Today I had the opportunity to attend Tableau’s “Design tricks for great dashboards” webinar. The speaker was Andy Cotgreave, a visualization expert and Tableau veteran. During the webinar, Andy touched on ‘framing’. As a level set, framing is about the context in which data is presented, which is critical because author bias can creep in to lead the witness, err, dashboard viewer into arriving at a biased opinion. This has been a concern of mine for years and it was great hearing Andy describe the problem so effectively.

Although the view of the data is static and there should be only one version of the truth, we are all unique individuals with different opinions, backgrounds, etc, thus, we don’t all interpret the data in the same way. I’ve always stated that if you showed one report, without context in a management meeting, each member of the audience would interpret the report differently. Essentially, a report can be twisted to fit a lot of different narratives. During the webinar, Andy used these two visuals to explain this scenario.

The first is a well-known infographic (created by Simon Scarr) that was very impactful. This striking visual depicts the loss of life as a result of the military engagement in Iraq. Certain design choices were made, like the deliberate choice of colour (red) and using the visual metaphor of dripping blood, to convey a very polarizing view.

But what if some simple changes were made to the infographic? I’m not talking wholesale changes to the layout and charts, I’m simply talking about tweaking the colour, orientation, and headline. As you can see below, taking the EXACT same infographic, rotating it 180 degrees, swapping out the red for a blue, and modifying the headline totally changes the narrative to something more positive.

At first glance of the original infographic, my visceral opinion was negative, and I had thoughts about how destructive and senseless war can be. When viewing the modified infographic, my first impression was “hey look, fewer people are dying”. Definitely a more positive narrative than the original infographic invoked.

It is our role, as the data literate, to ensure that when building visualizations our personal bias and opinions don’t influence the interpretation of the results. Easier said than done though. When tasked with creating new visualizations, I focus on the questions that the audience is looking to answer. Once the questions are understood, the focus shifts to providing the facts required to answer those questions. My preference is to not inject headlines or commentary into the visualizations.

  • But without the commentary aren’t the visualizations open for interpretation, thus propagating the ‘multi-version of the truth’ scenario?

Definitely, especially when the audience is non data literate and doesn’t have the experience in interpreting analytics. So how do we bridge the gap to resolve this? To me, the best way to solve this problem is by focusing on finding correlation, and to a certain extent, causation (this is a slippery slope to injecting personal opinion though, so beware) and adding that as context to support the analytics. When the data to support the decision-making process resides across different data sources, or BI platforms, there is an opportunity to tell a larger, more complete, data story. When commentary is placed into each visualization, legibility may be impacted when bringing together multiple artifacts into the data story. Not only that, an opportunity is lost to establish correlations that transcend single visualizations and/or platforms. An effective data story should contain:

  • The facts required to answer the audience’s questions
  • The necessary visualizations to convey the data story
  • Non-biased commentary that guides the audience to correlations in the data
  • When using multiple analytics solutions to tell the story, the emphasis should be on the data within the visualization and not technology that created the visualization

By sticking to the facts, a greater emphasis will be placed on the raw data and the correlation versus forcing the audience into one potentially polarizing view or opinion.