📊 How to (Not) Lie with Data: Insights You Need 🔍

Ethan Decker is the President of Applied Brand Science, a company that shows marketing leaders how to use the laws of marketing to grow their businesses. With a background in sociology and urban ecology, Ethan combines research with experience to debunk myths in branding and advertising, offering clear, actionable guidance for brand growth.

How Do You Make a Chart That Doesn’t Lie?

How do you present data — a graph, a map, a visual — that’s not just useful but is also honest?

If you’ve ever had to present data, you know you can tell a lot of different stories with the same source material. You can lose people’s attention. You can confuse them. Or you can lead them astray.

For instance, here are four different ways to show the US Presidential election. This data is from 2016, when Hillary Clinton was up against Donald Trump.

These four charts are from the same source data: the actual votes for the two candidates in each precinct. However, they differ in two key ways:

1. How the data is aggregated:

   • By precinct?

   • By county?

   • By state?

   • By electoral college votes?

   • By winner or by vote margin?

2. What story the maps depict:

   • A close race?

   • A red sweep?

   • A divide between cities and rural areas?

The most common way to show a US election map is the top left (called a “choropleth”), where each county (or state) is filled by the color of the winning party. I think this is a crappy chart, and I’ll show why below.

3 Tips for Presenting Data

Here are three tips for presenting data that I wish I knew back when I was starting out in business.

1. Work Backwards from “So What”

When you’re building your presentation, don’t start with the question, then the methods, then the recruit, then the exhaustive data. Start with the “so what”:

• What do you strongly recommend based on the data?

• What are the key implications?

• What should they DO with this info?

Make charts that show that perspective most clearly. Then build the rest of the report around that.

Put the recruit, methods, and detailed graphs or tables in the appendix. This will make your data presentation eminently USEFUL.

2. Show the True Magnitude

When presenting data, a lot of dishonesty enters via how you choose to show the sizes of the numbers — and especially the comparisons.

• It’s easy to make a tiny difference look huge by zooming in on a small sliver of the data or truncating the scale of the Y-axis.

• It’s easy to fudge comparisons by choosing which days, quarters, or years to show in the charts.

Instead, practice good “chart hygiene”:

• Show the scales honestly.

• Include the important data that provides context for the numbers.

• Design your charts so that the size and heft of the data is obvious at a glance.

3. Design for the Eye, Not the Brain

If a busy colleague or client only has time for a quick look at the chart, they should still be able to grasp what it’s saying.

This takes some skill. And even for the veteran presenter, it takes time to play around with different ways of showing the info until you find the one that works best.

• Line graphs vs bar charts.

• Colors vs grayscale.

• Bubbles vs fills vs conditional formatting.

The “right” way to design the chart is the one that focuses the eye on the vital comparisons and doesn’t bog it down with muddy visuals or lots of excess data.

Delete, delete, delete!

And while I’m not a fan of excess “decoration” — using cutesy icons for the bars in a bar chart, or snazzy 3-D formatting with drop shadows — I’m a fan of trying to make even a chart look nice, or lovely, or even beautiful.

Back to the Election

So how does this show up in election maps? First, as I mentioned above, the common view — the choropleth — is a crappy choice.

The Popular Vote

Switching to 2024, Kamala Harris received 74.3 million votes (48.3%), and Donald Trump got 76.8 million (49.9%). That’s a difference of 1.8 points.

The choropleth is terrible in two ways:

1. It’s WAY off of the 1.8% margin of victory — to the EYE.

   • The choropleth is 72% red and 28% blue, not 49.9% vs. 48.3%.

   • It massively messes up the magnitude.

2. It gives a poor sense of how communities voted.

   • It way over-represents people in sparsely-populated places (like the large states in the West that are up to 85% unpopulated).

   • It way under-represents huge cities like New York, LA, Dallas, and Chicago, with millions of people smashed into a pixel or two of red or blue.

Again, this is how the data appear to the EYE, not the brain.

A Better Chart for the Popular Vote

A much better chart for the popular vote is the margin of victory per county, like this one from the NY Times.

This chart does three things beautifully:

1. It does a much better job of showing the people.

   • For example, in Colorado (a state of 6 million people), half the population lives in the Denver metro. The rest of the state is pretty empty (it’s why we like it out here), and 43% of it is public lands. As some have said, “land doesn’t vote.”

2. It does a great job of showing the blend of voters even within a county.

   • Trump landslides are deep red; Harris strongholds are dark blue. The rest are pale and closer to the true 50/50 data.

3. It’s a story-at-a-glance.

   • You instantly get a sense of cities vs towns, West vs East, coasts vs inland.

   • You instantly feel how close the popular vote was.

The Electoral College

Each state gets a certain number of “Electors” to choose the president, roughly in proportion to the state’s population — from Wyoming’s 3 (for 600,000 people) to California’s 54 (for 39 million people). All but two states are “winner takes all.”

Of the 538 Electors, Trump won 312 and Harris won 226. That’s 58% to 42%, or a 16-point margin.

A Better Chart for the Electoral College

A much better chart for the Electoral College story is something I call the gumdrop map. I’m particularly fond of this one from PBS.

This chart does three things swimmingly:

1. It shows the data.

   • The actual 538 data points. Because why not?

2. It doesn’t massively distort the magnitude of each of those Electors by filling space.

   • In one glance you get a sense that it’s not a total blow-out, but that Red probably won.

3. It’s pretty.

Final Thoughts

I don’t even know why the choropleth is still so common. Maybe because, at least state-by-state, it’s just easy. Maybe people don’t know there are better options. Maybe it’s just legacy.

Regardless, I think it’s inaccurate, it’s not useful, and it’s not pretty.

I wish I had some sway in how news sites reported the election. And I wish I knew these three tips back when I started in the work world.

So do your best to make charts that are useful, accurate, and beautiful.