When Brookings Institute asked me to come give a data viz workshop, I asked them to show me how they’re currently visualizing their data.

They sent me to their blog, where I found this:

Who is our audience, when we’re blogging?

Anybody in the world with an internet connection.

So with that audience in mind, take a closer look at this chart. The y-axis is labeled “standardized coefficient value.” How well do you think that’s gonna go over with a public audience? I have a PhD in research and I can barely translate that into plain language.

And the asterisks above some of the columns?

They correspond to, not just p-values, but the three different levels of p-values.

No one cares.

I mean, out of 100 members of the public, I estimate that only 5 even know what p means. And perhaps 1 cares about the three different levels.

This is too precise for the general public.

And getting overly precise is exactly how we lose people.

Now, if Brookings was going to an economics conference to talk to other wonks about their study, they would definitely need to say things like “standardized coefficient value.” Out of 100 members of that audience, I’d guess that 99% know and care about the three different levels of p.

But if Brookings was speaking to that crowd, they’d probably want to ditch the silly sun graphic. And the carrots (?) growing underground.

Those graphic flourishes are likely to lose them some credibility with that academic, wonky audience.

This data visualization is what happens when we try to make all audiences happy with the same graphic. We actually turn off everyone. You need different visuals for different groups.

Precision is audience-dependent.

And precision can show up in many different places, like:

How many decimal places do you really need to report? Two can seem overly nerdy to the general public but 3 or more might be appropriate for some of your more technical audiences.

Do you bother to mention statistical significance? If it’s important and this is going out to the public, work on wording that’ll be clear.

Are phrases like “standardized coefficient value” necessary? Or is there another way we could analyze this data that’s more concrete and immediately understandable, if we’re reporting to a broad audience?

So when you’re running your graph through the Data Visualization Checklist and you reach this guideline about precision, remember that it isn’t about what precision you want, it’s about the level of precision your audience needs to come with you in your data storytelling.