What do you learn when 110 people discuss data and communications in a knowledge café? Quite a lot. On Tuesday, IHME and Forum One hosted an event on Communicating Data for Impact, centered around the White Paper
we published last year with guidance for getting the right data in the right format to the right audiences. The format made for fascinating discussions and a tremendous learning experience (see key insights and videos below).
We had three presenters that approached the topic from different sides: Deep Dhillon, CTO at Socrata, provided the "Meta Context" from Socrata's perspective of providing a platform for publishing large numbers of datasets. Noah Iliinsky, User Experience Expert at Amazon Web Services, provided specific advice on visualizing data. And yours truly focused on the different kinds of audiences and used specific examples from the Global Burden of Disease study to show how IHME is addressing the needs of different audiences (see the slides here).
The knowledge café format (or rather our version of it) turned out to be very useful for spirited discussions and a learning experience for everyone involved. We started with a plenary session with about 110 participants where co-author Nam-ho Park introduced the Communicating Data paper and the concept of a knowledge cafe. Each presenter then provided a 3-minute cliff hanger for their session. We split into 3 groups, and each presenter provided more detail on their topic (5-10 minutes), followed by an interactive group discussion (10-15 minutes). After each session, the speakers rotated, so every participant had a chance to see all three presenters and discuss each topic. So while I gave the same short presentation three times, the ensuing discussions were remarkably different. At the end, we reconvened the whole group for a final panel discussion and Q&A. All in all, the event ran for a little over 2 hours. A great experience for everyone involved; just check out the Twitter stream.
10 key insights
- Understanding your users' question(s) is crucial to identifying and designing the proper data communication tool.
- Work with different formats to educate your audiences. An infographic may draw attention to your data with very few data points, interactive visualizations can encourage and enable people to explore underlying or contextual data to broaden understanding
- You can categorize your users into 4 key audiences: researchers, data analysts, data actors, casual users. More on that in the White Paper. Any individual can fall into different groups, depending on the data. A Minister of Health is a data actor for health data, but probably a casual user for sports results.
- An interactive data visualization can tell a story or enable users to explore and find stories. To drive change, you may want to use exploratory visualizations to find the relevant stories, then create your own visuals to drive home your points.
- Use all available channels to drive audiences to your data, including social media, media outreach, infographics, policy reports, conference attendence, etc. It is important that the users are pointed to relevant tools that fit the specific needs of their audience group.
- Data visualizations should rest on four pillars: purpose (why?), content (what?), structure (how?), formatting (the icing on the cake). Don't start at the end. More on that on Noah's blog.
- To improve data visualizations and other tools over time, analyze web metrics, conduct focus groups, encourage feedback from users, engage in conversations. Following best practices for visualizations is required but not necessarily sufficient to hit the nerve of the audience and make your visualization go viral.
- When have you done enough analysis to present your data? Depends: In academia, you will always need to get to 99.9%. If 80:20 is enough, you can stop way earlier. In most cases: whenever your deadline is reached.
- Analyzing and communicating data also provides valuable insights into data gaps. As part of communicating data, you should point out what additional data should be collected to provide better answers.
- Quote of the day: stay away from "decision-based data making"
Noah Iliinsky's talk
Deep Dhillon's talk