Updated August 2026. Every figure links to its original source.
Pie Chart Statistics
The pie chart is 225 years old, and the median pie chart people actually build has five slices.
William Playfair published the first one in 1801 in The Statistical Breviary. It had three slices. In the 225 years since, the format has been attacked in print since 1914 and measured in the lab since 1926, and the research is more mixed than the reputation. Cleveland and McGill found position judgments 1.96 times as accurate as angle judgments in 1984. Simkin and Hastie found pies and bars equivalent when the task is genuine part-to-whole estimation.
The five-slice median comes from our own logs: 13,058 charts exported by 5,496 people between January and August 2026, of which 77.9 percent had six slices or fewer. Everything below is sourced and linked.
History and origins
Almost everything known about the pie chart's origin traces to one paper: Ian Spence's 2005 study in the Journal of Educational and Behavioral Statistics, which examined Playfair's original plates.
The first pie chart appeared in 1801, in William Playfair's The Statistical Breviary.
Source: Ian Spence, Journal of Educational and Behavioral Statistics, 2005
The world's first pie chart had exactly three slices, showing the Asiatic, European, and African land areas of the Turkish Empire.
Source: Ian Spence, JEBS, 2005
Playfair introduced three entirely new chart forms in a single 1801 volume: the circle diagram, the pie chart, and a Venn-like diagram.
Source: Ian Spence, JEBS, 2005
Playfair invented the pie chart about 15 years after the line graph and the bar chart, which he published in 1786.
The pie was his last major graphical invention, not his first.
Source: Ian Spence, JEBS, 2005
Playfair used his own invention in only two publications beyond The Statistical Breviary, a handful of pie charts across his entire career.
Source: Ian Spence, JEBS, 2005
Playfair hand-colored every pie chart himself, once for every copy of the book, after the copperplate impressions were made.
Source: Ian Spence, JEBS, 2005
The pie chart's intellectual ancestry runs back to Ramon Llull's Ars Magna of 1305 to 1308, nearly 500 years before Playfair.
Source: Ian Spence, JEBS, 2005
The first person to publicly embrace the pie chart was Charles-Joseph Minard, who put pies on statistical maps by 1858, 57 years after the invention.
Source: Ian Spence, JEBS, 2005
The first published attack on the pie chart came in 1914, in Brinton's Graphic Methods for Presenting Facts, the first widely circulated textbook on statistical graphs.
Brinton wrote that the circle with sectors is not a desirable form of presentation. The argument is 112 years old.
A pie chart offers at least five natural reference anchors at 0, 25, 50, 75, and 100 percent, against two or at most three for a divided bar.
Source: Ian Spence, JEBS, 2005
In French the pie chart is named after a cheese, le camembert.
Source: Ian Spence, JEBS, 2005
Playfair's Turkish Empire chart, redrawn from the three land areas he plotted. Three slices, hand colored.
Playfair split the circle into three sectors and colored each one by hand. It was the first chart to show empirical proportions and distinguish the parts by color. Proportions here are illustrative of the form, not a transcription of his measurements.
Nightingale and the rose diagram
Florence Nightingale's 1858 diagram is constantly described as a pie chart. It is the mathematical inverse of one, and the reason she built it that way is itself a lesson in chart perception.
Nightingale's 1858 polar area diagram covered 24 months of Crimean War deaths, split into two 12-month roses.
Source: Friendly and Andrews, SORT, 2021
Her chart is not a pie chart: its wedges have equal angles and varying radius, the exact inverse of a pie's varying angles and equal radius.
Source: Friendly and Andrews, SORT, 2021
Nightingale scrapped her first design after realizing a linear radius makes a doubled death rate look four times as large, then replotted deaths on a square root scale so area matched the count.
A working statistician correcting for area perception in 1858, 126 years before Cleveland and McGill measured the same effect.
Source: Friendly and Andrews, SORT, 2021
Her data showed an annualized 60 percent mortality from disease alone in the first seven months of the Crimean campaign, exceeding the Great Plague of London.
Source: Friendly and Andrews, SORT, 2021
At Scutari the annual mortality rate among the sick reached 415 percent in February 1855, a wedge that would have projected 40 inches from the center on a linear scale.
Source: Friendly and Andrews, SORT, 2021, quoting Nightingale, 1858
Nightingale was elected the first female member of the Statistical Society, now the Royal Statistical Society, in 1858.
Source: Science Museum, London
The radial wedge chart predates Nightingale by about 29 years: Andre-Michel Guerry published the form in 1829.
Source: Friendly and Andrews, SORT, 2021
Perception research
Four experiments do most of the work in this argument. Two of them are usually quoted against pie charts, and two of them are usually left out.
Cleveland and McGill concluded in 1984 that radical surgery on pie charts was needed, and offered dot charts as the replacement.
Source: Cleveland and McGill, Journal of the American Statistical Association, 1984
Their accuracy ordering runs position on a common scale first, then position on nonaligned scales, then length, then angle and slope, then area, then volume and shading.
Source: Cleveland and McGill, JASA, 1984
Position judgments were 1.96 times as accurate as angle judgments in their second experiment, across 51 subjects with usable data.
This is the number behind essentially every piece of advice to use a bar chart instead of a pie chart.
Source: Cleveland and McGill, JASA, 1984
Heer and Bostock replicated the ranking on Mechanical Turk in 2010 with 50 subjects per chart, and position encoding still significantly outperformed length.
Source: Heer and Bostock, ACM CHI, 2010
That replication tested seven encodings across ten charts each, 70 trials in total, with pie chart angle and circular area among the least accurate.
Source: Heer and Bostock, ACM CHI, 2010
Both the 1984 original and the 2010 replication score accuracy as the base 2 logarithm of the absolute error plus one eighth, the standard log absolute error measure in chart perception.
Source: Heer and Bostock, ACM CHI, 2010
Rectangular area judgments were no more accurate than circular area judgments, and comparing shapes with a 1 to 1 aspect ratio produced the worst performance of all.
Source: Heer and Bostock, ACM CHI, 2010
The entire crowdsourced replication cost 367.77 dollars and used 186 participants, a sixfold saving over the lab equivalent.
Source: Heer and Bostock, ACM CHI, 2010
Simkin and Hastie showed in 1987 that when the task is genuine part-to-whole estimation, pie charts and bar charts perform equivalently.
This result is routinely omitted from summaries of the 1984 research it followed.
Source: Simkin and Hastie, JASA, 1987, as reported in Spence, 2005
Spence and Lewandowsky found in 1991 that pie charts beat bar charts on compound comparisons of the form is A plus B greater than C plus D.
Source: Spence and Lewandowsky, Applied Cognitive Psychology, 1991, as reported in Spence, 2005
The first psychological test of the pie chart, by Eells in 1926, found readers estimated proportions faster and more accurately from pies than from bars.
Across ten analysis tasks with 180 participants, the pie chart matched bar charts and tables on seven of them and was the fastest chart of all for clustering.
It lost only on correlation, anomalies, and distribution. Chart choice is task dependent.
How people actually read a pie chart
The standard criticism assumes readers judge the central angle. Two studies, 90 years apart, say they mostly do not.
Asked how they read a pie chart, Eells's 1926 subjects reported arc length 51 percent, area 25 percent, angle 23 percent, and chord length 1 percent.
Angle came third. The cue the criticism targets is not the cue most readers use.
Source: Eells, 1926, as reported in Skau and Kosara, EuroVis, 2016
Skau and Kosara found angle is the least important visual cue in reading pies and donuts, and that arc length may be the most important.
Source: Skau and Kosara, Computer Graphics Forum, EuroVis, 2016
A donut chart was read as accurately as a pie chart, with mean log error of 1.000 for the baseline donut against 1.032 for the baseline pie.
Removing the center of a pie does not affect how precisely it can be read.
Source: Skau and Kosara, EuroVis, 2016
Strip a pie down to angle alone and error nearly doubles: 1.967 log error for angle-only pies and 2.279 for angle-only donuts, against 1.032 for a normal pie.
Source: Skau and Kosara, EuroVis, 2016
Donut hole size barely matters. Inner radii from 0 to 80 percent of the outer radius produced log errors between 1.16 and 1.33, and only the thinnest ring at 97 percent degraded to 1.553.
Source: Skau and Kosara, EuroVis, 2016
In a random sample of infographics on Visual.ly, 36 percent of those containing charts used some form of pie or donut chart.
Source: Skau and Kosara, EuroVis, 2016
Mean log absolute error, lower is better. Kosara and Skau, EuroVis 2016. Every variation reads worse than a plain pie.
Exploded, enlarged, and other variations
Kosara and Skau tested the decorative variations people reach for when they want a slice to stand out. All of them cost accuracy, including the one that changes no geometry at all.
Every tested pie variation increased reading error. Mean log error was 1.151 for a plain pie, 1.236 exploded, 1.338 with an enlarged slice, 1.487 squared, and 1.570 as an ellipse.
Simply pulling one slice out of a pie raised reading error by about 7 percent, despite changing neither the angle, nor the area, nor the arc length.
The authors called the result unexpected and concluded exploded pies should be avoided.
Enlarging one slice's radius to emphasize it makes readers overestimate that slice by a factor of 1.6 on average, while the drawn area only grew by a factor of about 1.28.
Across 6,420 trials with 108 participants, every pie variation showed the same systematic bias: small values overestimated and large values underestimated.
How often pie charts are actually used
Two corpus studies, 20 years apart, both find the pie chart much rarer in published work than the volume of argument about it would suggest.
In a corpus of more than 41,000 web visualizations, pie charts made up under 5 percent of output in four of the five tools studied, as low as 0.4 percent on Plotly and 0.6 percent in D3.
Source: Battle et al., ACM CHI, 2018
Even where the pie is one of only four chart options offered, users pick it least: 23.2 percent pie against 34.7 percent line and 31.8 percent bar on Chartblocks.
Source: Battle et al., ACM CHI, 2018
Bar, line, scatter, and map together account for between 64.8 and 90.5 percent of real-world web visualizations depending on the tool.
Source: Battle et al., ACM CHI, 2018
Only 0.05 percent of 20 million crawled web pages contained an SVG chart of any kind, roughly 10,000 pages.
Source: Battle et al., ACM CHI, 2018
In a survey of six 1998 publications, pie charts made up about 10 percent of charts in news magazines and were rarer still in the business and scientific press.
Source: Ian Spence, JEBS, 2005
Playfair's three inventions, the line graph, bar chart, and pie chart, accounted for about 50 percent of all diagrams in news magazines and 60 to 80 percent in the business and scientific press.
Source: Ian Spence, JEBS, 2005
Color vision and slice colors
A pie chart identifies its categories by color, which makes color vision deficiency the hardest constraint on the format. These are the numbers to design against.
Color blindness affects roughly 1 in 12 men, about 8 percent, and 1 in 200 women, an estimated 300 million people worldwide.
Source: Colour Blind Awareness
The US National Eye Institute confirms about 1 in 12 men have color vision deficiency, and red to green confusion is the most common type.
A red slice adjacent to a green slice is the single most common avoidable pie chart mistake.
Source: National Eye Institute, NIH
About 3 million people in the UK, roughly 4.5 percent of the population, are color blind, and around 40 percent of color blind pupils leave school unaware of it.
Source: Colour Blind Awareness
A person with moderate red to green color blindness can accurately name only about 5 pencils out of a standard box of 24.
A useful way to picture how many distinct slice colors a chart can really carry.
Source: Colour Blind Awareness
Low contrast text was found on 83.9 percent of the top one million home pages, the most common accessibility failure and the one that most often makes chart labels unreadable.
Source: WebAIM Million, February 2026
16.2 percent of all home page images had missing alternative text, which for a chart image means the data is simply gone for screen reader users.
Source: WebAIM Million, February 2026
What the critics actually measured
The case against pie charts is often quoted as attitude. Underneath it there are specific numbers, and they are worth citing precisely.
Edward Tufte wrote in 1983 that the only worse design than a pie chart is several of them, because the viewer must compare quantities in spatial disarray both within and between pies.
Source: Tufte, 1983, quoted in Stephen Few, Perceptual Edge, 2007
Stephen Few argues pie charts are only reliably readable when a slice is close to 0, 25, 50, 75, or 100 percent.
Source: Stephen Few, Save the Pies for Dessert, Perceptual Edge, 2007
Perceived area follows a power law with an exponent of about 0.8, meaning area is systematically underestimated, while relative line length is perceived almost perfectly.
Source: Kosslyn, Graph Design for the Eye and Mind, 2006, quoted in Few, 2007
Asked to judge a circle whose area was 16 times another, Few's students guessed anywhere from 6 to 50.
Angle judgments are biased in a predictable direction: acute angles below 90 degrees are underestimated and obtuse angles above 90 degrees are overestimated.
Source: Robbins, Creating More Effective Graphs, 2005, quoted in Few, 2007
Our own usage data
Numbers below come from anonymized event logs on this site, January 29 to August 18, 2026, covering 13,058 finished chart exports by 5,496 people. This is the part of the page nobody else publishes, so it is free to cite as long as you link back.
The median exported pie chart had 5 slices, and the mean was 5.6.
The crowded twelve-slice pie used in most criticism of the format is not what people are actually making.
77.9 percent of exported charts had six slices or fewer. Only 8.9 percent had ten or more.
Five slices was the single most common chart size at 19.3 percent of exports, narrowly ahead of two slices at 18.1 percent.
The two-slice spike is people making a single proportion, which is the simplest possible use of the format.
78.7 percent of exports were pie charts and 20.6 percent were doughnut charts, with bar and line charts together under 1 percent.
Only 4.4 percent of exports used the 3D effect, 573 charts out of 13,058.
3D pie charts are heavily searched for and rarely shipped.
PNG accounted for 51.2 percent of exports, JPEG 19 percent, PDF 17.1 percent, and SVG only 12.7 percent.
Pasting data beat typing it and beat CSV upload: 6,338 paste selections against 4,870 manual and 2,267 CSV.
92.7 percent of palette selections kept the default palette rather than switching to one of the eight alternatives.
Of 1,037 palette events, 961 were the default. Whatever colors a tool ships with are the colors most charts will wear, which puts the accessibility burden on the tool.
Share of exports by number of slices, January 29 to August 18, 2026. Pie Chart Generator first-party data.
Cite this page
Writers and researchers are welcome to use these figures. Please credit the original source next to each statistic, and link back to this page so readers can check the full list. Copy one of the lines below.
Plain text attribution
Pie Chart Generator, "Pie Chart Statistics", August 2026. https://www.pie-chart-generator.com/pie-chart-statisticsHTML link
<a href="https://www.pie-chart-generator.com/pie-chart-statistics">Pie Chart Statistics</a> — Pie Chart GeneratorFrequently asked questions
When was the pie chart invented?
In 1801. William Playfair published the first pie chart in The Statistical Breviary, and it had three slices showing the Asiatic, European, and African land areas of the Turkish Empire. Playfair had already invented the line graph and the bar chart about 15 years earlier, in 1786. The pie chart was his last major graphical invention.
Are pie charts accurate?
Less accurate than bar charts for precise comparison, and more accurate than their reputation suggests for part-to-whole reading. Cleveland and McGill measured position judgments as 1.96 times as accurate as angle judgments in 1984. Simkin and Hastie found in 1987 that when the task is actually estimating a part of a whole, pies and bars perform equivalently, and Spence and Lewandowsky found in 1991 that pies beat bars on compound comparisons such as whether A plus B exceeds C plus D.
How many slices should a pie chart have?
Small numbers, and real usage bears that out. Across 13,058 charts exported from this tool between January and August 2026, the median chart had five slices and 77.9 percent had six slices or fewer. Only 8.9 percent had ten or more.
Is a donut chart worse than a pie chart?
No. Skau and Kosara found the donut chart is read as accurately as the traditional pie chart, with the baseline donut actually scoring slightly lower error than the baseline pie. Hole size barely matters either, from a solid pie up to an inner radius of 80 percent, and only the very thinnest ring degrades accuracy.
Should I explode a slice to highlight it?
No. Kosara and Skau measured mean log error of 1.151 for a plain pie against 1.236 for an exploded pie, even though exploding a slice changes neither its angle, nor its area, nor its arc length. Enlarging one slice's radius is worse: readers overestimate that slice by a factor of 1.6 on average.
How common are pie charts in practice?
Rarer than their reputation implies. A corpus of more than 41,000 web visualizations found pie charts made up under 5 percent of output in four of the five tools studied, as low as 0.4 percent on Plotly and 0.6 percent in D3. Spence found pies made up about 10 percent of charts in the popular press in 1998, and less in business and scientific publications.
What colors should pie chart slices be?
Colors that survive color vision deficiency, which affects about 1 in 12 men and 1 in 200 women, roughly 300 million people worldwide. Red and green confusion is the most common form, so a red slice next to a green slice is the single most common avoidable mistake. Direct labels on slices remove the dependence on color entirely.
Build a pie chart with five slices or fewer
Paste your numbers, pick a palette, download a PNG or SVG. No signup required.
Open the pie chart generator