We live in a world where data visualisations are done through intricate code and graphic design. From Tableau to Datawrapper and Python and R, numerous possibilities exist for visualising compelling stories. But in the beginning, all data visualisation was done by hand. Visualisation pioneers like W. E. B. Du Bois and Florence Nightingale hand-drew their visualisations because there was simply no other way to make them.
For Du Bois it was his team of black sociologists who explained institutionalised racism to the world using data visualisations, while for Nightingale it was her diagram showing the causes of mortality.
And, even as computers developed, it was often easier to visualise using analogue means. This article will explore the history of hand-drawn visualisations and the case for presenting them in this style. It will also show examples from experts who have opted for the pencil over the screen. You'll also learn some top tips to help get you started.
A short history
When Mary Eleanor Spear wrote her pioneering visualisation book Charting Statistics in 1952, she emphasised graphics that could be easily hand-drawn. For example, the 'range bar chart' (a predecessor of the boxplot) is a simple summary graphic for one numeric variable -- a relatively simple visual to create without a computer.
Unlike a histogram, which would require deciding on breakpoints a priori and counting the number of cases that fall into each bin, a boxplot relies only on a few summary statistics. The analyst would calculate the median, the quartiles, and account for outliers, then get out their ruler and pencil to draw the visualisation.
John Tukey popularised these ideas in his 1977 book Exploratory Data Analysis, where he also emphasised that graphics could be easily hand-drawn.
The idea of Exploratory Data Analysis (now commonly abbreviated EDA) is to compute summary statistics and make basic data visualisations to understand a dataset before moving forward. Every graphic in the EDA book was made by hand by Tukey, although he was so precise they can be mistaken for computer-generated charts.
Jacques Bertin, another visualisation pioneer, was also focused on making a data analyst as effective as possible without a computer. One of his strategies was to create a ‘Bertin matrix’, a physical representation of an entire dataset, which could be reordered by the use of long skewers stuck through it. His graduate students would work to find an ordering that showed structure in the data, then photocopy the physical matrix to retain a version of the data before moving on.
Is digital better? A case for hand-drawing visuals
So, handmade data visualisation is not something new. In fact, it is the original form of visualisation! But, as computer tools have evolved to make it easier to create data visualisations, more and more visuals are ‘born digital’. That doesn't mean the need for handmade visualisation has disappeared, or that computer-generated graphics are better. There are several reasons why I advocate for journalists to experiment with hand-drawing visuals:
It gets you thinking outside the box. If you are someone who is adept at using computer tools to generate visualisations, you may only think of the visual forms most easily generated by your tool.
A handmade visualisation can lend a feeling of friendliness to a story. Quite often, computer-generated visualisations feel sterile and can be inaccessible to certain audiences.
Handmade visualisations feel less ‘truthy’, so they can be a great way to convey uncertainty.
Making visualisations by hand is a concrete way to learn the way that data values are coupled to visual marks.
Sometimes a handmade visualisation is a product you make for yourself, to help you brainstorm, understand your data, or just as a creative outlet. Other times, a handmade data visualisation can become your final product, published for others to read and experience.
And there are many handmade visualisations to draw inspiration from. The book Infographics Designers Sketchbooks is filled with behind-the-scenes looks at how visualisations began their lives. While some of the authors do their sketching in code, the vast majority begin by drawing on paper. So, hand-drawn visualisation can also be a step on the way to something computer-generated.
Perhaps more interesting are the hand-drawn visualisations that end up getting published in one way or another. In the category of personal visualisation, the project Dear Data, by Giorgia Lupi and Stefanie Posavec is a prime example.
The reports were works of art, like an autobiography in data visualisation.
Lupi and Posavec are both professional designers, and their client work (typically computer-generated) can be seen in a variety of contexts.
For Dear Data, they took another approach. Every week for a year, they each collected data on an agreed-upon topic about their lives (like laughter, doors, or complaints) and generated a hand-drawn visualisation of that data on a postcard. They mailed the postcards transatlantically to one another.
While data visualisations often aim to accurately convey information to a reader, that wasn't the goal for Lupi and Posavec.
Instead, they wanted to convey some sense of their lives to one another. Readers aren't asked to decipher the precise values they put on the page, but rather to draw inspiration from beautiful forms, and enjoy what is closer to a narrative or memoir of the authors’ lives.
People who bought the Feltron Report weren't doing so in order to learn something new about the world, but to appreciate Felton's work. Again, the reports were works of art, like an autobiography in data visualisation.
What the experts say
Research on data visualisation often focuses on how effective a visualisation is at conveying the precise information it encodes. In 1984, William Cleveland and Robert McGill published a paper called Graphical perception: Theory, experimentation, and application to the development of graphical methods.
This paper (cited more than 1,600 times!) outlined the results of their experiments on graphical perception. If you have heard arguments for the use of bar charts instead of pie charts, the data likely came from this 1984 study.
Their study showed how bad people are at judging areas (of circles or other shapes) and cautioned against the use of area as a method for graphical encoding. IEEE Vis, a professional community and conference for computer scientists studying visualisation, continues to publish papers along these lines.
For example, the paper Ranking Visualizations of Correlation Using Weber's Law, demonstrated which data visualisations made it easiest for readers to assess correlation visually.
However, the goal of visualisation does not always have to be to encode information in such a way that it is easy to read off exact values. Often, the most important thing is to give a truthful impression of the data. And, the most technically correct visualisations may not always be the best way to convey that impression.
Often, the most important thing is to give a truthful impression of the data.
One important component of visualisation is attention -- a person can't read and understand a visual unless they pay attention to it. Visualisation critic Edward Tufte often advocates for the simplest possible visualisation, by maximising the data-ink ratio.
Darkhorse Analytics produced a gif example of what this process could look like. In many cases, it is better to reduce the amount of visual clutter and non-data ink, but other times it seems Tufte takes this too far, such as his redesign of the boxplot that ends up as a broken line with a dot in the centre.
Data visualisation expert Nathan Yau advocates for what he calls ‘whizbang’. Whizbang is the cool factor (often animation or interaction) that draws people into your visualisations. In a world filled with digitally-generated visualisations, a handmade visualisation might be just the whizbang you need to draw readers in.
Data journalist Mona Chalabi has embraced this idea, creating many hand-drawn visualisations that are published as finished pieces in The Guardian and elsewhere.
Chalabi is the data editor at The Guardian, so she knows the ‘rules’ of data visualisation. But she also understands when it makes sense to bend or break them. Her OpenVisConf talk, Informing without Alienating, discusses her philosophy of making graphics that inform as many people as possible.
Chalabi considers the context in which her visualisations will be seen. She also draws her visualisations using familiar objects, to help readers understand things like units. For example, she created a visualisation to answer the question ’How much pee is a lot of pee?’ using common soda bottle sizes:
In another piece, she showed sugar consumption over time in the US and the UK, using sugar sprinkles.
Beyond the familiarity granted by the objects Chalabi draws, the hand-drawn nature of her visualisations make them feel less precise. Again, this is her intention. Numbers and computer-generated visualisations often ‘feel’ true, but there is always some amount of uncertainty that surrounds them.
By drawing her visualisations, Chalabi is able to convey some amount of variability. When you look at her visualisation of how much air pollution is emitted and inhaled by people of different races, you won't be able to read off the exact numbers. (In fact, you can't read off numbers at all -- the chart does not have labeled axes!) Instead, you will be able to see which group's share is largest, and by how much.
Chalabi is working from her gut, but researchers at Bucknell University have begun to study how different groups interpret data visualisations differently. So far they have focused on a particular rural population, but you can imagine how this work could be extended to other subgroups. One of their key findings is that visualisations are personal.
Often, we imagine that we can generate an idealised representation of a particular dataset, but everyone comes to our visualisations with their own identity and prior beliefs. For certain groups, a visualisation may work well, and for others not at all.
Bucknell’s researchers point out that many of the field’s historic studies on perception relied upon a homogeneous group of people (often, college students, who tend to be whiter, richer, and, of course, more educated than the general population).
Hand-drawn visualisation in practice: an exercise to get started
Hopefully, by now I've convinced you there are many benefits to drawing your visualisations by hand. But how do you actually begin doing it? My main recommendation is to just put pencil to paper and start. Many people believe they can't draw, but almost everyone drew as a child, and simply got out of the habit. As with other skills, the more you practice the better you get.
You don't need any particular materials to draw visualisations. A ballpoint pen and a piece of paper will do in a pinch, but having a few colours makes things more fun. When I do the following exercise with students, I encourage them to bring along any materials they have lying around (I'd guess you have some fun art supplies tucked into a corner of your desk somewhere) and I also show up with some more things to share.
Again, these supplies don't have to be expensive. I bought mine at a local surplus store and spent less than $50 on a bag full to bursting that I've used for several sessions already. Some supplies to consider:
- coloured construction paper
Beyond these standard school supplies, I've brought along balloons, pom-poms, watercolour paints, wire, string, and other ephemera. I find that having lots of stuff to play with opens up creativity. If you want to get fancy, the supplies suggested by Sketch-a-Day are a good start.
Sketching, ideation, and iteration are key elements to design thinking, another area from which I draw inspiration. Artists and designers regularly generate many sketches before settling on a final idea to flesh out. If you begin by making (say) four sketches of a visualisation before picking a particular form, you may find that your later ideas are much stronger than your first one.
Again, working by hand makes it easier to quickly generate a few ideas. If you want to stretch yourself, try making 10 sketches before you begin, or take inspiration from Nathan Yau and generate 25 different possibilities.
Do this quickly! Spend 30 seconds to a minute on each sketch, so you don't get too bound to a particular idea. Remember, you can come back to the one you love after the sketching period ends. If you run out of ideas, it's okay to draw subtle variations on an idea you want to explore, but aim for sketches that are as different as possible. Again, this is a good way to exercise your creativity.
If you are working with others, consider holding a ‘design charette’. Explained by Kara Pernice, from the Nielsen Norman Group, as "a short, collaborative meeting during which members of a team quickly collaborate and sketch designs to explore and share a broad diversity of design ideas".
You don't need any particular materials to draw visualisations.
After you have quickly sketched a few ideas, spend a minute assessing them. Which one do you want to explore further? Now you can begin work on your ‘finished’ product (which may just be for you or could be aimed at future readers). Consider whether you want to use tools like a ruler or a compass to make your work more precise, or if you want to leave it fluid.
Think about your use of colour. Without the computer, you don't have any pre-defined colour palettes to choose from, so you may need to be more intentional. On the other hand, maybe you only have five coloured markers, so those will be your colour choices.
Try this exercise first with a very small dataset. I recommend fewer than 10 rows, but at least two variables. Constraints breed creativity, so having to focus on just a few values will make you stretch more for your set of ideas.
Also, if you need to make a mark for each observation in a dataset, remember that the larger the dataset, the longer it takes! Some handmade visualisations capture many values beautifully, but that can require a more advanced skill.
Take inspiration from Tukey and use tracing paper to trace a computer-generated graphic.
When I have done this exercise with college students, I’ve used that college’s racial demographic data, published by the Office of Institutional Research. This dataset leads to questions of categorisation (do they let students select multiple races, or do they simply select ‘multiracial’? Why are international students categorised separately?), which made their way into my students' work.
At SRCCON 2018 in Minneapolis, I brought data about the city’s lakes. If you are looking for small datasets, Wikipedia can be a good resource. As you get more experienced visualising by hand, you may want to try larger datasets. That's fine! In fact, I think using a combination of computer tools and art supplies is an interesting marriage. You can use the computer to summarise the data for you, but then do the mark-making on paper.
If you want to create a precise visualisation, but also give it a handmade feel, take inspiration from Tukey and use tracing paper to trace a computer-generated graphic. This can also be useful when you want to draw something which might be beyond your artistic skills (say, an outline of the United States).
Throughout this piece, I have been using the terms hand-drawn and handmade interchangeably. It is likely that you want a two-dimensional product of your visualisation if you will be sharing it electronically, but if you have the flexibility for something three-dimensional, try adding in more of the craft supplies I mentioned above.
At SRCCON, participants played with the dimensionality of Minneapolis' lakes, creating 3D visualisations out of paper and wire.
No matter how you decide to work, the practice of hand-visualising can bring you closer to your data. In Numbers in the Newsroom, Sarah Cohen suggests you should "memorize some common numbers on your beat”.
What better way to familiarise yourself with those numbers than physically mapping them on paper? This is also a great pedagogical practice, because when students begin visualising data they can get caught up in the technical details of a computer tool, without grasping the underlying connection between data and visual. Removing the computer makes it more concrete.
Of course, I'm not the only one who thinks this. Stefanie Posavec, of Dear Data leads workshops on handmade visualisations, and the companion book Observe, Collect, Draw! is a guide to help you do just that.
I have also drawn inspiration from Jose Duarte, whose Handmade dataviz kit provides a good starter pack of materials to jumpstart your creation process. Duarte's work is again physical and playful, and he encourages everyone to try it for themselves.
Whatever you do, I believe bringing handmade data visualisation into your practice can lead to more whimsy and fun in your life. Even if your hand-drawn visualisations never make it to print, they can help you think creatively about what you produce digitally. And, in the right context, a handmade visualisation could make your data more accessible to a broad audience.