We often speak of finding and following our passions. Bret Victor, a talented data visualizer and deep thinker on many topics, believes in finding and following great ideas as guiding principles. I met Bret several years ago and became reacquainted with him and his work again this year. One of his recent projects was the creation of the interactive graphics that accompany Al Gore’s newest book Our Choice. Unlike many interactive infographics, Bret’s are both beautiful and uncompromisingly useful, perfectly suited to the audience and task.

Bret and I share an interest in ideas that can serve as guiding principles. We both approach our work in the service of these principles with a keen sense of responsibility. One of the guiding principles of my work can be expressed as:

Individuals usually make better decisions when they learn to think critically (a.k.a., scientifically) and have broad access to information. These individuals can have a positive effect on others if they learn to present information clearly, accurately, and truthfully. Better thinking, based on good information, paired with effective communication, can produce a better world.

Bret talks about the power of guiding principles in a video-recording that you can watch online. One of his own guiding principles, which he pursues in innovative ways, is the idea that the creative process works best when we have an immediate connection with the object of creation. Immediate feedback while creating works around limitations of our brains, allowing us to see what we would otherwise have to imagine with great difficulty. I recommend that you watch this video. You’ll find Bret’s presentation both brilliant and inspiring. If your interest is piqued, you can learn more about Bret Victor and his work at www.worrydream.com.

Take care,

When Business Intelligence software vendors that are to blame for keeping the information age advancing at a snail’s pace have the chutzpah to give advice about data science (the in-vogue term today for data sensemaking), I find it difficult to remain silent. The latest entry in the “Advice from the Clueless” category is an interview with Timo Elliott of SAP Business Objects, titled “What Is a Data Scientist? SAP’s Timo Elliott Says Leadership,” which appeared in Forbes on February 22, 2012. I’ll warn you now that my comments in this blog post are dripping with disdain. A less acerbic response would lack honesty.

Rather than writing a thorough review of Elliott’s comments, which isn’t warranted, I’ll just feature a few quotes from the interview, followed in each case by a short rejoinder.

Timo Elliott, Senior Director, Strategic Marketing, SAP Business Objects

To begin, let’s put Elliott’s comments into context by looking at his experience:

Elliott performed analytics for Royal Dutch Shell for about a year ending in 1988, when he joined BusinessObjects in Paris as the eighth employee. He has been with the company, now part of SAP, for more than 20 years.

“About a year” of analytics experience 24 years ago? Well in that case, let’s hang on his every word.

We have now entered an era in which technology is no longer the primary bottleneck to extracting meaningful business value from data. The primary bottleneck is actually human leadership, according to Timo Elliott, Senior Director, Strategic Marketing at SAP BusinessObjects. In other words, to expand the impact of data on your business, it is time to balance the focus on the technology of analyzing data with development of leadership in order to make sure that technology is put to good use.

SAP would love for us to believe that technology is no longer a bottleneck. Human leadership is indeed part of the problem, but savvy leaders would realize that technologies are still in fact a big part of the problem that we face. A good leader would advise the organization to stop relying on vendors like SAP when attempting to make sense of data.

Historically, what we now call “data science” has been somewhat limited to Web companies, which have had wide access to their activity logs and have been able to devise excellent products from them. But now, through the release of the latest crop of visualization and business intelligence (BI) products, that kind of capability now exists for all kinds of companies.

This is indeed beginning to happen, no thanks to SAP. Those who are using the term “data science” meaningfully and with integrity are doing so, in part, to distance themselves from the likes of SAP Business Objects, which has so far provided nothing useful but production reporting systems. When marketers like Elliott use terms like “data science,” they’re trying to give the impression that they’re on to something new and better without any real substance to support the claim. This is the same organization that introduced the embarrassingly impoverished Business Objects Explorer as “revolutionary.”

The reason many BI projects ultimately fail is too much focus on technology.

I couldn’t agree more. This is also the reason why so many BI products fail. Focusing exclusively on technology without understanding the needs and abilities of those who will use it—a common pitfall in software development—produces products that only a software engineer could love. This is especially true of products that support thinking, which must interface seamlessly with human perception and cognition. SAP Business Objects knows how to build production reporting systems, but not how to build tools for interacting meaningfully and efficiently with data.

“You have Scotty down in the engine room,” Elliott says. “He’s the guy who understands the technology perfectly, but he’s not the one leading the ship. You need the whole crew. You’ve got Spock, who’s an analytics person; you’ve got Bones, who’s the human relations person who does the emotional side; you’ve got Uhuru [sic] on communications. But the key person is Captain Kirk. Captain Kirk doesn’t know how the fusion generator works. He is a decider. His job is to lead people into whatever the situation is and make those tough decisions.

Much like former President George W. Bush, the “deciders” in many organizations are not in touch with the data. Relying on deciders in leadership (the Captain Kirks of the organization) will only work if they actually do have some idea of how that fusion generator propels the ship.

In some ways, a data scientist is equal parts Captain Kirk and Mary Leakey, the best-known member of the team that discovered and interpreted the early human skeleton “Lucy” in Egypt. The data scientist is part ship’s captain, part anthropologist. The data scientist is aware of the complexity of the systems at hand, but is less a deep technology expert than a comprehensive evaluator of the modalities of data used in an organization.

Okay, I’ll start by admitting that I don’t know what “comprehensive evaluator of the modalities of data” means. Probably nothing, giving the fact that marketers like Elliott don’t have to make sense as long as their words sound impressive. Unfortunately, data scientists—those who understand what’s going on based on evidence derived from data—are rarely given leadership positions. Knowing what’s really going on is seldom given priority in organizations.

“What’s not data science is a business person with a business question going to an IT organization saying, ‘Give me this report,’ and the IT person coming back and saying, ‘Here’s your report,'” Elliott says. “That is not data science. Why? Because it’s not about the interface. That’s the business person basically trying to do their current job in the current way, using a little more data. That can be worthwhile, and I’m not saying IT organizations don’t provide value when they do that, but it’s not data science.

I agree that this scenario doesn’t reflect data science. But this scenario accurately represents the very model that SAP has fostered for years and continues to foster in its products today.

“Ultimately, the reason why a lot of the companies that have people called ‘data scientists’ are successful is not only because of the data scientists and their skills, but also because the people that run those companies are keenly aware how much of a difference data can make to their businesses.”

The suggestion that organizations with people called data scientists, rather than data analysts, business intelligence professionals, or decision support specialists, are more successful than others is pure nonsense. The term data scientist in most organizations is just the latest term that’s being used by people who do precisely the same work as those who use the other titles. Changing what you call these folks doesn’t magically improve their work.

Even as data-science technology is on the upswing—IT spending per head…may actually jump 60 percent in the coming years, according to Gartner—there is a growing realization among the most data-savvy companies that the culture is just as important as the technology.

And if Gartner says it, we know what that means, don’t we? After all, Gartner’s magic quadrant claims that SAP Business Objects is the second most visionary vendor in business intelligence, second only to IBM, neither of which have demonstrated any real vision in their products for years.

“Most of us are just really bad at analyzing information,” Elliot says.

Yep, this is absolutely true. Why? Because most people haven’t learned to think critically, haven’t learned basic analytical skills, and have grown less capable to the degree that they rely on dumb technologies such as SAP Business Objects to do their thinking for them. If SAP wants to provide leadership in data science or whatever you choose to call the work of data sensemaking, they themselves have some learnin’ to do. Until then, perhaps they should remain silent and concentrate on listening.

Take care,

Jeff Heer of Stanford University and Ben Shneiderman of the University of Maryland have co-authored a wonderful new paper titled “Interactive Dynamics for Visual Analysis.” With so much emphasis today on the visualizations themselves, Jeff and Ben are encouraging us to also attend to the interactions with those visualizations that are required for effective analysis. Data exploration and sensemaking (a.k.a., exploratory data analysis) requires constant and fluid movement from one view to the next, rapidly changing how we’re viewing the data to pursue each new question that arises. Interactions are required to alter the view, and it’s important that those interactions be quick and easy, otherwise our minds will be distracted from the flow of analysis.

In this paper, a taxonomy of 12 interactions, organized into three categories, is proposed:

Jeff and Ben are two of the smartest, most articulate, and most productive researchers in the field of information visualization. This paper is well worth your time. Read it and then consider how well the data analysis tools that you currently use support these interactions.

Take care,

Data visualizations can be designed to look beautiful, if you possess the required visual design skills. The question is, “Should data visualizations be beautiful?” For years a battle has raged between infographic designers who emphasize the importance of aesthetics and data visualizers with a more practical bent who focus on the degree and quality of understanding that results. Those in the aesthetics camp argue that if an infographic is not eye-catching, no one will look at it, and that compromises in the quality of communication are justified as a means to capture the reader’s attention. Those in the optimal-understanding camp argue that the reader’s attention is wasted if the visualization does not clearly and accurately tell its story. In truth, most people have joined one camp of the other, not because of deep thinking on the topic, but because of preferences formed by their experience or lack of it. I’ve tried to occupy a middle ground, pointing out that visualizations can be both aesthetically pleasing and fully informative, without compromising either concern, but that this takes a high degree of visual design and communication skill. While the battle rages, however, fundamental questions are being ignored.

Should data visualizations be beautiful?

What qualifies as beautiful?

If you believe as I do that data visualizations, despite secondary variations in purpose, are always meant to inform, then their effectiveness is determined by the degree and quality of understanding that results. Therefore, a data visualization should only be beautiful when beauty can promote understanding in some way without undermining it in another. Is beauty sometimes useful? Certainly. Is beauty always useful? Certainly not.

What’s always required is that a visualization work for the human eyes, which means that it should not be displeasing to the eyes. A few basic principles of visual aesthetics can be followed—good color choices, legible fonts, proper placement and spacing, etc.—to achieve this result. Making a visualization beautiful is rarely required and it is usually not worth the effort unless your audience is huge and the information is really important. In addition, it can often work against the goal of informing. Making a data visualization beautiful in a way that compromises the integrity of the data always works against you. Even when the information is not compromised, however, beauty can work against you by drawing attention to the design of the visualization rather than the information that it seeks to communicate. Think back over your life and ask: “Were the people who influenced and taught me the most all physically beautiful? If they were wrapped in a different physical package, would that have affected their ability to influence me or my ability to listen to them? Did I ignore information that wasn’t delivered by stunningly attractive people?” Beauty is not the goal of visualization and it is usually not required to achieve the goal.

On those occasions when making a data visualization beautiful is truly useful, we must face the fact that beauty is indeed “in the eyes of the beholder.” What qualifies as beautiful for some is not beautiful to others, beyond the basic aesthetics that I referred to earlier that are rooted in visual perception. Most of what we deem beautiful is a product of culture and experience. If you love wine, as I do, you probably no longer prefer the wines that you found pleasing in the beginning. The fruit-bomb California Zinfandel’s that I loved in the past are no longer palatable to me. I now prefer wines that were crafted in the European tradition to produce greater subtlety and depth of character and to pair well with food.

To further illustrate this point, I’ve found that, when arguing the importance of beauty in data visualization, people often illustrate their position using works by infographic designers such as David McCandless. To my eyes, however, even when I ignore the fact that the information has been ravaged, I rarely find his work beautiful. Obviously, some people see his work differently than I do, but that’s the point that I’m making. Beauty is a fleeting target. What qualifies as beauty varies with the tastes of the audience.

If you’re a gifted graphic artist and communicator and have the skill that’s required to craft beautiful data visualizations when they’re needed, that’s wonderful, and I wish you well. Just don’t hinder the advance of data visualization by arguing that it must always be beautiful. Remember that the goal is to enlighten.

Take care,

Boris Evelson of Forrester Research has been singing off-key about data visualization recently and he doesn’t seem to realize that he’s tone deaf on this topic. Have you ever noticed that when people become recognized as experts in a particular field, they sometimes think this magically grants them expertise in other fields as well? Expertise requires study and years of practice, practice, practice. I’m particularly sensitive to this tendency when BI generalists give opinions about data visualization without taking time to understand it. This bothers me because people put their trust in “experts” and make costly decisions based on their opinions. My ire was most recently raised when reading statements by Evelson about data visualization as quoted in an InfoWorld article by Chris Kanaracus.

Evelson and I exchanged strong words back in 2009 when he deigned to list the features of “advanced data visualization” in his blog. His list was nonsense and I said so. Long after the dust settled, Evelson contacted me to ask if I’d be willing to advise him on matters related to data visualization. He should have asked my advice before his interview with Kanaracus.

Here’s the section of the article “Tableau BI visualization tools with user-centric design” (InfoWorld, January 18, 2012) that cites Evelson’s opinion:

Until the in-memory addition, Tableau wasn’t necessarily something a company already invested in a BI platform from SAP or Oracle would need, according to Forrester Research vice president Boris Evelson. “These days all of the other vendors have perfectly fine data visualization capabilities,” he said. “Now they let you do this in-memory, which very often is what the business users want. They don’t want to be restricted to the underlying database structure.”

At the same time, Tableau and its competitors need to further differentiate themselves. Microsoft is pushing PowerPivot as an extension of Excel with not much of a learning curve, while Spotfire features integration with Tibco’s middleware stack and offers advanced analytic capabilities, he said.

However, “whatever [Tableau] is doing, they’re doing it right,” as Forrester client interest in the company has jumped significantly of late, Evelson added.

I can imagine the mixed feelings of Tableau’s leaders when reading Evelson’s words: grateful he said that “they’re doing it right” but cringing to have these words spoken by someone who doesn’t actually understand what they’re doing and what makes it right.

Tableau did not suddenly become relevant to organizations with big BI product stacks when they introduced in-memory data handling. All along, these organizations have needed what good data visualization vendors like Tableau and their kin have been providing—effective ways to explore and analyze data—because the big BI vendors haven’t provided it and still don’t, which brings us to Evelson’s most naïve and potentially harmful statement: “These days all of the other vendors have perfectly fine data visualization capabilities.” After I read this, my wife mistook my convulsions as a seizure. Evelson’s statement couldn’t be more wrong. To date, none of the big BI software companies support data visualization in a manner that is “perfectly fine” or even reasonably adequate. They allow you to view data in graphs, but do so in embarrassingly inadequate ways. This inadequacy is especially apparent when we narrow our focus to exploratory data analysis, which requires meaningful and rapid interaction with data. Neither PowerPivot from Microsoft, Business Objects Explorer, nor any of the other attempts that I’ve seen by big BI vendors to enable exploratory data analysis have advanced past kindergarten. To draw on my Biblical roots for a moment, good visual analysis products such as Tableau, Tibco Spotfire, and SAS JMP lead people who have previously stumbled around in the dark using clunky BI products to exclaim “I was blind but now I see.”

Finally, back to our friend Boris Evelson. The best experts in any field are the people who started out as and continue to be the best students. When we stop being students, our expertise ceases to grow. When people recognize you as an expert and begin to hang on your every word adoringly, it’s tempting to wear that mantle with pride, refusing to ever again assume the role of student. If Evelson wants to express useful opinions about data visualization, he’s got some learnin’ to do. This is true of many BI thought leaders. Until then, they should stick to what they know.

Take care,