Pie chart maker

What a pie chart is actually for

A pie chart answers exactly one question: what share of the whole is each part?

It answers that question better than a bar chart does, and this is the part people forget when they repeat that pies are always bad. A pie is a closed shape. The reader sees immediately that the parts are components of one total, that they add to everything, and that nothing is missing. A bar chart of the same numbers shows five quantities and gives no visual signal that they belong to a single whole.

What a pie cannot do is support precise comparison. People read angles poorly — far worse than they read length. Two slices at 22% and 25% look identical. So the rule is: use a pie when the shares are the message and the differences are large; use bars when you need the reader to rank things accurately.

Pie versus bar: the honest comparison

	Pie	Bar
Signals "these form a whole"	Yes, inherently	No
Reading exact values	Poor — angles	Good — length
Ranking many items	Poor past four	Good to about fifteen
Comparing across two datasets	Bad — two pies are very hard to compare	Good
Space efficiency	Compact, roughly square	Needs width or height
Best number of categories	2–5	5–15

If you find yourself squinting to work out which of two slices is larger, that is the chart telling you it should have been a bar chart.

When to use a pie chart

It works well for:

• Market share with a handful of players plus an "Other"
• Budget or spend split across three to five categories
• Traffic or revenue by source, where the point is the dominance of one source
• A single dominant share — "62% of revenue comes from one product" is a pie's strongest use, because one large slice against the rest needs no precise comparison
• Yes/no or two-way splits, where a pie reads instantly

Reach for something else when:

Your situation	Better chart	Why
More than about six categories	Bar chart	Small slices become unreadable slivers
The parts do not sum to a meaningful total	Bar chart	A pie asserts a whole that does not exist
Comparing the same split across periods	Stacked bar	Several pies side by side cannot be compared
You want a headline number in the middle	Donut chart	The hole is usable space
Values can be negative	Bar chart	A negative share of a whole has no meaning
Differences under about five points matter	Bar chart	Angle perception is not that precise

The six-slice ceiling

Past about six slices a pie stops working, for a reason that is arithmetic rather than aesthetic. Once you have eight categories, several will be under 5%, and a 5% slice is 18 degrees — a sliver too narrow to label inside, too small to compare, and too similar to its neighbours to distinguish by colour.

Two ways out, and they are not equivalent:

Group the tail into "Other." Correct when the small categories genuinely do not matter individually. Keep "Other" under about 10% of the total — if it is bigger than your largest real slice, you have hidden the story rather than simplified it. Say group anything under 3% into Other.

Switch to a bar chart. Correct when the small categories do matter and you were only using a pie out of habit.

How the slices should be arranged

There is a convention, and following it makes a pie noticeably easier to read:

1. Start at 12 o'clock and go clockwise. That is where the eye begins.
2. Largest slice first, descending. The reader gets the ranking from position, not from angle comparison.
3. "Other" last, regardless of its size, because it is a residual rather than a category.
4. Percentages in the legend, not crammed into the wedges. Small slices cannot hold a label, and leader lines pointing out of a pie make it look like a diagram of a cell.

ChartGPT applies all four by default. Say keep the original order if your categories have a meaning that sorting would destroy — age bands, or a sequence of stages.

Two worked examples

1. Traffic sources

Data:

Source	Sessions
Organic search	52400
Direct	21100
Referral	15300
Social	11900
Email	4200

Prompt: Pie chart of sessions by traffic source as percentages. The point is that organic is more than half.

What you get: five slices, largest first from 12 o'clock, percentages in the legend rather than on the wedges. The title states the organic dominance. Email at 4% is on the edge of viable — one more small category and this becomes a bar chart.

2. Budget with a long tail

Data:

Category	Amount
Salaries	4200000
Marketing	980000
Tools	310000
Office	240000
Travel	120000
Legal	85000
Training	60000
Misc	45000

Prompt: Pie chart of budget split in $, group anything under 5% into Other

What you get: salaries, marketing, and a single "Other" bucket. Eight slices would have produced five slivers; three slices tells the actual story, which is that this is a payroll-dominated budget.

Five things that ruin a pie chart

1. Too many slices. Covered above, and it is by far the most common failure.

2. 3D and exploded slices. A tilted pie makes slices at the front appear larger than identical slices at the back — it distorts the only quantity the chart encodes. Exploding a slice does the same thing in a different way. ChartGPT does not draw either.

3. Slices that do not add to 100%. If your categories overlap, or respondents could pick more than one answer, the parts are not parts of a whole and a pie misrepresents them. Multi-select survey answers are the usual culprit — those are bars.

4. Two pies side by side. Comparing "before" and "after" as two pies asks the reader to compare angles across two shapes, which almost nobody can do. A stacked bar with two bars, or a grouped bar, shows the same change legibly.

5. A colour ramp. Sequential shades of one colour imply the categories are ordered by magnitude. Distinct hues are correct for categorical splits — and they need to differ in lightness too, so the chart survives greyscale.

Getting your data in

A pie needs two columns: category and value. That is it.

You can give either raw counts or percentages — if you give counts, the percentages are calculated for you; if you give percentages, they are used as supplied.

Naming your units — dollars, thousands, sessions, headcount — means the legend and subtitle read correctly. Note there is no value axis on a pie chart, so units appear in the labels and subtitle rather than along an edge.

How to make one

1. Paste two columns, category and value.
2. Say pie chart so it does not infer the type.
3. Name your units so the legend and subtitle read correctly.
4. Say what the chart is for — "the point is that organic is more than half" writes the title.
5. Adjust in words. Group anything under 3% into Other. Keep the original order. Sort largest first.
6. Export — SVG for slides, PNG for a quick paste, PDF for print.

Related chart types

• Donut chart — the same chart with the centre opened up for a total or a headline figure.
• Bar chart — when precise comparison matters more than the sense of a whole.
• Stacked bar — when you need to compare the same split across several groups or periods.

Exporting

SVG, PNG and PDF, on every plan including free. SVG matters here — curved edges are where PNG scaling artefacts show first. See pricing for what the paid plans add.