We need to talk about pie charts and their flippant use.
They seem to be the default visualisation style for reports and presentations, but a lot of the time they’re bad at communicating the main message. Sometimes they’re even misleading.
“But pie charts have their place, they’re good at showing part to whole!” I hear you roar.
Hold that thought.
I’m going to illustrate why pie charts are weak, using a chart from a real report. I’ve changed the title and labels, but the numbers are real.
No colour, no pie
Take a look at the chart 1 in grey tones.
Do you immediately understand which is the most common, second and third most common way for the respondents to remember to take their medicine?
A pie chart with similar sized slices isn’t good at showing which part is bigger or smaller. You’re asking your audience to compare the slices at different angles, which takes quite a bit of cognitive effort.
To be able to interpret the pie chart you’d need labels on slices to tell which is second, third and fourth most common way for people to remember to take their medicine.
You’re also relying on colour and the chart key to make sense which slice represents which variable. What if a member of your audience is colour blind?
Around 4.5% of the UK population has some form of colour blindness, with 1 in 12 men being affected compared to 1 in 200 women. Depending on the condition, it may be difficult for a colour blind person to distinguish between similar intensity colours.
Additionally, the pie slices aren’t organised from largest to smallest, so the reader can’t rely on the order of the variables to tell them which is the first, second, third, fourth and fifth most common way to remember to take their medicine.
Now take a look at chart 2.
You can instantly tell which is the most common way for people to remember to take their medicine.
This bar chart doesn’t rely on colour and it doesn’t have to have numbers for each bar, because the chart grid shows the scale and unit of measurement, and the bars are organised from largest to smallest.
Such a beautifully clear and simple chart.
Colour me pietiful
For a moment, let’s ignore the point about not relying on colour. Would the pie chart be better with colour (see chart 3)?
The colour and the key helps, but it’s still slow to read which slice corresponds with which item in the chart key. Also, you still won’t be able to figure out which slice is second, third and fourth biggest. It’s still bad at communicating the variable relationships in the data.
Adding colour to a bar chart makes it look a bit more cheerful (which isn’t to be sniffed at), but it doesn’t depend on it (see chart 4).
It’s as clear as the bar chart in grey tones, but looks a more attractive. It’s a nice touch, but the visualisation doesn’t rely on it. The relationships in the data are very clear with or without colour.
I’ve got 99 problems but a pie ain’t one
At this point you, dear reader, should be convinced that pie charts have some pretty serious weaknesses and therefore are best avoided.
However, let’s ignore the points about making your audience compare pie slice sizes at different angles, not having any logical order to slices and having to rely on colour.
Here’s a pie chart in full colour and with the slice percentage labels (see chart 5).
Now you can read the labels and see which slice is second, third and fourth largest.
Do you notice anything unusual? [Hint: The slices add up to 128.3%.]
Remember the point about pie charts being good at showing part to whole? ‘Parts’ should add up to exactly 100%. Not more and not less.
When the data was collected, the respondents could choose more than one option. This is an example where a pie chart isn’t the right chart type, because the variables in the data are not mutually exclusive.
Many of the respondents have at least two different ways to remember to take their medicine. Maybe someone sets a notification in an electronic calendar, but they also get a reminder from their doctor.
If the variables in your data add up to more than 100%, it’s a sure sign that a pie chart isn’t appropriate.
Let’s take one more look at a bar chart, this time with percentage labels (see chart 6).
You can still understand the main message of the graph as quickly as without colour and without percentage labels, but this time with more accuracy.
Another last goodpie
Hopefully I’ve managed to convince you to steer clear of pie charts.
Before making a chart you should make sure to:
- understand how the data was collected
- know what the variables are
- decide what you want to tell the audience
If you want to have a chat about designing effective content with data, get in touch with me by email k.puustinen[at]kainos.com
Resources to help choose the right chart type
Financial Times — Visual vocabulary poster
Office for National Statistics — Chart type advice