I was recently working with an online course student to visualize data for her country’s government officials.

These aren’t her real categories or real numbers, but you get the idea.

Before: Reading Numbers

Her “before” version looked something like this:

She had small n’s — a unique situation!

My definition of a “small” number is less than 100 — an arbitrary cut-off point that I learned from a past supervisor, and I’ve stuck with for years.

Not only was the total under 100, but each of the cells was under 100, too.

After: Skimming Visuals

There are several ways to bring the table to life.

During a 1:1 consultation, we talked through a few ideas together.

Idea 1: Heat Table

This easy visual — made via Excel’s Conditional Formatting — would be a great addition to the appendix.

If you’re familiar with the 30-3-1 approach to reporting, then you know I put allllll the tables in the appendix.

In other words, I aim for zero or few tables within the body of the report. The report’s body should focus on key findings with graphs, not dilute the data with eeeeeeverything in a table.

Idea 2: Clustered Columns

This is Ann K. Emery’s least favorite chart of all time. I’ll never let any of my students use one of these bad graphs!!!!

It’s the default option — we simply highlight the summary table in Excel, and insert a chart — and we get this lazy chart.

There’s too much going on, so it doesn’t tell us anything.

The color-coding by category is off.

There’s a separate legend, which is an accessibility no-no.

Idea 3: Small Multiples Histograms

With a re-orientation and re-coloring, it’s easy to build small multiples histograms.

But, the vertical columns didn’t fit her portrait Word doc:

The the horizontal bars would be totally fine.

My only hesitation was that bar charts are… boring.

I’ll never let my students have only bar charts in their reports, slides, dashboards, or infographics. Our viewers deserve variety.

Plus, this was a unique small n situation. Let’s capitalize on that!

Idea 4: Icon Arrays

To bring the small n to life, we tried an icon array.

The student liked this option because:

• It’s very, very, very fast to create — much faster than creating and formatting a chart.
• It was similar to her original table — the same rows and columns.
• Right away, she could see how much bigger Category A was than the other categories.

Yes, I realize the irony…

I just wrote a blog post asking you to humanize your data with realistic people icons, which would be a step above shapes like circles and squares.

We tried realistic people icons, too.

They look fine close-up.

But they simply didn’t fit. In other words, there would have to be fewer icons, and/or they’d have to be much larger, to even be legible.

They just look like chicken scratch, sigh.

Behind the Scenes

I’ve written and spoken about symbol fonts a million times.

Webdings, StateFace, and WeePeople are personal favorites.

Yep, these are simply Webdings n’s!

Never, ever add a bajillion individual circles to your Word doc or PowerPoint slide. What a waste of our own time. Just use symbol fonts.

Made Within Minutes

Can you type n’s into your table?

And change the font into Webdings?

Even if you’re the slowest typer in the world… I promise it won’t take long to type some n’s.

Then, simply adjust the font colors, so that you’re using one brand color per category.

Bonus: Download the Materials

Want to see the Excel file and Word doc used in the blog post? Download them here.

Something wonderful — and terrible — has started happening.

Researchers, evaluators, and scientists complete my online courses. Or they bring me in for private workshops.

And they start applying what they learned. GREAT!!!

But it’s a little off. OOPS.

This is entirely my fault. I’ve been recommending that we use icons for years now — but I haven’t been clear enough about how to use them, and how not to use them.

In this blog post, you’ll see 8 different types of icon use. The good, bad, and ugly.

(1) Multi-Color 1990s-Era Clip-Art: NOOOOOOOO

Let’s start with what not to do: clip-art.

Icons and clip-art are similar, but not identical.

Icons use one color, so they hold up well in grayscale printing, or when very small.

Clip-art uses several colors, so it doesn’t hold up well in grayscale printing, and it’s hard to see when it’s small.

For these obvious reasons — clip-art is outdated and hard to see — clip-art gets The Big Red X.

(2) Cheesy, Posed Stock Photography: NOOOOOOOO

Gosh, I hope this one’s obvious.

I love adding photographs to reports, slideshows, and infographics.

But not the cheesy, posed type, like this.

Another Big Red X.

(3) One-Color Icons to Replace Words: NOOOOOOOO

One-color icons are okay.

But icons can only accompany words, not replace words.

(4) Icons to Accompany Words: Yes!

Let’s start by adding words: males and females.

Then, we’ll turn those choppy phrases into full sentences to remove guesswork.

We’ll bold a few key words to make it more skimmable.

We’ll color-code the keywords to match the icons (blue keywords for blue icons).

Finally, we’ll color-code by category: blue for one group, and green for another. You’d obviously use your brand colors, not mine, but the technique is the same: one hue per category.

Even if you’re the slowest typer in the world, these bare-minimum edits should only take a minute.

(5) Icons to Categorize: Yes!

Next, let’s add some actual, ya know, graphs.

Bar charts are fine, but they’re super boring if that’s all we have.

These are binary variables — a.k.a. yes/no variables — so two-slice pies are perfectly fine.

(Here’s my list of pie chart guidelines if you’re not familiar with my stance on pies.)

This is where we can add icons to categorize.

You see the difference, right?

I’m not a fan of using icons on their own, like the previous examples. They feel overly-simplistic and dumbed-down. But I am a fan of using icons to accompany other graphs, and to aid with categorization, like this:

We’re not limited to pies for binary variables.

We could use pies, donuts, or square pies.

No matter the graph type, the icon usage is the same: They’re accompanying the graphs, and aid with categorization.

(6) Icons to Show Proportions: Yes!

We can also use icons as the visualization itself.

Collages of icons are called icon arrays. The square version is also known as a waffle chart.

For bonus points, we could add the male/female icons to aid with categorization, too.

You see the distinction, right? The primary icons (the collages of squares or circles) are used to visualize the data, not just decorate it.

(7) Overly-Simplistic Icons to Humanize: NOOOOOOOO

The only thing worse than one dumbed-down Gingerbread Person is a hundred of them.

NO MORE GINGERBREAD PEOPLE!!!

(8) Realistic Icons to Humanize: Yes!

I am a fan of using realistic icons to humanize the data.

You see the differences, right? We’re not using single icons just to decorate, like the first couple examples. We’re using icons to categorize and/or to show proportions.

And we’re not using dumbed-down Gingerbread People. We’re using realistic human silhouettes.

More Examples of Realistic Icons to Humanize

I want to normalize these realistic icons for you. Here are a couple examples from the wild:

• The proportion of female chairs in board or CEO positions in German companies
• COVID-19 won’t spread as easily when Oregon reaches herd immunity

Using Smaller Denominators

All icons — squares, circles, Gingerbread People, and realistic human icons — get tricky to read when they’re tiny.

In some projects, we can simplify our fractions.

I tried changing 42% of men to 42 of 100 men, and then to 21 of 50 men. That worked fine!

But when I tried changing 37% of women to 37 of 100 women, and then to 18.5 of 50 women… yuck. If our unit of analysis is people, we can’t talk about or visualize half a person. That’s confusing to our audiences.

So while I’d prefer to simplify the fraction, I can’t in this specific example.

Traditional and Storytelling Versions

Hopefully you’ve heard me talk about traditional vs. storytelling graphs before. If not, you can watch this video for a quick overview. Both versions are correct, but they’re correct for different types of audiences.

The Bottom Line

There are several different ways to use icons in data visualization, and some are better than others.

I don’t recommend clip-art (version 1) or cheesy stock photos (version 2).

I don’t recommend icons to replace words (version 3).

I don’t even recommend icons on their own (version 4).

Instead, at a bare minimum, we should be using icons to accompany other graphs and to categorize (version 5).

Better yet, I’d prefer that we use icons (like little squares or circles) to show proportions (version 6).

And when our unit of analysis is humans, I’d prefer that we use realistic human icons (version 8) and never the dumbed-down Gingerbread People (version 7).

Your Turn

Which version of icons are you currently using? Which one might you try in the future? Comment below with all your ideas! Bonus points for sharing links to your public-facing projects so we can learn from each other.