Aaron Williams

"Our audience has become much more accustomed to reading and understanding charts"

Share on facebook
Share on twitter
Share on email

Aaron Williams works at the intersection of journalism and data science as a data reporter for the investigative team at the Washington Post, covering topics ranging from the opioid epidemic and gun control, to race and immigration. As a reporter, Aaron leverages his expertise in data visualization, statistics, and journalism in order to create clear and effective visual stories.

Your definition for “infographics”
Infographics are how the world communicates right here, right now. Infographics typically take data that’s usually structured in a column/row format and shows insights from that information. But, data, in scale, size and form, is changing and becoming something that’s transforming the world as we know it. So, it’s unclear if my definition will hold in the future. For example, we’re seeing more and more infographic-like content produced by vloggers, social media influencers, and others interesting in conveying information visually (You decide on whether or not you think any of it is actually good!).

Which are your obligatory references?I’m inspired by everything from the talented journalists and designers at news organizations around the world to hip hop fashion designers. I think all visual art aims to communicate something, which means there’s inspiration everywhere.

Recently, Observable, an online interactive notebook, has not only become my preferred platform for building graphics but also my go-to repository for fascinating deep dives into data and visualization. You can find a graphic for almost anything there.

Is there any special project that you would like to highlight/stand out?
In 2018, My colleague Armand Emamdjomeh and I did an analysis of the persistence of racial segregation in the United States. To this day, it’s still the most ambitious project I’ve worked on.

To produce this map, we used a mixture of census data, node.js and traditional GIS tools. We’ve since used the metholodogy to create other dot-density maps on other topics in the newsroom.

How is your work process?
I always start with the data.

Before I sketch or code anything, I spend as much time as I can investigating the dataset. I usually do this via the command line or in Microsoft Excel. I get to know the columns and look for quirks in the data. I look for outliers, unusual themes in the data, and really anything that makes me think someone would want to learn about this data.

From there, I’ll usually move the data into something like Jupyterlab or R Studio where I can do a more robust analysis. This where I model the data into a format that I might visual later. At this stage, I might even begin to create simple charts to flesh out any of the findings I found in the first step.

For example, if I noticed some outliers in the data, I’ll chart how far they are from the mean or median and see if that’s interesting. Most of the charts and visuals I create here do not go into the final visualization I create.

I create these visuals to help myself and my collaborators understand the scope of the data and ask others what they think about the analysis. Do they think what I found was interesting? What am I missing?

Asking these questions help me refine my analysis and hopefully lead me to the most interesting part of the data, which in turn leads me into how I might visualize it.


I used RStudio to do a more in-depth analysis of opioid overdose death data after exploring the columns in Visidata.

Next, I’ll grab a pencil and paper and do some rough, freehand sketching. At the Post, I learned to embrace sketching with paper before writing code. Drawing on paper is a cheap way to explore ideas without putting too many resources into creating a visual. I can also easily iterate on ideas before picking a few that merit translating to code.

Scribbles from my notebook when working on this project.

From here, I’ll move to Observable. Observable’s “magic notebooks” have changed how I work because I can quickly have a working JavaScript environment without installing anything but a web browser. And because it’s entirely on the web, I can edit and share my work from any computer with an Internet connection. This has become invaluable because it allows me to send work-in-progress graphics to stakeholders and allow them to tweak parts of the graphic without writing code. This means even someone without knowledge of d3.js or JavaScript can send me feedback on colors, spacing, labeling and more.

Furthermore, the ability to export my rendered graphics as either PNG or SVG files means, if the project doesn’t need to be interactive, I can create the majority of the graphic in code and then move the exported file into another program like Adobe Illustrator polish for publication.

Early versions of opioid overdose graphic published in late 2019. All of these graphics were protoyped in Observable.

Using this approach, my colleague Brittany Renee Mayes created a set of visuals for an analysis we did on the history of the opioid epidemic in the U.S. And because I did this in Observable, you can how it works in my published notebook.

How was your first “immersion” in this fascinating world of visualization?
I started my journalism career while I as college student in San Francisco. I was writing stories about the neighborhood I lived in but also did some web design on the side for extra money. It wasn’t until around my senior year of college that I heard about the Interactive News team at the New York Times (then under Aron Pilhofer). I learned there were journalists who used code and design to tell stories. After attending the New York Times Student Journalism Institute, I began to learn as much as I could about this world.

Do you have any particular goal to face in 2020 interns of visualization?
I joined the Washington Post investigative desk in last year, so, my goal for this year and beyond is to take my visual skills and apply them to investigative reporting. Often, newsrooms will have a distinction between their investigative desks and visuals desks, so, I look forward to bringing a visual approach to all the work I do going forward. I also hope to find smarter ways to work with massive datasets that often can’t be visualized without smart approaches to loading the data. The information produced worldwide is only getting bigger and so I want to be able to analyze and visualize that information as quickly as possible.

What kind of data or information do you dream about visualising in a graphic?
I’ve always wanted to work more with audio in my projects. I think that are interesting ways to visual music beyond the waveform and it’s something I hope to try either as a side project and feature story in the future. 

Graphics departments at newspapers are evolving. Now they work on more complicated, polyhedric assignments, including different tools and narratives. More than a traditional piece of graphics… from this trend? Where our craft is focusing at?
Graphics are fairly commonplace in most news stories, which means our audience has become much more accustomed to reading and understanding charts. I think this allows us as data visualization enthusiasts and wonks to push the boundaries of the medium. I’ve seen more adventurous chart types and interactions that we didn’t see in the past. Internet speeds are still variable depending on where you are in the world but smart phones are fairly commonplace. This hopefully means more and more people are seeing data visualizations in their life and becoming literate in the language of charts (provided we are making sure our work is accessible).

That said, I’ve also seen great restraint.

What I mean by that is there’s a habit in this industry to sometimes create an incredibly sophisticated graphic for the sake of doing so. However, a complicated chart, no matter how innovative the code or design is, is worth it if the audience can’t understand what they are looking at.

In response, I’ve seen more and more graphics that aim to be clear and concise first before being clever. This doesn’t mean the charts are simple or that they aren’t innovative. On the contrary, I’m starting to see graphics that balance minimal visual complexity with well-reported insights into a topic. I imagine we’ll see this more as graphic designers become graphic reporters.

Would you dare to say this is not a graphic or this is a graphic? Could you define a clear line?
Just in the short time I’ve been doing this work, the idea of a “graphic” in news has changed meaning hundreds, if not, thousands of times and will likely continue to do so. I think it’s counterproductive to codify a graphic into a specific form. However, there is such thing as a bad graphic and I think it’s important to learn what works and what doesn’t.

If you weren’t a graphics guy, you’d be a…
I’d probably be DJ at a house party somewhere or coffee shop owner whose cafe doesn’t provide WiFi.

What are we going to find in your speech in Malofiej?
I’ve been thinking a lot about the history of visualization as it pertains to progress and social issues in the U.S. In my talk, I’ll walk through how I think graphics can do more than just visualize data.

What do you think?