Data Viz Things I Want to Try in 2019

So, we're a couple weeks into 2019, and I've been thinking a lot about the kind of work I want to do the rest of the year.

For several months, I've been focused on creating a sustainable, reusable platform and process for all of New America's data visualization work. I'm almost finished—we're about to roll out our data visualization charting and componenent library (I've named it Teddy 😬) to all New America staff next month. Here's an early prototype (I'm still working on a nicer home for it that will be used internally). All of the code is open source on Github.

It was a big lift, both on the design and coding side, and also because it's going to be accompanied by a newly designed workflow and process for the graphics team, which took a lot of thinking, planning, and convincing organizational stakeholders to put in place.

My hope is that these new systems will support almost all of the organization's data viz needs, while allowing the graphics team flexibility to focus on innovative projects where we'd like to invest more time and energy.

And with that added flexibility, here are some of the things I'd like to experiment with in 2019:

Scrollytelling

There's been big move towards scroll-based user interactions in the interactive news space. It makes sense. Let's say you have a series of graphics with multiple different states and layers. Usually, when we want to show these different layers to our audience, we tend to use interactions like clicks, swipes, or hovers, with user interface components like dropdowns, buttons, sliders, and tooltips.

Scrollytelling is powerful, because a "scroll" is much less of a burden for the user than a click. To quote Joshua Porter, “Scrolling is a continuation, clicking is a decision.”

Scrollytelling lets us tightly couple the content on the page to the data that's being displayed. It doesn't disrupt the narrative flow of a story, it augments it. For complex graphics, scrollytelling can free us from the tyranny and ubiquity of the dashboard.

Unfortunately, scrollytelling gets a bad rap, but as long as you follow some basic guidelines, it can be a really powerful way of telling stories on the web. The biggest rule here is not to hijack the browser’s native scroll and reimplement a shittier solution in javascript.

Scrollytelling resources:

Data quizzes

I really like the idea of using a quiz format to bring a reader into a data visualization. The basic idea is that you present the reader with a concept, and ask them to test their assumptions about an issue by drawing the chart themselves. Afterwords, you show them how their assumptions stack up to the actual data, as well as the assumptions other readers have made.

Here are some nice examples:

Localization

I recently watched a talk by Nadja Popovich, a graphics editor on the NYT climate desk, at Tapestry 2018. She talked about how one of her goals is to make her visualizations on climate change personal for her readers.

This was her example: How Much Hotter Is Your Hometown Than When You Were Born?.

It touches on an important point: one of the best ways to personalize a data set for an audience is to localize it in the context of their community, and their lives.

Geography isn't the only way to do this, but it's one of the best. To me, maps are the highest form of data visualization. Other forms can be so abstract that they lose all meaning. I guess it's time to break open QGis again.

I would really love to do more on this topic, and find ways to bring the personalization of a data set set into my work (although I have tried this once, with a graphic on public university affordability).

Measuring Impact

I want to try to figure out how to measure the impact of my data visualization work. My team's mission is to improve audience understanding of the complex policy issues New America works on. So how do we know if we've accomplished this?

I'll tell you a secret—we don't. We really have no way of knowing whether or not something we've made has worked, or moved the needle on an issue, and it really freaks me out. We work across all of our programs to design and develop custom data visualizations, and then once we publish... that's it.

I have a couple ideas, but they're bit underwhelming.

Analytics

Like any modern website, we do track a range of analytics, but nothing specific to our data visualization work. I suppose I could set up some custom analytic events:

  • Amount of time the graphic stays in the user's viewport (although this is vulnerable to people leaving their browser windows open)
  • Interactions—what people are clicking on and hovering over in the graphic

UX testing

If we had more time and resources, we could test our actual prototypes with people from our audience. This is the dream, but sigh, I don't know if we'll ever get there. NPR has released their approach, called Hypothesis-Driven Design, but for a small team like ours, I'm not sure it's feasable.

Internal heuristics

To help us better allocate our time and decide which projects to focus on, I've started tracking hours spent on a project versus the number of pageviews the published graphic has gotten.

For example, if we spent 40 hours designing/developing a graphic, but it only received a couple thousand pageviews, then something isn't working. On the flipside, I can imagine what the response to this metric might be—if the right 5 people saw the graphic, it was worth it.

I need to flesh out my ideas here more fully, but it's definitely something we need to think about more systematically.