In his 1983 book, The Visual Display of Quantitative Information, author Edward Tufte introduced the term, chartjunk. It refers to all the visual elements in charts or tables that aren’t necessary to comprehend the displayed information. As a proponent of minimalism in data visualization design, Tufte argued that most of the “ink” in a graphic should be devoted to displaying the data. In charts, he viewed the ink used to convey non-data elements or redundant information as unnecessary or irrelevant. Outside of the data visualization community, most people would associate chartjunk with clutter.
A flashpoint in the chartjunk debate occurred when Tufte criticized the work of infographic artist Nigel Holmes. While Tufte advocated for a minimalist, statistics-driven approach, Holmes offered a more playful, artistic style that featured visual embellishments that Tufte despised.
On one side you have Tufte saying, “Chartjunk promoters imagine that numbers and details are boring, dull, and tedious, requiring ornament to enliven. Cosmetic decoration, which frequently distorts the data, will never salvage an underlying lack of content. If the numbers are boring, then you’ve got the wrong numbers. Credibility vanishes in clouds of chartjunk; who would trust a chart that looks like a video game?”
On the other side, you have Holmes saying, “Keeping things simple and clear does not mean dumbing down information, nor does it mean making it look boring and austere. That is why art is important. I mean art in the service of information, not art for art's sake. Sometimes art might mean just beautiful simplicity. At other times it might mean wit, or humor, or fun. My fundamental mantra is enjoyable clarity.”
Data storytellers find themselves caught between each side of this debate. Many of us who have read Tufte’s data visualization books have embraced a more minimalistic approach in our chart designs. However, as data storytellers, we are specifically tasked with building explanatory data visualizations for our data stories. In many ways, the responsibility to make insights engaging, persuasive, and memorable can align more with Holmes’s body of work—data journalism versus academic statistics.
In 2011, data visualization author and expert Stephen Few established a more pragmatic, middle-ground position. “In my opinion, nothing that supports the chart’s message in a meaningful way is junk. Sometimes more than minimal ink and sometimes even redundant content is needed to communicate a chart’s message and drive it home.” While ardent Tufte followers will struggle with this view, it reflects a more reasonable perspective on design choices that some denigrate as chartjunk.
When it comes to data storytelling, it’s not uncommon to use various non-data elements to support the messages in your charts. Stephen Few identified three ways in which non-data elements can reinforce the data storyteller’s intended messaging:
When the non-data elements don’t support one of these messaging purposes, they can turn into potential clutter.
For data storytellers, understanding how to avoid clutter can be useful because you want your data charts to communicate clearly and concisely. It’s especially important to avoid forms of clutter that can distort the underlying data or distract from your key messages. Clutter can add unwanted noise that can get in the way of the signals you’re sharing in your data scenes. For this reason, I thought it would be helpful to create a guide to avoiding different types of clutter.
As I was researching what contributes to clutter in data charts, I couldn't find any comprehensive resources on the topic. To assist other people in identifying what issues can lead to clutter in data visualizations, I’ve assembled a collection of potential clutter elements and organized them into five main categories:
As a data storyteller, you want to minimize the cognitive burden placed on your audiences to interpret your charts and removing clutter from them can help with this task. If you reduce this form of noise in your charts, audiences will be able to spot and follow your key takeaways more easily.
Within these five categories, some forms of clutter will be more problematic than others. Even though certain elements may be considered mild or subtle forms of clutter, a combination of them can quickly make a chart more difficult to read and interpret. Without having any definitive studies on what constitutes harmful clutter, I am simply offering an opinion based on my own aesthetic preferences and professional experience.
The frame represents the field in which the graphical objects will be displayed and serves as the backdrop of the chart. If the background is too complex or lacking in contrast, it can interfere with people’s ability to process what’s being displayed in the chart’s foreground. In most data storytelling scenarios, a data visualization won’t require an outer border for its frame that just adds extra weight. However, each chart should have an ample margin or white space around its edges so the chart can be easily distinguished from other content.
Recommendation: Overall, you should try to avoid adding design elements—especially in the background—that will just make the data more difficult to see. For example, adding a detailed image to the background of a chart can make the data series or points in the foreground harder to see and interpret.
The graphical objects in a chart display the data and are the “high” data-ink areas of the chart. However, depending on how you format these elements, you can make the data values easier or harder to process for your audiences. For example, adding a 3D effect to a bar chart can make it more difficult for them to compare values. Other formatting options can make the graphical display more intricate and taxing to work with.
Other design choices for graphical objects relate to color or markers (line charts) can also make the display busier and harder to process. Because there’s no information encoded by them, each adds another variable to the equation and further complicates what the audience needs to process.
In some cases, a high number of data series or volume of data points in a chart can cause interpretation issues. When the data values overlap (i.e., overplotting), they can obscure interesting trends or patterns in the data set—the data then becomes the noise unto itself. You may need to prioritize which data is essential and remove the extraneous information. For the remaining data, you’ll need to find the right balance of the spacing between data values, so they don’t feel too overcrowded or too sparse.
Recommendation: There’s rarely a compelling need for significant styling or formatting on the graphical objects. In general, the simpler the design, the easier it will be for people to interpret the numbers. With data storytelling, you often don’t need to display all the data you analyzed in the exploratory phase. You should try to simplify the data in your charts whenever it makes sense to do so.
The axes provide a horizontal (x-axis) or vertical (y-axis) reference line for measuring values based on a defined scale or assigning categories. Without labeling the data values directly, the axes may be the only means for interpreting them. While axes can be an integral part of most charts, you need to be mindful that the scales aren’t overly granular, which can add unwanted noise.
In some situations, gridlines may be helpful as reference lines embedded within the frame. However, they should be subtle (not thick, dark lines), so they don’t detract or interfere with the graphical objects. Tick marks are another formatting option on the axes that calls more attention to the intervals of the scale. In many cases, they aren’t needed and can be another minor source of clutter.
Recommendation: To avoid creating clutter with axes, you want to make sure they’re properly calibrated with the granularity of the data that’s being displayed. If gridlines, axis lines, or tick marks aren’t needed, remove them.
The text elements of the chart are crucial to interpreting the data. However, labels can quickly add noise to a data visualization if they’re not employed strategically. In data storytelling scenarios, not all the values need to be labeled—in fact, it’s strongly recommended to be selective with what values you focus on with your audience.
It’s also important to make your text as succinct as possible whether the copy is for data labels or overlaid annotations. For example, if you were listing different US states as categories in a chart, it may be preferable to use each state’s two-letter acronym rather than their full names if the audience is familiar with them (Pennsylvania vs. PA). The length of the copy in the annotations will depend on whether you’re presenting the charts synchronously or asynchronously. If you’re explaining the chart directly to the audience (synchronously), the annotations don’t need to be as detailed.
When you display numerical values in data or axis labels, it can be helpful to simplify the values by rounding the values, removing decimals, and using abbreviations. For example, it’s much easier to process $4.1M rather than $4,138,842.25, especially when there are multiple values displayed and the exact values aren’t essential to the point you’re making.
Sometimes, you need to help the audience connect labels to different parts of the chart. Leader lines between the labels and the data points can help users to pinpoint specific values. However, if you overuse them, they can make the chart messy and confusing.
By default, many charts include legends, which act as lookup keys for connecting categorical references to different parts of a chart. Rather than increasing the cognitive load for your users by having them look back and forth between the legend and data series, it can be better to direct label the categories on the graphical object.
Finally, the font design in terms of type (serif vs. san serif), size, weight, and color (adequate contrast) can make a difference in how legible the content is and whether it becomes distracting or not.
Recommendation: In data storytelling, you don’t need to label everything—even though it may be the default approach used by your data visualization tool. You want to be selective with what you highlight and draw attention to. You also want to be mindful that the content should be scannable and easily understood by your audience. The more you can simplify the labels, the more time your audience can spend on processing the message.
In explanatory scenarios, you overlay objects such as arrows, brackets, boxes, or annotation callouts on to data visualizations to draw attention to key data points (sometimes, they are inserted into the frame in the background). You need to determine how much space should be allocated to these objects and be careful that they don’t obstruct any data that is needed for interpretation or context.
Typically, the overlay objects are designed to be noticed, but they shouldn’t interfere with the focal points or key features of a chart. They are meant to reinforce and support the message within the chart whether the delivery approach is synchronous or asynchronous.
To make charts more engaging and add emphasis to the content, you can add icons and pictures to a chart. Tufte would view these more artistic embellishments as unnecessary. However, as Few pointed out, they can enhance the memorability and engagement. At the end of the day, if it isn’t clear to the audience why the images are included or why they’re relevant, they are in danger of becoming extraneous decorations that just add clutter.
Recommendation: In data storytelling, you need to guide the audience to the parts of the chart that are meaningful. The overlay objects can play a key role in this area, but they should still be subtle and not overpowering. If you feel the need to add imagery or other graphical elements to a chart, they must be relevant so they can reinforce the targeted message. You want your chart to be engaging and memorable for the right reasons—and not the wrong ones.
In summary, you want to be mindful of how various forms of clutter can interfere with communicating your key messages. Personally, I’ve found a minimalist approach to be beneficial in my work. I recommend eliminating unnecessary clutter so it takes less effort for audiences to process your charts.
However, there are situations where embellishments and non-data ink can help engage, draw attention, and make your points more memorable. If they serve a purpose and are integral to what your chart is communicating, they shouldn’t be viewed as junk or clutter.
In the example below, you can see how a combination of small edits to a chart design can remove clutter and enhance how it communicates.
As a data storyteller, you need to have a practical and balanced approach when it comes to handling clutter. At times, you may be aligned with Tufte’s minimalist approach, but at other times, you may consider the value of Holmes’s “enjoyable clarity” approach.
In a recent interaction I had with author Alberto Cairo, he shared a few insightful observations about clutter. First, a chart design can be a little messy and cluttered but still be beautiful and readable. Second, while design choices may be subjective, that doesn’t mean they’re arbitrary—everything should be deliberate and done for a reason. Third, what is and isn't clutter is mostly informed by personal taste or judgment, not empirical evidence or universal principles of design.
For me, it really comes down to doing what’s right for your audience and the desired message. The better you know your audience and the purpose behind your chart, the more likely you’ll be able to avoid clutter and convey your insights in a clear, meaningful way that informs and drives action.
Effective Data Storytelling teaches you how to communicate insights that influence decisions, inspire action, and drive change.