In this chapter, first I will go into more detail about the differences between dashboards and business analytical applications. I go through the history of these applications. I then go into some detail on the confusion surrounding each, specifically as it relates to data visualization. Next I’ll talk about what makes for a bad visualization.

Understanding the Problem

When I talk about dashboards in this book, I’m specifically referring to the information dashboards used by businesses, nonprofits, and government organizations, among others, to monitor one or more organizational objectives. Dashboards found inside one’s car—or, say, on a panel in the International Space Station—are another type of dashboard altogether. These dashboards are often mechanical in nature and are primarily used to monitor physical systems. There are key differences between the dashboard you might find on your computer and the one you might find in the control panel of a mechanical device. The metaphor, as I’ll explain in what follows, is useful to the extent that both communicate necessary information quickly and efficiently. But the differences between the two are also noteworthy, so let’s discuss them further.

Mechanical dashboards are designed very differently than their informational dashboard counterparts. Consider the airplane—where you would find a mechanical dashboard—in which the pilot is the operator of the vessel. The pilot relies on important information presented in the small area right in front of his or her eyes, called a flight instrument panel. However, the instrument panel isn’t the only source of help and information available to the pilot. The pilot must also rely on other devices and people including what he or she can see outside the windows of the cockpit, the copilot, and communications from control towers, among others. If any one or more of these resources fails—if, say, the flight dashboard blacks out or the communication system malfunctions—the pilot relies on training, experience, and intuition. In this sense, the dashboard then is but one important resource available to help an operator move something—a plane, a car, even an organization—safely to its destination.

The previous distinction is often misunderstood by dashboard vendors and popular visual designers who argue that we should rely solely on devices such as dashboards for decision making. Part of this is hype: as we are able to create applications quickly, we can model and display data better than we ever could before. Vendors have attempted to draw our attention to their work with flashy and sparkly metallic finishes for pie charts. But with that attention, these gimmicky devices do little to inform. In this way, they fail to deliver on their fundamental reason for existing. As you’ll see throughout this book, a primary reason dashboards and business applications fail is because they employ superficial visualizations not based on research.

But another, significant part of the misunderstanding involves the privilege status we as a society give to devices—to anything really—that purport to process and subsequently present information. Think about this way: the original deliveryman of our data was the news, which was an institution that we came to trust (for better or for worse) to vet the veracity of what was presented. In our present day, data comes to us from virtually everywhere, not just the news. The ubiquity of data around us has, in many ways, given us a false sense of security. The decisions we make give the impression that we have a clear view of all the facts, but this is not always the case. In truth, while the sheer volume of information has undoubtedly increased, the channels through which we receive this information are dominated by a few key players.

Think about information in your own daily life. How many different search engines do you consult to find something on the Internet? Probably not many. How many different news companies do you browse to find about what’s going in the world? You probably have a preferred source—and many news services share stories. This book is no different: I prefer Microsoft Excel for spreadsheet development (for good reason), but, for the sake of argument, how many other spreadsheet applications exist to this level of popularity?

We impart a great degree of trust in our information delivery systems, as we should. When they meet our needs, they are invaluable. But with this trust comes great responsibility to do the job and to do it well. What you develop in this book will seek to present information to meet this level of trust. Many organizations and institutions have been all too quick to take advantage of our trustworthiness of their data, encouraging myopia with regard to decision making. They are hoping we won’t cast a critical eye to their work, but unfortunately for them, that’s exactly what you’ll do later in this chapter.

Of Pilots and Metaphors

Let’s return to the airplane example. What pilots use in their cockpits is an evolving set of tools resulting from new technology and continuous research. Analog controls and dials were the norm prior to the mid-1970s (Figure 2-1). Growing demand for flying, however, would soon turn the skies into a congested highway. More information was required for the pilot, and before the advent of advanced digital systems, cockpits were filled with tons of indicators and signals competing for space and attention. In response, NASA studied cockpits in an attempt to develop a new system to encapsulate all the information surrounding pilots and to increase situational awareness of their environment. The result is what’s called a glass cockpit, as shown in Figure 2-2, which is an advanced digital representation of the most important information to pilots. Since its implementation, our skies have become safer with fewer accidents. Glass cockpits are mandated in all commercial flights by the NTSB.

And here’s the point I want to drive (fly?) home: compared to information visualization products on the market, pilot dashboards are drab and ugly. They exist to help pilots make decisions, but they don’t dictate such decisions to them. In fact, a real problem for glass cockpits is they can—and sometimes do—black out. In response, recent studies by the United States’ General Accountability Office have encouraged more training to pilots in the event that this happens. But what’s interesting is that even without the glass cockpit, pilots rely on their other information channels so as not to fly blind. This is a different direction than the visualization industry has taken us.

But what if aircraft had followed this trajectory? What if the view outside the window, the controls, and the copilot were replaced with one dashboard and then that dashboard was blacked out? Who would want to fly that aircraft? Would you? Would you choose to be a passenger on that flight? Would you want to be in the same sky as that aircraft? That flight instruments have followed a path different from information dashboards is because traversing our skies safely is far too important for our pilots to rely on colorful nonsense. In fact, glass cockpits don’t flash or sparkle, but they have a great safety record because they were designed using research.

A Metaphor Too Far: Driving Down the Information Superhighway

At its core, the problem stems from a world of vague definitions. This is not a new problem to technology development.

For a moment, let’s take a trip down memory lane. Growing up, I heard much talk about the information superhighway. I heard this term in the mid-1990s, at a time when I was just a child, when the home personal computer was taking off, modems still made crazy telephone sounds, and the idea that information could be transferred at anything greater than kilobytes per second seemed like a pipe dream. That term information superhighway seemed everywhere—on radio, on television, and in magazines. Even our first computer had the phrase printed on everything from the installation discs to its nascent pamphlets. Yet the definition of this term always seemed different depending on who you asked.

For now, I’ll ask the Oxford English Dictionary:

So, what was it? An abstract concept? A physical structure of a proposed fiber-optics network? The Internet? How could the definition of something so ubiquitous be so hard to pin down?

And here we find the same problem with dashboards. When hype surrounds a product, we hear terms for the product everywhere but care little of its definition. Today, many people are still confused about a dashboard is.

For many, real definitions aren’t required. What mattered then—and what matters now—is the metaphor. For the information superhighway, the idea that rapid information could traverse distance quickly was an exciting prospect. But more than that, in the middle of the 1990s, the superhighway became a national symbol of driving into yet unexplored territory. The information age would soon be upon us, and from our vantage point we needed to drive only a short distance down that superhighway to warm ourselves in the sunlight ahead. We would be able to shine a light on the new information. The sun ahead, we assured ourselves, would surely illuminate hidden truths.

And we may be no farther down that highway than we were a decade ago. Indeed, we may just be driving in circles. Today’s information visualization products are meant to illuminate, but they still carry the bloated metaphor of their mechanical brethren so far as to hinder the process of data sensemaking. They feature radial dials and gauges that dazzle, but research suggests these gadgets inform little.

What’s clear, however, is that if we are to keep our heads above water, how we come to understand and digest data cannot be ambiguous. The information superhighway was more than just a signal of the information to come down its road; indeed, its prolific use in the face of a vague definition is the manifestation of the data storm that now surrounds us with confusion. So too, if today’s information dashboards do not provide us with shelter from the data storm, then they work against us. When they add to our confusion, they are indiscernible from the storm itself.

A Brief History of Dashboards and Information Visualization

The original visual communication products that first appeared in the 1980s were commonly referred to as executive information systems—they represent what we likely today call dashboards. These initial systems were often unreliable and, for many businesses, expensive and impractical. Yet, nearly two decades later, they have become the hallmark of business intelligence packages. Several factors contributed to this rise.

Specifically, initial hardware problems were overcome by the middle of the 1990s when advances in data warehousing made implementing the necessary hardware and software to monitor system metrics more practical and cost efficient. A new way of querying data, online analytic processing (OLAP), played an important role in allowing analysts to quickly pull and report aggregated information from their data stores to executives.

At the same time, a new way to monitor business was being advanced by David Norton and Robert Kaplan, called balanced scorecards. This new way called for the development of metrics called key performance indicators that could monitor and measure the success of a business. This was part of a new way to measure business called total performance management. The synthesis of new technology and new ideas about measuring business carved the desire from executives to monitor more.

Yet it would not be until the Enron scandal—and the executive scandals that followed suit—that the need for computer based metrics took off. In response to Enron, the United States implemented a set of rigid standards to companies concerning their reporting requirements and information assurance standards. The new set of rules required a level of corporate governance that visual communication products, like dashboards, could help achieve.

Finally, the prominence of spreadsheet software in corporations and Internet technologies like Adobe Flash gave developers the means to more easily visualize complex ideas through charts and graphics in a short amount of time. As organizations began to fill their data stores, the need for ad hoc analysis was more necessary than ever. Such cheaper methods brought the idea of visualization to the fore. Organizations such as the United Nations Development Program and General Electric began to emphasize the ways in which data could help the world. Businesses and governments struggled to make sense of what they called big data. Social networking provided a new type of data based entirely around the consumer that organizations sought.

And yet, in many ways, the requirements of these systems—to quickly convey important information and to help analysts quickly divine meeting from data—have changed little if at all since their inception. Indeed, the skepticism that existed in the 1980s surrounding the ability for technology to truly improve business performance still remains in the minds of business executives today. While businesses embrace new technology, they are still hesitant to what they perceive as offshoring business decisions to a computer. Throughout the history of dashboards, those at the top of the company have relied on and blame their information technology for the inability for their technology to deliver. At the same time, business leaders have argued in newspapers and magazines that if the dashboard wasn’t working for a company, the executives needed to take a look at their metrics; it was their fault for measuring the wrong indicators. While companies argued internally over why their technologies and metrics weren’t good enough, visualization expert Stephen Few found fault in work not based in grounded research. In 2005, Few wrote the following:

But not all is doom and gloom. When we take a step back, we can remove ourselves form the swarm of information around us. We can look at key parts of the storm and organize them for data sensemaking. Indeed, this is what information communication helps us do.

A Quick Summary Before Taking a Critical Look

Looking back through everything presented so far, you may have noticed a few key themes. Don’t worry about looking back through everything; I’ll save you the time and trouble and summarize what I feel are the important parts.

• Beware not to take any visualization metaphor too far.
• Just because it’s information presented in a seemingly professional way, don’t trust it outright.
• Technology isn’t the problem, though it often receives the blame.
• Dashboards and business analytics applications help make good decisions, but you shouldn’t rely on them such that your view of other information becomes obscured.
• At the end of the day, it’s all about communication.

Dashboards by Example: U.S. Patent and Trademark Office

In 2010, the U.S. Patent and Trademark Office (USPTO) released several dashboards reflecting its internal operations and requirements. Figure 2-3 shows part of an older version of its patent dashboard, a large colorful series of radial dials and gauges. (When this text was original written, the USPTO was using the older dashboard shown in Figure 2-3. It has since updated the dashboard to follow many of the principles outlined in this book. However, I will use its previous iteration for illustrative purposes.) Take a moment to analyze the following figures. As you go through each one, consider the following: How well does this information communicate? Does this dashboard include extra “ink”—that is, does the dashboard include extra stuff not really required?

What did you notice? The following sections are my thoughts.

Radial Gauges

Gauges are a favorite among dashboard and visualization vendors despite communicating information poorly. Consider Figure 2-4 from the USPTO’s dashboard. What information do you gain from the graphic? Well, you can ascertain that there were 28.1 months of traditional total dependency.

But how did you figure this out? Well, the label in the center essentially told you this information. Putting your finger over the label for a moment, would you have been able to guess the number 28.1 or even 28 months? Not likely. Gauges and radial dials do not allow for precision in visualization.

Now, consider that only the label helped you out; then what is the point of the visualization? What do you gain from the radial gauge metaphor? The developer certainly spent a lot of time ensuring the gauge has a drop shadow, gradiating light source, and cool neon car colors. But none of these additions does anything to convey information. As you will see in Chapter 3, these extra colors and doodads amount to extra ink that services their function little. Information visualization expert Edward Tufte calls this chartjunk.

All the extra ink and junk gauges fill a disproportionate amount of space compared to the information they convey.

Now, consider the series of gauges in Figure 2-5.

Is the information in each metric compared easily across? Not really. They might be better understood presented as bullet charts, which I talk more about in Chapter 3.

So Many Metrics, So Little Working Memory

This dashboard is filled with many, many metrics for you to scan and analyze. What are the most important of these metrics? Does the dashboard let you know which have more importance than the others?

In many ways, this dashboard does not summarize as much as it could. The metrics are grouped together logically, but the long-page format means you must remember what you’ve read at a higher section if you need to compare it to a section below. At first thought, you may not realize there is anything wrong with this.

The problem comes from our own limitations in working memory. Research has shown that our working, or short-term, memory can hold a small amount of information for not very long. In most of our daily routine, our working memory serves us faithfully to remember thoughts when moving from one associated activity to the next and to remember the steps of a daily routine, such as brushing our teeth. Our working memory fills with the steps of what we must do: (1) grab the toothpaste, (2) grab the toothbrush, and (3) go on to shave after this activity.

However, we know from research that our working memory can hold about two or three “chunks” of data at any one time. This limitation affects how we interpret information, such as pie charts, graphs with legends, and radial gauges like those featured earlier.

The gauges display two pieces of information each: (1) the name of the metric presented and (2) the value of the metric presented. But, some of the gauges on the dashboard will display three pieces of information with an added goal metric. Returning to what we know about working memory, these pieces of information fill in the slots available to hold such chunks of data. If you must actively store this information and then scan the rest of the page, it’s likely to be not an easy task without trying to commit this information to your long-term memory. Further, when you find the metric against which you want to compare, the information in this metric will also likely need to fill its information into one of the slots of working memory. If all the slots are taken, your working memory will dump some information to make room.

You may try this for yourself; you may even find success at retaining information for longer than described. Consider, however, that this is the result of an effort on your part to actively retain this information. That is, it’s likely that by simply being aware of the limits of working memory, you are attempting to commit the information to more long-term storage, such as thinking about the information, repeating it yourself, or even writing it down. If you must do this in an information visualization, then consider whether the data presented in a visual manner is as illuminating as its visualization configuration suggests.

There are solutions to the limits of our working memory; one way is to take advantage of our preattentive processes. These processes are activated when we are looking at visual patterns. When we compare values in a bar chart, preattention comes to life. We don’t need to commit a visual shape size to memory; the mere act of placing it next to another similar shape that is taller, shorter, or the same size conveys simple information that does not require us to commit it to working memory in a way that we may lose it before using it. I’ll use Chapter 3 to discuss how best to use preattentive processes in your work.

Is the Logo Necessary?

Often it’s tempting to create a nice logo at the top of our work. I’m not a graphic designer, so I cannot weigh in on the design effectiveness of the logo in Figure 2-6. However, on the function of a dashboard, the logo serves little purpose but to take up space. If the end goal of a dashboard, as described in Chapter 1, is to use a single screen, then the logo serves to consume important and otherwise better spent real estate.

The top of the screen will be the first thing you see when you scan this dashboard, and the top-left corner of the screen is where, as I’ll demonstrate, the most important information should reside. The dashboard’s title, a logo, or its owner is important information, but if you are to feature it prominently, you must ask yourself what data it conveys. As a general rule, if the item of the dashboard does not convey data or information, it should be featured with less, not more, prominence.

Returning to the idea of privileged information, notice the language and images used to describe the dashboard. The logo says “Your window to the USPTO,” yet does this dashboard, as a whole, truly contribute to a better informational understanding into the government bureaucracy or into its goals or progress? I don’t think it does. But the trope of having a clearer view into the process by merely invoking the use of information visualization is readily seen.

While likely unintentional, the pair of binoculars with a view clouded by data seemingly reiterates the suggestion by vendors that all we must see is data. In reality, the poor viewer of the binoculars may be looking at too much data from afar while not realizing what information is available to him at hand. The image is a just another reflection of the myopia that diving into a sea of information can create.

The author Stephen Few lists 13 common pitfalls found on dashboards, which you can apply to your work as well.

• Exceeding the boundaries of a single screen
• Supplying inadequate context for the data
• Displaying excessive detail or precision
• Expressing measures indirectly
• Choosing an inappropriate media of display
• Introducing meaningless variety
• Using poorly designed display media
• Encoding quantitative data inaccurately
• Arranging the data poorly
• Ineffectively highlighting what’s important
• Cluttering the screen with useless decoration
• Misusing or overusing color
• Designing an unappealing visual display

When you consider the definition of a dashboard given in Chapter 1, you see that the USPTO’s dashboard violates some of the original guidelines outlined. For one, you don’t really know what the most important information is for you to consume. Second, while the information can be said to be visualized, the processes used in delivering this information to you aren’t ideal. You’re left wondering if visualizing in this case has really brought a significant benefit to simply presenting the information in, say, a report or another form. Finally, the information was not presented in a single view but rather was spread across an entire horizontal column.

In many ways, the USPTO dashboard may lend itself to commendation for encouraging transparency of their process. However, you must remember the privilege data is given when evaluating the performance of the information visualizations themselves.

Visualizations That Look Cool but Just Don’t Work

The unfortunate reality is that many visualization efforts fail to live up to a high standard in which the organizations that house them promote their effectiveness. On the one hand, like any field, visualization research is growing. Recent interest in the field, however, has encouraged work that is more hype than useful. Consider the dashboard in Figure 2-7, which is inspired by style that is seen commonly in corporate dashboards.

Perhaps the first thing you notice is the part shown in Figure 2-8.

It’s true that the company that made this dashboard presented it as a proof-of-concept and not from a real dashboard. But this graphic is featured prominently in its libraries; indeed, it’s a piece of eye candy meant to draw your attention to its work. In a real business application, it strains the imagination to see how a chart like this might ever be useful. Sure, it might have your attention, but it offers little in the way of its intended use: to communicate information quickly and effectively.

The chart in Figure 2-8 was not born in insolation; rather, it is a product of the hype that surrounds information visualization. Vendors add this type of chart to their chart libraries promising a variety of ways to present information. But no amount of variety will fill a need if it cannot do what it was born to do. But how did this hype begin? For that we turn away from private business and focus on data journalism.

Data Journalism

Data journalism is relatively new. Just as interest in using data for better business decisions has grown in the past few years, data journalism—using data to tell a journalistic story in new and novel ways—has grown in parallel. One notable example is the journalist David McCandless, who is known for his data visualization work in the mass media. Among other things, he is a contributor to the Guardian’s data blog, and his work has been featured in the Museum of Modern Art in New York. You can check out his work at www.davidmccandless.com/.

His work exists more in the colorful world of the chart of Figure 2-8. It’s not that his work is uninteresting—far from it. In presentations on data visualization I’ve done, I pass around his book along with books from other data visualization experts. His book almost always receives the most attention. That’s because his work is colorful and flashy. But if you take a close look at his work at the link provided, you’ll see concepts presented in a way that stress art over communication. In this book, I’ll draw a distinction between this type of data visualization (more appropriately described as data journalism or data art) and the business data visualizations presented here.

And yet it’s important to understand the business world hasn’t yet drawn the distinction. McCandless, for instance, isn’t alone in communicating art over facts. Other outlets, like the Washington Post’s Wonk Blog, feature work that’s overly complicated, communicates little, or is repurposed without enhancements from other outlets. You can see one of my favorite examples of overly complicated data journalism graphics at this article from FlowingData.com, which uses a Wonk Blog chart: https://flowingdata.com/2012/05/18/is-the-filibuster-unconstitutional/. Compare that to the simplified chart in Figure 2-9, which would be considered too boring by a data journalist. Which do you think is easier to read?

But aside from this work being a poor communicator, much of this work suffers from underlying confirmation bias built in. This is not hard to understand; the author’s personal conclusions and opinions compensate for data stripped of its ability to tell a story. You must remember that data art is often created with an agenda already in mind. Infographics, a type of data art, often exist to promote a product or provide stats for an organization convincing you to donate to them. In data art, the author’s narrative shapes the data presented rather than letting the data tell its own story.

As you develop your applications, you should let the data speak for itself. If you’re not careful, you can design applications that allow managers and stakeholders to create results they already agree with. It’s sometimes hard to identify this confirmation bias because, by its nature, the feeling of validation comes with a visceral feeling of pleasurable affirmation. In reality, confirmation bias works against you; rather than illuminating new insights, it provides you with what you already believe to be true. In that sense, it provides you with nothing of new value.

Why These Examples Are Important

The previous examples aren’t trivial visualizations found randomly on the Internet made by nobodies. For example, David McCandless has given talks about information design and data visualization at well-known conferences and to vendors. The point I want to underscore is not that McCandless or any other designer’s work is bad from an artistic standpoint—I readily admit that I am not equipped with the expertise to make this sort of evaluation. I want to note instead how interest in visualization has driven work unsuitable for decisions to be preferred by businesses. In reality, what is praised outside our offices does not and cannot deliver this to us. When you consider the influence McCandless et al. has had on the business side of information visualization, are you at all surprised by this type of chart (see Figure 2-10), whatever it is?

The Last Word

The world of information may at first appear daunting and dizzying. Information surrounds everything we do, and the information channels we’ve come to rely on to be our guides aren’t always using the best methods to communicate with us. However, learning about information visualization and data interpretation can arm us with the ability to understand our world more correctly. You should remember that information visualization is part of the decision-making process, but it shouldn’t replace your own decision-making faculties. You need to rely on data, but you also need to rely on personal experience, intuition, and those around you that you trust. When you prefer data to everything else, you become myopic, but if you understand where data interpretation sits with everything else, a real, true big picture is within your grasp. In the next chapter, I’ll go through the principles of visualization and discuss how you can use them to your advantage.

___________________________

¹Produced by “Arpingstone.” Released into public domain. https://commons.wikimedia.org/wiki/File:Hornet_moth_dh87b_g-adne_arp.jpg.

²Photograph taken by “Naddsy.” Reused in accordance with Creative Common Attribution 2.0 Generic License. https://creativecommons.org/licenses/by/2.0/deed.en.

³Created by Randy Schutt. Reprinted here under Creative Commons Attribution-Share Alike 3.0 Unported license.