"Well, It Depends..."

The answer to just about every question about how to do things in data visualization is, "Well, it depends..." Seriously. Ask me a question. Odds are 99-to-1 that's going to be my answer.

Why? Well, it's because data visualization brings so many disciplines together and the goals of data visualization are so varied and the nature of the content is highly interpretable based on point of view, bias, context, and 100 other things, that even the most basic questions depend on context, intent, audience, focal point, the nature of the data, etc.

There is no better example of this than a kerfuffle that occurred over the weekend on Twitter.

First, there was this tweet critiquing maps on the Georgia Department of Public Health COVID-19 status report.

This tweet criticizes these two maps and implies malicious intent on the part of the creators. And if you are quick to judge, and don't view the source, and lack empathy for the creators, and maybe don't even think about the use cases for the tool, you might quickly pile on, criticize, foam at the mouth, etc. As you can see from the number of RTs, that's pretty much what happened.

The professional data visualization community went bananas.

And then Jon Schwabish wrote a thoughtful piece "critiquing the critique" and there was much soul-searching and further consideration. And the pendulum of criticism swung somewhat the other way!

But the conversations around the nuances in all of this is where this gets really interesting. The primary topic - is this a "good" or "bad" map/viz? If so, why? What specifically "works" or "doesn't work"?

Ooooohhhhhhhh, let me say it again! "WELL, IT DEPENDS!!!"

You see, the "it depends" rabbit hole is endlessly deep and wide. Is the scale "fair"? Is the color scheme appropriate and fair? Does the answer change based on whether the use case is or is not intended to show change over time? What if the user is confused, regardless of intent? What if the data is changing dramatically over time and we can't predict maximum values to set a scale that is flexible to handle that change? Is it a great or terrible choice to show an "alert" bucket, and base that on the "top X" or should it be based on some measure of "outliers"? Etc., etc., etc.

I'm not here to report fully on this debate, provide a final judgment, or say much of anything except this conversation is a great living example for why "it depends" really is always the answer. Go read the original tweet, and go down some of the RT and reply rabbit holes. And read Jon's rebuttal, and Kenneth Fields' tweet and replies. And think deeply about how hard it is to make the right call sometimes. Oh, and don't forget to look at the original data dashboard so you take in all of this in the context of the supposed "original sin."

And if you want to leave a comment below, including a question about some of that Twitter thread, please do! I'll be happy to continue the conversation. But you've been warned. You know what the first three words of my answer will be.

==========

Ask me about virtual or in-person data storytelling and visualization training for your team

Learn more about data storytelling and visualization via my other LinkedIn Learning courses.