Don’t charts pie. Save them for birthdays.

Data contains knowledge that needs to be unveiled.

So, with data visualization (that always depends on what we want to show) we can transform data into insights, and we can communicate effectively but only if we are able to detect the information it carries within.

In every representation, we must create charts that clarify and provide the right canvas for analysis.

Obviously, there is not one single right option that will be the best one in every situation: we can choose a different chart (there are a tremendous number of charts available) that can be the best response for different answers.

Choosing the wrong visual aid or simply defaulting to the most common type of data visualization could cause confusion with the data consumers or lead to mistaken data interpretation.

The importance of choosing the right chart is one of the mandatory goals for everyone needs to prepare a data viz.

It's not always so easy choosing the right chart but, here, I don't want to focus on the rules we can use to choose the best type possible. I want to talk about pie charts, one of the most overused graphs in the world and (in most cases) not the best way to present data.

A pie chart is a circular graph that is divided in parts (slices) by lines coming from the center.

Honestly speaking, pies are not really the main problem, but I need to say that in most cases this type of chart deserves to be banned from our toolbox.

What we don't want to create are charts that often distort the information and make it more difficult for decision-makers to understand the messages they contain.

During several talks, often it happened to me to discuss about the idea to making better presentations, reports, and dashboards.

The first advice I always suggest, maybe one simple way, is to eliminate pie charts from the inventory.

My first suggestion is to try to replace this kind of chart with a horizontal bar chart, organized from greatest to least or vice versa: with bar charts, our eyes compare the end points. Because they are aligned at a common baseline, it’s very easy to assess relative size.

In this way is easier to see not only which is the largest segment, but also how incrementally larger it's than the others.

The story begins around 1801 when the first pie chart appeared in a publication entitled "The Statistical Breviary" by William Playfair.

In this publication, the author used different graphs to present areas, populations, and revenues of European countries.

Although, the term “pie chart” was coined years later (and it is not the only food metaphor that has been used to describe it).

We can start from four facts:

• We put charts to present information in a way that’s easier to understand
• People love circles
• Pie chart (with doughnut charts) is the most detested, misunderstood, defended, and talked
• Our brain is poor at comparing angles

The first point is pretty obvious: we must present data in a visually simple way. Our audience (we are not our audience) needs to understand quickly and clearly.

There are three different theories for why we love circles:

1. Evolution: circles and curvilinear shapes put us at ease compared to the sharp edges that remind us of the dangers like objects, arrows, animals’ teeth
2. Happiness: we equate circles with positive emotions (circular shapes like pleasant items such as a baby’s face)
3. Our own eyes: we look around through our rounded "lens"

Because pie charts were used for nearly everything, they’ve garnered a reputation for being ineffective, lazy, and just bad but, as William Cleveland, an American computer scientist and Professor of Statistics and Professor of Computer Science at Purdue University, wrote "pie charts have several perceptual problems. Pie charts do not provide efficient detection of geometric objects that convey information about differences in values."

In general, charts should make data easier to understand and make easier to compare different data sets doing that without increasing the complexity of what’s being presented.

Except in very rare cases, pie charts are extremely bad at making data easier to understand, and especially bad at helping us compare different data sets.

Pie charts may look pretty but are impossible to read quickly and accurately (it gets better if only two slices are shown).

Perhaps, as children, we learned to think about percentages through equal parts of a pie.

But when the parts are not the same, as often is the case in real life, it becomes difficult to envision the parts of a whole the pie chart was intended to facilitate.

People underestimate angles less than 90° and overestimate angles greater than 90°. Also, angles with horizontal bisectors (when the line dividing the angle in two is horizontal) appear larger than angles with vertical bisectors.

In short words, pie charts only make it easy to judge the magnitude of a slice when it is close to 0%, 25%, 50%, 75%, or 100%.

People just naturally aren’t very good at briefly distinguishing the differences between slices of a circle and, apart from this, it’s also nearly impossible to compare similar slices from two different pie charts.

We’re much better equipped to understand differences in rectangular shapes.

Data visualization uses four different elements to be effective: visual cues, a coordinate system, scale, and context.

A pie chart uses a polar system (when almost every chart uses a cartesian system, better known as X and Y chart) that is a coordinate system in which measurements are made by angular measurements.

In a pie chart the labels hardly line up, so we can create a cluttered and hard to read result (sometimes we can also use more than one pie chart, as they often make the data more complicated than before).

This kind of chart don’t work well at small size: a small pie chart is simply useless, whereas a small bar or line graph can still easily show differences in data. Small percentages (which might be important) are tricky to show in a pie.

We can also state that it’s much easier to distort the data on a pie than any other chart type. As we have said, we’re so bad at distinguishing between the different slices so, if you tilt or make “3D” a chart pie you quickly make it even harder to read and distort your data even more.

More notes:

• a pie chart only works with one set of data
• simplicity is always a key (putting more than 7 categories on a pie chart dilutes the meaning and confuses the viewer)
• we don’t represent negative values because we can’t show a piece of pie that’s already been eaten :-)
• don’t make viewers move their eyes back and forth from the legend

As a data professional we want to build something to tell a story immediately and accurately. Our users should not have to give it a minute to understand what is going on.

When we talk about data visualization what we need to do is always to consider the story we want to tell with data.

“When is it appropriate to use pie charts?”

“Never. Maybe.”