Beyond getting all the charts right, bringing them together for the intended audience represents a challenge and is a large part of the learning curve for practitioners. As you saw in Chapter 2, the bento box provides a powerful paradigm for starting to lay out pieces and bringing them together. It provides constraints, helping us think about placement and tasks. When you consider widgets and legends to be views of data, most charts come under the bento box paradigm as they support the primary views. In building visualizations, we consider the task the chart represents and work with our end users to understand their world. Sometimes, it's easy. The charts fall into place, their purposes clear, and they work well together.

In 2019, a Vantage Partners survey found that over 77 percent of data initiatives failed to achieve reasonable adoption, per the executives polled. Sixty-five companies from a mix of Fortune 1000 companies participated. Data is supposed to bring forth insight when used, visualized, and deployed at scale. So why did these initiatives fail? It wasn't the technology, the executives said, but a breakdown around people and processes. Their efforts failed at the last mile, where users interacted with their solutions. Returning to the bathroom, this is back to where we stand with dripping wet hands, and a dispenser that refuses to release its wares. It's the paper towel problem, but with charts.

Paradigms on how to create meaning beyond a single chart vary widely. As we bring charts together, we hope to transport our users, to elevate the experience of fact finding to an endeavor that's successful and near-frictionless. Just as with the bento box, we want form to support the function in a way that delights while informing. Yet often, we struggle to build unity within the entire visualization.

Most data initiatives are hoping to move visualization beyond functionally providing paper towels to building an experience more like our restaurant visit. Beyond charts on a page, modern visualization paradigms push boundaries through interaction, expansion, and connection. These techniques provide us options but add more complexity as we look to interweave sundries of charts together.

Synthesizing data serves as a key function. Far from a niche component of certain jobs, data is a profession in its own right. In the early 2000s, we were awed by the terabytes of data we made. Today, we measure in zettabytes. Beyond creating this data, we have to understand it. A part of understanding lies in the research we've covered so far. The other part directly intersects with our history as a profession. So, how did we get here?

The First Charts

Throughout the long arc of history, humans have sought to make sense of their data visually. Numerous cave paintings often feature animals, but with interesting artifacts beyond what would be expected, such as the use of various symbols within a picture or seemingly systematic groupings and placements. Are these evidence of a hunt well done, documentation of dates, or annotations of star constellations? Research shows a number of these cave paintings may be the latter (Sweatman & Coombs, 2018). Patterns line up with where certain stars would be across a myriad of caves in ways too uncanny to just be portraits.

Beyond paintings, other artifacts shed light on how we visualize numbers. The Ishango bone in Figure 3.1, for example, has puzzled experts since its discovery in 1950. Is it a mathematical tool, a period tracker, or a lunar calendar of other sorts? The marks are grouped, their sizes different with a distinct pattern.

As an early visualization, the bone itself creates spatial limitations. The marks are ticks, etched all around the bone using stone or another tool. Yet their groupings, sizes, and the patterns allude to a story, one for which we still have no definite answers. Like cave paintings, the bone provides us clues but not explicit details. Other items, like clay tablets, have preserved transactions and debts throughout a long span of time. Linguistics and mathematicians alike study these to understand how we've used numbers throughout the span of time to communicate and count but also recognize patterns.

Beyond tracking counts and stars, maps are prominent visualizations that document everything from ocean tides to geological changes to ever-changing land use patterns. The Turin Papyrus Map, dated back to 1150 BC, might be one of the oldest surviving maps. It served as both a topographic and a geologic map, identifying where materials were and routes for construction. It abstracted and systemized the types of stones available, using legends to clarify. The Marshall Islands stick chart shown in Figure 3.2 documents ocean wave patterns and currents using a variety of materials, including shells, sticks, and fiber. Individually created, they abstract information for sailing (Finney, 1998).

Bone, stone, clay, and other natural materials feature prominently in the early visualizations we have preserved. Modern data physicalization continues this tradition, using mediums such as knitting, Play-Doh, and clothing to share representations of data. Other experiments explore the sonification of data, allowing the patterns to be heard. Technology always played a role in how visualization was made and shared. Paper and papyrus served as key innovators in the greater proliferation and preservation of data. As lightweight mediums, they rolled and folded while allowing better preservation and transportation.

Advances in astrology, mathematics, and science leveraged paper as an innovation and abstraction to find patterns. Planetary rotations and motions could be tracked as curves and compared. Math concepts could start as doodles to think about a problem or pattern and evolve into paradigms we recognize today. During the Islamic Golden Age between the eighth and fourteenth centuries, innovations around cryptology, language, and algebra pushed the world in novel ways and have direct ties to the innovation of paper. As writing became cheaper to do and more broadly transportable, chart making further created a secondary documentation means around numbers and concepts, as seen in Figure 3.3. Advances in geometry, algebra, and trigonometry benefited by being able to sketch and test ideas.

Advancements in mathematics furthered how societies tracked populations. Maps could capture growth but also changes in landmarks. Statistics matured from state-level tallies for taxation purposes to broader analysis techniques in the nineteenth century. Probability became more integrated in statistics, as various disciplines started leveraging both probability and analysis to predict future events based on existing patterns.

While data visualization has roots in the statistical presentation of data, it is not a neutral tool. In addition to supporting technical advances, statistics were used to further certain political agendas, such racist ideals fed by eugenics. Several statistical methods have direct ties to the eugenics movement, such as Fisher's T-Test. In addition to causing harm, visualization can help us make change.

Standardizing Visualization

William Playfair is widely recognized for inventing common charts, such as the line, bar, pie, and area charts. In Figure 3.4, you can see his time series chart on trade deficits. Note the annotations nestled in the lines. The first time he clarified the lines of imports and exports explicitly with “line of '' preceding each and repeated what it stands for after the lines cross. The areas between the lines are colored in and labeled. The gridlines go from 0 to 200 in intervals of 10. One hundred is clearly called out as the midpoint with a more prominent line. The title clarifies what is being measured (imports and exports) with what filters (from Denmark and Norway) over a given time period (1700 to 1780). The axis is on the right. The years are marked below with additional annotations clarifying units of measurement.

These charts are a turning point in visualization. They take trends and abstract them into a visual system of communication—an idea we will explore in depth in later chapters. The gridlines on this chart likely serve two purposes: they provide the anchors in which to map the points but also reference points for consumers to understand the message.

Other turning point visualizations include John Snow's cholera map and Florence Nightingale's coxcomb chart, shown in Figure 3.5. In the era of COVID-19, both of these visualizations are particularly striking. In the summer of 1854, a mysterious illness gripped a part of London. Within a week, 10 percent of the people in the area had died from the illness. At the time, most people and doctors assumed it was related to miasma, or bad pockets of air. Snow, already an esteemed medical doctor, began mapping out the data as he had a hunch it was not miasma, but the water. Snow's map and accompanying papers helped shift the tide by encouraging people to boil their water to rid it of exposures to cholera.

Only four years later, Florence Nightingale created her famous coxcomb chart. Included in her 800-page book that she also sent to Queen Victoria, the diagrams clearly highlight the role sanitation plays in avoiding deaths. The coxcomb chart showing the number of soldiers who died using the areas of the circle segments clearly shows the early excess deaths. Better sanitation in later months reduces those losses. Nursing the war-injured and ill with few supplies and resources, Nightingale sought to highlight the burden while also showing where improvements were made. Using data proficiently was critical to Nightingale's success and pivotal to nursing.

By 1900, American sociologist W. E. B. Du Bois brought data to life at the World's Fair in Paris, showcasing data sketches alongside portraits and articles around how Black Americans lived in the United States. The color line he discussed in his work was on vivid display in France. The charts were large, framed, and meant to be turned and pursued. Hoping to use sociology to create change, the data clearly showed the broad-ranging effects of slavery and racism.

Du Bois and his team used a variety of charts to showcase his findings, some of which are shown in Figure 3.6. Light lines sketched between bolder colors helped annotate trends. Photos humanized the data, putting faces and stories juxtaposed with the data.

By 1915, engineer and statistician Willard Cope Brinton published the first textbook on data visualization, Graphic Methods for Presenting Facts (1915). Brinton's work is a departure, as it is focused on making charts accessible to the masses. It discusses examples of graphics, their purposes, and how to re-create them. It covered a wide array of chart types and started cataloguing them in ways that would feel quite familiar. We can credit Brinton for shaping the path toward broad use of charts, regardless of industry.

The Shifting Role of Data Visualization

In A Whole New Mind, Daniel Pink (2006) highlights the drastic shift in our economy since the information age. “We are moving from an economy and a society built on the logical, linear, computer-like capabilities of the Information Age to an economy and society built on the inventive, empathetic, big-picture capabilities of what's rising in its place, the Conceptual Age,” he writes. Within the Conceptual Age, we are inundated with technology, abundance, and a globalized workforce.

Data visualization plays a critical role. As limitations around data storage drop, parsing out meaning grows even more difficult. Moreover, working with smaller, disparate sets of data fails to yield the types of findings companies seek. Data itself has become lower-grained over the years, moving from vague aggregates of groups to incredibly personal and tracked down to small minute actions of an individual. While data science leverages these findings for predictions, visualizations serve as a means to communicate and build trust in these models.

Where visualization has existed as a task in a number of domains, recent years have shifted it to a profession in its own right. Those visualizing data don several titles:

• Analyst
• Data scientist
• Designer
• Developer
• Data journalist
• Informaticist

These titles often intersect with specific tools or programming languages. It is common to see titles like D3 designer or tableau developer listed as defined positions within a company. Many roles require a level of data management and transformation in addition to the ability to present information. Others may specifically focus only on front-end creation, dedicating greater focus to user experience and user interface design. Some roles may specifically only focus on training end users to become more proficient in consuming data visualizations or working with executives to formulate a cultural shift to one that is data-informed.

Organizations like Data & Society and the Data Visualization Society seek to further elevate and professionalize the work. As these groups identify members, follow-up initiatives expand to include understanding the work, building standard practices, and ultimately shaping the discourse and vision for the work. These organizations are starting to build surveys, chat groups, and content. Beyond establishing terminology, these groups must also begin supporting professionals with training opportunities and pushing standards. Emerging practice professions rely on these types of networks not only to build on fundamentals but to formalize by building broadly centered certifications that go beyond use of a tool, establishing codes of ethics, and working to shift legislation to protect practitioners and the broader population.

This paradigm shift within visualization is essential. Historically, those working within visualization made charts from smaller, disparate datasets with limited impact. Most modern datasets, however, capture the human experience at a deeper scope than ever before. With today's data, we can understand what makes us think, believe, and act in unprecedented ways. Beyond just seeing patterns, visualization can be used to further shape how we behave. Our challenges are less about finding information and more about identifying meaning amid all the data we have available.

This paradigm shift in data requires us to think more broadly about the impacts of what we visualize. The COVID-19 pandemic also corresponded with an infodemic. Beyond the abundance of charts in certain geographical areas, political agendas interacted heavily in impacting how the data was interpreted. Skeptics leveraged open data to spin the interpretation against masks and other precautions, creating a subculture that propagated misinformation by way of self-analysis over expert interpretation of the data.

Kate Starbird (2020), sociologist and researcher of crisis informatics, also noted the risks of public sensemaking. As the public engages in “collective sensemaking,” rumors form and proliferate, further skewing and fracturing understanding of an event. With COVID-19, this has meant wide variances around interpreting the effectiveness of various precautions and how to handle them. As Starbird writes, “In the connected era, the problem isn't a lack of information but an overabundance of information and the challenge of figuring out which information we should trust and which information we shouldn't trust.”

We saw this collective sensemaking play out in the proliferation of visualizations created around COVID-19. Tableau Public, a free platform for visual analysis, allows users to quickly and easily explore data. By March 9, 2020, there were over 931 visualizations containing the terms COVID and Coronavirus in total, shown in Figure 3.7. They contained a heavy mix of case data, as well as other explorations. Attempting to isolate down to just case data by using terms like case, death, and infection dropped the number a little, with the record high being 143 different visualizations posted on just March 31, 2020. These numbers are still an undercount and represent the “work only on one platform”, using terms in English.

Additional analysis showed notable declines in sensemaking as governmental policy started informing lockdowns, mask usage, and other practices (see Figure 3.8). By late May of 2020, individually made case tracking dashboards shifted down to a steady cadence with a focus moving to peripheral topics and personal stories.

The COVID-19 infodemic is an important turning point for visualization practitioners. Beyond making charts to support a story, visualization often is the story in its own right and plays a key role in communicating and understanding the world. We find more charts integrated within the news, particularly as more news moves online, where charts can be embedded interactively within the story. The New York Times, the Guardian, Financial Times, and a number of other organizations have dedicated data journalist teams. Additionally, Twitter, Reddit, and other platforms allow citizen analysts to also share and present data. Yet by themselves, charts widen the possibilities of interpretation. Later chapters will explore how and why by exploring data literacy but also how we create abstraction systems.

COVID-19 visualizations also highlighted the divergences within the visualization practitioner community. Some felt it was their duty to get the information out there, while others called for pause and direction. These dialogs will start to inform and normalize values within the profession. Visualization as a domain, intersects with so many industries and verticals. While practices and skills may center on a common core, ethics will have to consider how topics are handled. These events serve as cultural flashpoints.

Maturity within the Profession

So many of us learn to make charts on the job or through statistics courses. Process improvement paradigms like Six Sigma are a powerful way to see charts drive action. Visualization courses exist under the guise of several names, such as data journalism, analytics, informatics, and numerous others. As an evolving profession, academic majors dedicated to data visualization are growing but still limited.

Most practitioners learn on the job and through community ties. Yet growth trajectories often share a common thread and arc. Much as with learning a language, there are clear stages. These paradigms also mirror the broader evolution of the profession itself. We'll introduce them lightly here but revisit the concepts in later chapters.

Pictorial Representation

In these early stages, charts are abstract representations of data. Their encoded meanings are still fuzzy, and the selection of proper charts still feels ambiguous at best. Surveys from organizations like the Data Visualization Society highlight how newer professionals feel the load of attempting to learn the language of charts and balance in aesthetics for effective communication.

At this stage, practitioners are attempting to learn how to gather requirements, explore data, and present it so others can use it. They are attempting to match charts to tasks and nouns by translating requirements, deciphering descriptions and needs to the chart tasks we showed in Chapter 2. Sometimes, user needs for charts are explicitly stated: they want a map, trend, and maybe some pie charts to allow for quick categorical comparison. Most common are requirements that are loosely defined and highly ambiguous. Newer analysts are left wondering if the comparative task is best served by line charts, stacked bars, or even heat maps. They may favor charts that allow a variety of tasks to perceivably be performed, such as grouped bar displays or stacked items. The nuance of how and when to use specific charts is still evolving, often leading to compositions that fail to guide end users enough on tasks.

Beyond selecting charts, tying them together is an intimidating task. Early works may look like Figure 3.9. In this example, the practitioner knows the end user of the report needs to encourage donors to increase donation amounts, frequencies, or ideally both. Additionally, within the build specification, brand colors of red, yellow, teal, and lilac must be used on the company's midnight-gray and black standard. The users want a map so that they can localize their pitch, and they need to understand the composition of how users map to the newly rolled-out campaigns. Lastly, they must be able to access information about the individual donor.

Within this example, we can see an allusion to the bento box, framing to constrain the charts in place but also to differentiate them. The color requirement adds complexity but is quite common in the field. The pie charts start to allude to the power of how the campaigns are organized: we can see that most donors fall in the frequency program and, despite smaller amounts, drive close to half of all donations. The map shows donations, but the colors obscure some of the pattern finding. The table serves as a bridge, sharing information about the last donation date and what campaign a possible donor matches. Tooltips may show additional information. Depending on complexity, more text may be used to guide the user. A large part of the interaction relies on the drop-down filters.

Perceptually, we recognize that this work is hard to navigate. The colors overwhelm and the button at the bottom manages to be both visually distracting and hard to see simultaneously. The colors themselves obscure meaning, particularly in this arrangement. The weights from the charts affect how the colors are seen: the pie charts are large, the table is filled with a wall of color, while the map has smaller, lighter splotches everywhere. Both the table and the pies feel complete, while the map gets lost in the shuffle, feeding a type of reading pattern that feels more like a boomerang.

The charts are literal matches to most of the requests: we can see composition with the pies, we have a map, and there is a table at the donor level. In this state, practitioners rely so much on tools like chart choosers to select the individual charts. Structured frameworks help organize charts by size or importance. It is frustrating for the practitioner attempting to translate loose requirements to abstract charts and equally challenging for consumers attempting to navigate a dashboard like this. Yet a vast number of data visualizations land in this category. It is a key reason that most data initiatives, books, and training courses address this level of maturity. Experienced practitioners can easily pick out opportunities to reduce the signal. Color needs to drop drastically. We should revisit chart selections. We might even consider dropping the literal frame.

The pictorial stage is about embracing the feeling of being overwhelmed. It is understood that visualizing data requires more nuance than initially seen from early exposure points. Communities, training, and mentors serve as powerful guides at this stage. A number of communities have training initiatives both publicly and privately to train on more automatic chart selection as well as reducing visual complexity. Most growth trajectories at this stage move toward minimalism and refining work toward what we know from perceptual research discussed in earlier chapters.

Perceptual Refinement

Beyond basic charts, practitioners must also learn to compose visualizations together elegantly. The perceptual stage focuses on making the literal charts more precise as well as working to de-emphasize the entire piece. Design choices start to consider distractions, reducing visual clutter and centering on the message. Minimalism is espoused as a core value with an emphasis on shifting toward precision as accuracy. This is the most common next step for practitioners.

Minimalism is also a key stage in maturation. It is experimentation at one extreme that helps practitioners distill down to core, shared practices. As with other professions, it focuses on using a smaller set of tools better and more precisely. As a visual medium, the library of possible charts starts wide and quite ambiguously—minimalism reduces this to a core vocabulary that covers a wide swath of most visualization practice. Consider these charts to be like the tables and chairs of restaurants. For dine-in service, you need a place to sit. Bar charts are one of the most basic and precise ways to communicate data. Learning to use and present data well with bars, lines, and scatterplots covers a wide library of common charts used in practice.

If we revisit the donations dashboard at the perceptual stage, we will notice wide and extreme shifts, as shown in Figure 3.10. We have dropped down to a monochromatic view, selecting the color with the greatest contrast against our midnight-gray: the lilac. In addition, we dropped the explicit bento box framing and moved to more figurative ones. The charts are significantly changed. The story of our donor profile is more explicit: we can clearly see the outsized donation amount that named sponsorships provide while also realizing that those in our frequency campaign cover most of our donations with the juxtaposed bar. The scatterplot acts as a bridge to individual donors, allowing us to interact and get far more details about donors within the clearly delineated donor profile while being far more targeted about the choices we have in selecting possible donors. Do we want to prioritize those we haven't heard from in a while yet still maximize possible dollar amounts? The scatterplot makes this decision quite efficient.

Moreover, this dashboard adheres to common practices of removing ink, minimizing color, and mostly following a logical reading pattern. A light line highlights the profile pane as a separate entity, allowing it to be read apart from the left side of the dashboard. It is functional and tidy. The information within the pane is greatly expanded with the path to action far clearer. We want to collect certain donations and these explicit options exist.

Yet the work itself is not the only thing that changes at this stage. It is also at this stage that practitioners start embracing the jargon associated with their work. Charlotte Baker-Shenk and Dennis Cokely (1991) provide a cultural paradigm that gives us powerful insight into how language, sociopolitical dynamics, and identity intersect, as shown in Figure 3.11. Role provides a pivotal context: are you part of IT, marketing, or a distinct visualization group? The politics within the org chart would also affect identity: is visualization centered closer to IT, marketing, or communication, or is it wholly independent? Socialization includes common or valued education paths as well as celebrated heroes. Who does data visualization “right” and what counts as a mistake? The language developed, using shared context, becomes harder to crack from the outside.

Within cultural groups, charts take on even more nuanced naming conventions, such as a column versus a bar chart, with the direction of the axis being the lone differentiator. Naming conventions may also be tool-dependent. Tableau, for example, provides a smart drag-and-drop interface for chart building. Practitioners introduced to visualization by way of GUI tools like Tableau may talk about “columnar bars” or “bars on columns'' rather than “column charts.” The nuances align practitioners to their tool communities and serve as cross-community shaming vessels. These activities serve to acculturate the practitioner to norms associated with their preferred tools, work community, and even job title.

Beyond terms, maligning charts that are commonly not used well reduces the load in selecting charts but also demarcates the move from casually producing charts to becoming fluent in data visualization. It is a separator, a line that clearly defines “us” versus “them” around educated chart use and a clear, strong pendulum shift from the pictorial stage. “Pie charts are bad” in particular serves as a hallmark of and rally cry between the general population and educated chart use.

Emerging practice professions rely on these extreme shifts. They serve to identify in- and out-group members by way of attitude as you saw with the norms of chart practice above, and allow greater formalization of the work. As visualization moves from a task done by anyone to a distinct trade and skill set, taboos and attitudes create explicit boundaries. Charts that are heavily misused or presented incorrectly—out of either misunderstanding around their use or seemingly endless relatability—drop out entirely. The identification against something else pushes members toward the professional group. It becomes a principle to present data as precisely as possible.

Computers have shifted us from printing static visualizations (with considerable effort and cost) to interactive visualizations (at much lower effort and cost). We are rather early in this shift with many practices and skills still mostly in the printing of static visualization work. However, the reduction of effort and cost, thanks to computation, gives us more capacity to focus on the semantic aspects of data—by both authors and their audiences.

These shifts alone aren't enough. Beyond perception, other factors influence how our compositions are read and understood. We see broader patterns and influences as practitioners and researchers look to neighboring disciplines, such as user experience, design, and cognition.

Building Subsystems

The perceptual stage is all about learning to build functional visualizations. It is core to learning how to define what a chart should do and achieving it in an efficient manner. As one end of the spectrum, it is not enough. The end users are begging for more. At this point, practitioners start pushing boundaries and testing theories—are pie charts really that bad? Is there a way to reduce precision but remain true to the message?

Figure 3.12 illustrates this shift. Unlike the diametric relationship from pictographic to perceptual, later stages that we propose in this book seek to preserve the lessons learned while expanding to meet a broader range of users. They build from and wrap around the lessons we have learned while seeking to solve problems in new ways.

Summary

This chapter closes out Part A and our brief look at perception. From this point, we are going to assume that you're comfortable with why certain charts are chosen from a perceptual lens and dig deeper into the semantics within the visualization in our next section. There, we will dive into semantics and further clarify and define semantics. We'll also explore how chart types range between more concrete and abstract representations of the data and what data literacy really encompasses. Where we briefly showed the influence of maps on visualization tasks, we will zoom into the parameters that drive that discussion and how those concepts affect other graphs.