When Up Isn't Up: An Argument for Subtractive Visualization

I want to take you higher. Lift your spirits! You want to climb the ladder to the highest position in your organization. Raise up those hands! Lift me up with you! I really look up to you.

Don't be down in the dumps. You focus too much on the down-side. Don't stoop down to their level. Stop peering down your nose toward me.

Up and Right = More and Good

Yeah, we all know that up is good and down is bad. Clearly. And research has shown this to be true across cultures. Up really means "better" and "more". Studies have shown that, at least in left-to-right language cultures like English, to the right is also "better/more". This has deep cultural significance and goes back centuries, at least.

So when we create data visualizations, it's a good best practice that values that appear higher and to the right on a chart should represent "good", and "better", and "more".

Keep this in mind as you look at the scatter plot below, where you see poverty rates going from low to high numbers from left to right on the x-axis. And on the y-axis, you see property values in certain neighborhoods in Boston–higher means higher value.

So think about this - more poor is to the right. But is "more poor" good? No, it's better to be less poor. So while the logic starts off right–higher numbers to the right–it breaks a standard convention that research supports, which is that better should also be to the right. What to do?

Reverse the axes! But don't just reverse them and have 38 on the left, and 2 on the right, because that's also strange–higher numbers should be on the right. So also make sure to change your labeling so it puts "more-ness" to the right.

In case you can't read those axis labels - they say "less wealthy" and "more wealthy". And now the "better" values are up and to the right, as expected.

More Isn't Always Better

This idea can be expanded in other ways. One of the most important examples is when thinking about data many of us have been obsessed with for the past year: COVID data. This is a related, but slightly different idea, which is that "more" isn't always better, such as more infections, and more deaths. So when we visualize case counts and death counts going up, up, up, we're potentially breaking a neurological convention. I think this could be part of the reason why we get numb to the numbers (interesting word pairing there...), beyond the obvious fact that we can't relate to large numbers of deaths.

So how do you visualize higher numbers and show that it's bad? There are a lot of different ways, and I have had an idea for a project floating in the back of my mind for months: using subtractive visualization to depict this tragedy. Rather than showing the values going up over time, I decided to start a visualization with the final number and simply remove elements over time, to reflect deaths occurring. It's a powerful idea, and I think a more emotionally impactful way of thinking about death–it is about loss and disappearance...it is not additive.

Below is the final "data art" piece, meaning it's not meant to be journalistic. It is 100% factual, based on real data from Johns Hopkins University. And it includes clear political commentary because the data is juxtaposed against tweets from President Trump during the pandemic. The intent is to trigger an emotional response. But for our purposes here, simply pay attention to the map and the dots slowly disappearing over time, based on data, tied to the date and counties where deaths were occurring.

My one request when you watch this is to view it on a large monitor, full-screen, with your audio turned on, and just sit and listen and watch the whole thing. Imagine you're in a dark room (like in a museum multimedia exhibit.) And try to feel the impact, not to intellectualize it.

Remember, "up" and "right" are good and your visuals should honor that. But sometimes "up" and "more" aren't actually good, so try to honor that as well. And consider subtractive visualization when appropriate.

==========

Ask me about virtual or in-person data storytelling and visualization training for your teamhttps://bhv.io/teach-land

Learn more about data storytelling and visualization via my other LinkedIn Learning courses.https://bhv.io/LILBillShander