believer in beauty and aesthetics.

I think what I have not said is how hard that can be to achieve. Great design is an enormous uphill battle, as you know. Great analysis, equally so.

At the heart of our, disagreement? It seems you do not care for the phrase "visually stunning", because for you it has the connotation of beauty for its own sake. For me, it simply means "beautiful, proportional and engaging". Adrian's viz is all three, and also analytically excellent. Hence, Analytically Amazing, Visually Stunning. I believe it proves my point (first paragraph, last paragraph) very well.

To each his own with the phrasing, I suppose! Thanks again for your comments and thoughtful analysis. I have learned a great deal from your work and, as I said, would never seek to marginalize viz best practices in any of my visualizations. However, I will always try to make them beautiful, proportional and engaging (stunning!) while adhering fully and completely to those principles.

Submitted by Stephen Few (not verified) on

Ross,

If by "visually stunning" and "beautiful" you mean visual design that uses aesthetics to support the task of informing, then we definitely agree. McCandless' work is not an fitting example of this. I think it's important, however, to always use visual aesthetics to support the goal of informing and not as a goal in its own right. Saying that any aspect of design is of equal importance in and of itself to the content is in my opinion an error. One is the goal and the other serves the goal. "Visually stunning" and "analytically useful" are colleagues but not peers.