Episode 15: 3 Things I Learned From The Economist Data Team - Featured Data Viz by The Economist Data Team

 
 The Economist. Image via their  website .

The Economist. Image via their website.

 

Welcome to episode 15 of Data Viz Today. What can we learn from the top-notch Data Team at the Economist? Host Alli Torban dives into the 3 things that she learned from their team as they put together their predictive model for the U.S. midterm elections. (Hint: the complexities of predictive model building, visualizing uncertainty, and annotations!)

Listen on Apple Podcasts, Google Play, Google PodcastsStitcher, SoundCloud & Spotify.

 "Who’s ahead in the mid-term race" image via their  website .

"Who’s ahead in the mid-term race" image via their website.

  1. Building a predictive model is both really complicated and really exciting.

  2. Conveying uncertainty is crucial both visually and when reporting results.

  3. Annotation and clean formatting go a long way in making visualizations that anyone can understand.

  • 01:30 - Thanks to Heidi Horchler for my new cover art! Check out her Daydream Odyssey Project and her Instagram!

  • 02:14 - Workshops by Jon Schwabish and Stefanie Posavec:  Chicago and DC.

  • 02:25 - Alberto Cairo’s online free data viz course.

  • 03:30 - I describe how they put together the predictive model. Read the full methodology here.

  • 07:53 - Conveying uncertainty is crucial both visually and when reporting results.

  • 08:10 - Try conveying a probability as 2-in-3 rather than 66% to keep your reader from doing any mental rounding.

  • 09:15 - Try adding a shaded confidence interval.

  • 09:55 - Annotation and clean formatting go a long way in making visualizations that anyone can understand.

  • 10:08 - Consider adding an annotation to a histogram that sums up what the graph is saying.

  • 10:50 - Consider showing the same information in two different ways (e.g. spatially and then categorized by color).

  • 11:40 - Try website testing tools to help you test the understandability of your data viz, like the 5-second test on www.useabilityhub.com

  • 12:30 - Alex told me that the design mockups were done using ‘R’ for prototyping and Adobe Illustrator for design. The final visualizations in the story were built using D3 and React so that they can be updated dynamically when they rerun the model each day, and this also allows them to make the visuals responsive for mobile users.

  • 12:50 - Full story, full methodology

  • 13:00 - James’ Quora Q&A

  • 13:15 - My final takeaways are:

    1. Predictive models are super fascinating, but we should also keep in mind that the real world is really complex, which a model can’t fully capture.

    2. Think about the ways that you can show uncertainty - not just visually but also in the text, like how they convey a probability as 2-in-3 rather than an exact percentage.

    3. And lastly, really prioritize adding annotations and clean formatting so that your charts are understandable to anyone.

  • 13:45 - What’s their advice to designers just starting out? Matt McLean said to force yourself to look at your designs from the perspective of a reader who knows nothing about the subject. Overcoming the problem of assumed knowledge is key to creating visualisations that actually show what you think they are conveying. Also, learn to code.

  • 14:15 - If you want to see more, follow the team on twitter at @ECONdailycharts.

  • 14:27 -  Get mapping right away with my free-mini course “Make Your First Custom Map in Under 30 Minutes”.