Where is Data Visualization Innovation?

Lately, the data visualization community has wondered whether the field is getting a bit stale. Are we past the days of experimental visualization work? Why do visualizations like The New York Times’ 512 Paths to the White House from 2012 or Periscopic’s Gun Deaths project from 2013 feel few and far between now when they were widely viewed as groundbreaking visualizations at the time?

Shirley Wu took a finer point on it in a recent set of posts: “Let me come out swinging: our industry’s creativity has plateaued…. the craft of telling data-informed stories on the web has plateaued.” Moritz Stefaner made a similar argument on his blog, saying “we see less experimental data visualizations, data viz showpieces or key visuals, and less presence in exhibitions.”

I have a different perspective and will argue here that there is plenty of exciting innovation happening in the data visualization field—it’s just a different kind of innovation than we’ve seen in the past.

Shirley provided a host of reasons why she and others are arguing the craft has stalled, which I think I can sum up into five main areas:

1. Introduction of D3. Clearly, a free, flexible, and open-source language like D3 revolutionized the way organizations thought about presenting data. For a time, D3 resulted in people making everything interactive—every bar chart needed a hover or tooltip. But after a while, it became clear that those kinds of interactive visuals weren’t bringing in readers, and the return on investment just wasn’t worth it.
2. Economic incentives. Creating intense, interactive visualizations is time-consuming and expensive. We knew this back in 2016 when The New York Times’ Graphic Editor Archie Tse made the case that interactives were expensive and not being used as much as readers increasingly shifted to their mobile devices (Gregor Aisch made a similar case at the 2016 Information+ Conference). More recently, fragmentation in the media environment and the rise of dis-, mis- and mal-information on social media has further changed the incentives for organizations to publish their data. Will Chase reminded us in the recent conversation with Alli Torban and Duncan Geere, that Axios differentiates itself from other media organizations by telling shorter data stories: “this very distinctive, short style, very few words. It’s a lot of bullet points and catchphrase sections that always denote the same thing. The goal is to it minimize the amount of time that you spend reading the article and maximize the amount of information you get.”
3. Changing consumption patterns. Americans are reading fewer books than in the past (12.6 books read on average in 2021, down from 15.3 books in 1990), which has affected our broader consumption habits as well: how we read the news, watch videos, and use social media. Readers today have less time and inclination to read big in-depth pieces like The New York Times’ award-winning Snow Fall scrollytelling story from 2021. And without the community of a single social media platform like Twitter, there aren’t spaces for big projects to be celebrated as breaking ground and charting new paths.
4. Scrollytelling fatigue. In her post, Shirley recalls reviewing the 2022 Information is Beautiful shortlist awards and being “inundated by scrollytelling pieces. It’s not that they weren’t good, but that after a while I couldn’t distinguish one from the next….That was the year I realized I was experiencing scrollytelling fatigue™️.” This sense is likely shared by others in the data visualization field. But despite this fatigue among the field, it’s clear that scrollytelling and scrollytelling-like stories grab attention and engage readers. It’s not that other forms don’t or can’t, but this approach seems to have won out, leading to the lack of other visualization types.
5. Eras Tour (in dataviz). All fields change. They all evolve. They all go through eras. Think about the last 200 years or so of art history. We had Impressionism (1860-1890ish), then Post-Impressionism (1880-1905), then Modern Art (1900-1970), then Contemporary Art (1970-Present). Pick your field, hobby, or sport and it has changed and evolved. Data visualization is no different. (Moritz and I talked about this evolution on a recent episode of the PolicyViz Podcast, which you should check out.) We may miss the era of “Romantic Data Visualization,” but does that mean the modern era is worse? Or has it just changed? Or, most importantly, has it responded to the technological and cultural forces that have shaped it? (Related note: Amanda Makulec and Elijah Meeks have explored what they call the “4^th wave” of data visualization in a recent Nightingale blog post/white paper.)

It feels true that we don’t see a lot of those big experimental projects pushing the boundaries the way we once did. And that’s likely for the reasons described above—they didn’t accomplish the goal that creators set out for them. To engage. To excite. To inform. To help decisionmakers.

If big interactive data visualization pieces with all sorts of shapes and colors and menus were so effective and so valuable, we would still see them. Organizations all over the world would still create them. I may be leaning too much on my economics training here, but incentives are incentives. Trends move and change to respond to those incentives. If organizations found that expensive, expansive data visualization projects didn’t accomplish their goals of bringing in more readers or subscribers, or increasing revenues, then it’s no wonder they don’t get published as frequently as they used to!

Of course, we do still see some of these big projects, for example, like Alving Chang’s recent story about belonging in The Pudding, Kevin Wee’s Tableau dashboard on podcasts, Moritz’s work with the German Federal Foreign Office on climate vulnerability, Tim Meko’s graphic for a Washington Post story on gentrification in Washington, DC, or the Urban Institute’s—yes, I’m plugging my own organization—Upward Mobility Initiative dashboard.

But the broad use case of these bigger (and more expensive) data visualization interactives and stories have changed. The media landscape has changed and fractured, and the underlying financial incentives have changed. Did FiveThirtyEight recently shut down because the work they were doing got worse? Some might argue yes, but the truth is probably closer to the idea that Disney (its parent company) found audience preferences had changed and it had become more efficient and cheaper to use other tools, likely with some AI.

This all seems to be coming to a head in a feeling I’ve seen and heard from several people in the field: that data visualization “innovation” is dead (indeed, Shirley’s post is titled “what killed innovation?”). That the lack of big experimental projects that capture our attention means innovation is no longer really happening in the field.

And it is here—towards the end of this post, no less—that I want to push back. Data visualization innovation isn’t dead, at least not to me.

The reason I don’t believe it to be true is because it depends on how you define the word “innovation.” Is “innovation” only big immersive data visualization projects that we all celebrate, love, and share? Or is “innovation” new ideas, methods, or devices that help us communicate data?

Look at the array of data visualization and design tools we have today that we didn’t have in the pre-2012 era: Tableau Public (2012), RAW Graphs (2013), Plotly (2013), Datawrapper (2012), Flourish (2017), PowerBI (2015), Canva (2012), Figma (2016), CODAP (2015), RStudio (2011), Jupyter Notebooks (2014), and Svelte (2016). Isn’t this innovation? Everything people had to create with lines and lines of code can, in some cases, be done with a click or a drag, shared and embedded in moments.

And look at where we’re headed with the integration of AI into data visualization tools. Microsoft’s Co-Pilot, Tableau’s Einstein Analytics, and Google’s machine learning integration into Looker are just the beginning of ways in which we will use AI in our data visualization processes. Hell, just use ChatGPT to write Python code!

Data science education has also progressed immensely, driven in no small part I believe, because of the popularity and utility of data visualization. The popularity of free or mostly-free data and data visualization tools and platforms like CODAP, Google Sheets, Datawrapper, and DataCamp have vastly accelerated how data science is taught in grades K-12. In the next few months, you’re going to see a new release of the National Learning Progression for K-12 Data Science Education, which (I believe) will suggest a fundamental shift in how data science is taught in the K-12 space. 

And don’t forget about data visualization research. We’re seeing exciting work areas such as AI and data visualization, accessibility and inclusivity, automation, and more. Upcoming conferences for CHI in a few weeks and VIS this fall promise to showcase some of this exciting research. We’re also seeing new and growing opportunities to learn data visualization, both within and beyond the college and university system. The growth of (mostly free) newsletters—many, in my experience, on Substack—and the expansion of online courses are opening doors for a new generation of data scientists and visualization creators.

Finally, I’d be remiss not to mention the incredible advances in areas like data equity, accessibility, sonification, data physicalization, and more.

For those of us who have been working in the data visualization field for a while, it may feel a shame to not see as many expansive, experimental, daring data visualization projects we can all share and celebrate. But there is innovation. There is innovation in tools and processes and collaborative organizations. Maybe they are not all public and maybe they are not all so visual and easy to explain, but they should be celebrated. The data visualization field seems to have entered a new era. The excitement is what we make of it.