Pie charts don’t get a lot of love from data viz practitioners, but users prize it for its simplicity and the niche it fills.

As a former data journalist, I instinctively shied away from visualizing my data with pie charts.

Similar concerns spring to mind: pie charts don’t really say a lot, and they’ve got a knack for distorting the truth, even when they’re used in appropriate situations.

But it wasn’t just my data viz senses tingling. In a fast-paced newsroom, I found it easier – and more eye-catching – to simply shout out figures rather than painstakingly optimize a pie chart.

Working on this in Flourish took me at least 3 minutes…
… while this took half a minute, and packs more visual oomph.

Percentages already imply part-of-a-whole, and unless the other slices are relevant to the overarching story, why the need for slices in a circle?

When push comes to shove though, it’s easy to see why many of my colleagues – and other organizations – love to use pie charts. They’re quick(er) to design, and feel familiar to many viewers.

As it stands, the newsroom folks I worked with weren’t the only ones dealing with this impasse. Much has been said about pie charts and their (mis)use over the years, and I figured to take stock of what’s been raised – and what remains contentious – so far.

The case against pie charts

There’re a few key reasons why pie charts should be forbidden.

Mike Raper, who’s Director for Organizational Analytics for Talent Dimensions, shares my grievances: pie charts take up a lot for showing so little, and it’s difficult to interpret due to its shape.

Microsoft Data Scientist Thomas Dorfer stresses the latter: humans are pros at reading lengths and heights, but we struggle to guess values from angles.

Dorfer’s argument that we interpret pie chart values from angles may be incorrect though. A 2016 study conducted by Tableau’s Robert Kosara and Fluence product manager Drew Skau claims that humans read pie charts by area (or in combination with arc length).

But not everyone hates pie charts with a passion. Some data viz writers, like Kai Wong and Steve Young, suggest that pie charts aren’t evil – we’re just using them wrong.

Wong’s take, in particular, is a timely reminder that accuracy (or precision) isn’t the only yardstick that a data viz’s effectivenes is measured by. Given that angles are less accurate visual cues than length, determining a pie chart’s worth solely by how accurate it visualizes data is a fool’s errand.

But his prognosis on where pie charts can shine falls short for me. If all pie charts are good for are to show minority or majority, then I’d stick to my way of shouting out percentages instead.

What about showing parts-of-a-whole then? I’d readily go with Dorfer on this one: tree maps > pie charts. They’re so much more versatile, and can handle multi-level data as well.

Die die must pie (chart)

What if we really, really want to use pie charts? In what situations might they suffice?

Personally, a pie chart’s redeeming quality is in its ability to emphasize something. For all its flaws, a pie chart with a disproportionately large slice instantly grabs attention.

Raper concurs, but adds that it “takes a bit of modification” – detaching the main slice, tweaking color, direct labelling and the like – before one can get the most out of a pie chart.

Storytelling with Data stresses two use cases:

  • to give a general sense of the part-to-whole relationship, or
  • to convey that one segment is relatively small or large.

The first comes with a rejoinder: comparing the precise sizes of each slices is less important than comparison to the whole.

But is using a pie chart really optimal in these scenarios? Tree maps, sankeys/alluvials (using them interchangeably since there’s some debate over whether the two are truly distinct), and sunbursts all pack more punch – heck, even a donut chart feels more useful, with its white space in the middle leaving room for additional information.

Maybe pie charts start to make more sense when time and cost considerations in a newsroom or business come into play. Shouting out percentages is less time-consuming, but readers would likely prefer a visual interlude in between walls of texts.

Still, a pie chart for the sake of a pie chart isn’t a good justification for using one, is it?

Anyhow, there’re still a couple of things to pay attention to if pie charts are still on the cards:

  • Order slices according to how one might read around a circle. A clockwise flow is ideal.
  • Avoid distorting effects (e.g. 3-dimensional, exploding). Not a critique unique to pie charts though.
  • Minimise the number of pie chart slices as much as possible, to avoid visual clutter.
  • If there’re too many categories, consider de-emphasising less important ones by giving them a dull color, or subsuming them into a catch-all “other.”
Source: Storytelling with data

A new critique of pie charts?

So, where else can the pie chart conversation go from here?

For starters, I’ve always wondered whether the use of pie charts inadvertently over-simplifies analysis.

The need to minimize the number of categories is more pressing in pie charts, and that necessarily glosses over distinctions between different entities. The priority of emphasis also means that pie charts spotlight just a sliver of the data that the the person behind the chart feels is most important.

What about dynamic, interactive pie charts? Beyond visual spectacle, do they fare any better than their static counterparts?

Then there’s the business angle too. Time is everything, and if adherence to data visualization principles is too time-consuming, then pie charts likely won’t disappear anytime soon.

Or perhaps the conversation can move towards designing – if not, properly using – simpler alternatives to pie charts. Tree maps, alluvials, sunbursts are superior options, but they feel foreign at first glance. I personally had a tough time clarifying how to properly use alluvials to my colleagues too.

For now, I’d stick to edible – and more palatable – pies.

Leave your musings here

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: