
Use a pie chart when you have two to five parts of one whole and want the biggest share to stand out. Use a donut chart for the same data when you also want to show a total or one key number in the center. Both are equally readable when labeled well: research by Skau and Kosara (2016) found donut charts are no less accurate than pie charts. If you have more than five parts, values that are close together, or several groups to compare, skip both and use a bar chart.
This guide is about choosing between the two. For step-by-step instructions in Excel, Google Sheets or PowerPoint, see our how to make a pie chart guide. For tools, see the best free pie chart maker comparison.
Update note (October 30, 2026): We added donut chart variations (progress rings, half donuts and nested donuts), advice for dashboards, waffle and 100% stacked bar alternatives, and links to the perception studies we cite. The October 29 update added the example charts, comparison table and decision flowchart. Details are at the end of the page.
A pie chart is a full circle split into slices, where each slice's angle, area and arc length show its share of the total. A donut chart (also spelled doughnut chart) is the same chart with the center removed, so readers judge mostly by arc length and the empty middle can hold a label. Both answer one question: what share of the whole does each part make up?
| Pie chart | Donut chart | Bar chart | |
|---|---|---|---|
| Best for | 2 to 5 parts of one whole | 2 to 5 parts plus a total or KPI | Ranking, close values, 6+ categories |
| How people read it | Angle, area and arc length | Mostly arc length | Length on a shared baseline |
| Accuracy for part-to-whole | Good with labels | About the same as a pie | Best for precise comparison |
| Center space | None | Holds a total or headline number | Not applicable |
| Main pitfall | Similar slices look equal | Thin rings make small arcs hard to see | Hides the "one whole" story |
| Pick it when | One slice should stand out | The total matters as much as the split | Readers must compare exact values |
Typical part-to-whole questions both charts answer well: market share among four competitors, website traffic by source, or votes on a single-choice survey question. Before you build either one, check three things: the parts add up to a meaningful 100%, there are five or fewer categories, and the goal is to show shares of the whole rather than exact differences between slices.
The charts below use fictional data so you can see the trade-offs instead of imagining them.

With four distinct shares, both charts work. The pie makes the 45% slice look dominant because area and angle add emphasis. The donut shows the same split and uses the center for the total, so readers get "how much" and "how it divides" in one glance. Choose the donut if that center number adds to the story; otherwise the pie is simpler.

With eight categories between 10% and 15%, the pie becomes a set of near-identical wedges, and a donut has the same problem. The sorted bar chart shows every value on one baseline, so a one-point difference is visible. Use a bar chart here, or group the smallest slices into "Other" if the whole matters more than each part.
When one answer wins most of the vote, say 70% of survey respondents, a pie or donut shows the dominance instantly. A donut can print "70%" in the center to reinforce the message. A bar chart also works, but the "share of the whole" impression is weaker.

| Your goal | Categories | Best chart | Why |
|---|---|---|---|
| Show a simple part-to-whole split | 2 to 5 | Pie chart | Fast, familiar view of a few shares |
| Show the split and a total or KPI | 2 to 5 | Donut chart | The center holds the headline number |
| Compare or rank values | Any | Bar chart | Lengths on a shared baseline are easiest to compare |
| Part-to-whole with many categories | 6 or more | Bar chart or treemap | Avoids thin, unlabeled slivers |
| Show one share as a count of 100 | 2 to 4 | Waffle chart | A 10 by 10 grid where each square is 1% is easy to count |
| Compare the same split across several wholes | 2 to 5 per whole | 100% stacked bar chart | Every bar is 100%, so the mix lines up on one axis |
| Show how a split changes over time or across groups | Any | Stacked bar chart | One baseline instead of several separate circles |
Switch away from circles when you have more than five or six categories, when slices are close in size, when readers need exact differences, or when you are comparing several datasets, such as market share in 2025 and 2026. Several pies side by side force readers to match slices across circles. A grouped or stacked bar chart puts everything on one axis. For layout and labeling rules, see our bar graph design best practices.
Most of the advice above is about a full pie or a full donut with several slices. Three donut variations come up often, and each follows its own rules.
| Variation | What it shows | Use it when | Avoid it when |
|---|---|---|---|
| Single-value donut (progress ring) | One share against the rest, such as 72% of a goal | You have one headline percentage or progress toward a target | You need to show more than two parts |
| Half donut (semi-donut or gauge) | A share or a score on a 180-degree arc | A KPI card or gauge needs to fit in a short, wide space | Readers must compare several slices |
| Nested donut or sunburst | Rings inside rings, such as region and then country | The data is a hierarchy and the inner ring is the main message | Readers must compare the outer rings, which have different lengths |
This is the strongest case for a donut. With only two parts, "done" and "remaining", readers do not need to compare similar slices. The center shows the exact number, and the ring shows it at a glance. Use one strong color for the share and a light gray for the rest, and print the percentage in the center, such as "72% of Q4 target".
A half donut takes up about half the height of a full one, which helps in dashboard cards. It works for one value or two or three parts. Keep the numbers in text below the arc, because a 180-degree arc makes angles even harder to judge. In Chart.js, a doughnut with circumference: 180 and rotation: -90 draws a half donut.
A nested donut puts one ring inside another. Readers can compare the inner ring well, but outer arcs are longer than inner arcs with the same share, so a direct comparison between rings is misleading. Use it for a hierarchy with a clear parent ring, and switch to a treemap or a grouped bar chart when readers must compare the rings.
On a dashboard, a donut with a number in the center works best as a KPI card: one metric, two or three parts, and the total in the middle. Do not place several donuts side by side to compare teams or months; use one 100% stacked bar chart instead. If the tool shows tooltips on hover, still print the key labels, because shared screenshots and printouts lose the hover.
In their 1984 graphical perception study, William Cleveland and Robert McGill ranked how accurately people judge different visual encodings. Position along a common scale ranked highest; angle and area ranked lower. That is why a bar chart wins whenever precise comparison matters, and why a pie chart works best for rough shares.
Pies are not always worse, though. Ian Spence and Stephan Lewandowsky tested proportion judgments in 1991 and found pie charts did as well as bar charts for judging one part's share of the whole, and better when readers had to combine parts, such as "A and B together". So a pie is a fair choice when the question really is "what share of the total?"
A common claim is that donut charts are worse because they remove the center where angles meet. Drew Skau and Robert Kosara tested this in 2016 by separating a pie's cues (angle, area and arc length). Their results suggest people do not read pies mainly by angle, and that donut charts are no less accurate than pie charts. Charts that could only be read by angle did worse. A useful way to think about a donut: it is close to a single stacked bar bent into a circle, which readers judge by length along the ring.
In practice, labels matter more than the shape. When each slice shows its name and percentage, readers use the text, and the small accuracy differences between a pie and a donut stop mattering.
| Tool | Pie or donut | Quick fixes |
|---|---|---|
| Excel, Word, PowerPoint | Insert > Insert Pie or Doughnut Chart | Sort the source data descending; add data labels from Chart Elements |
| Google Sheets | Chart editor > Pie chart; set a donut hole under Customize > Pie chart | Sort the range first; turn on slice labels |
| Chart.js | type: 'pie' or type: 'doughnut' (cutout sets the hole size) | Sort the data array and build the "Other" group in code |
| Power BI | Built-in Pie chart and Donut chart visuals | Group small categories in the data model before charting |
| Tableau | Built-in Pie mark type; no built-in donut | For a donut, put two pies on a dual axis and make the smaller top one a blank circle that holds the total |
To convert a cluttered pie into a bar chart, switch the chart type, put categories on one axis and values on the other, and sort the bars from largest to smallest. Our guide on how to make a bar graph covers the steps. For more chart tools, see our chart maker roundup.
| Mistake | Fix |
|---|---|
| 3D or exploded slices | Use a flat 2D chart |
| Too many slices | Group small ones into "Other", or switch to a bar chart |
| Several pies to compare groups or years | Use a grouped or stacked bar chart |
| Shares that do not add up to 100% | Use a bar chart, or fix the data so it forms one whole |
| Vague title | "Share of [total] by [category]", with a subtitle explaining any grouping |
Write alt text that states the chart type, the topic and the main values, and point to a data table when there are more than a few slices. For example:
Donut chart of website traffic by source in May. Organic search 45%, Direct 25%, Social 18%, Referral 12%. Total 8,000 visits.
Give screen reader users a real HTML table with a caption and header cells:
<table>
<caption>Website traffic by source, May</caption>
<thead>
<tr><th scope="col">Source</th><th scope="col">Share</th></tr>
</thead>
<tbody>
<tr><td>Organic search</td><td>45%</td></tr>
<!-- more rows -->
</tbody>
</table>
For color, WCAG 2.2 asks for a contrast ratio of at least 4.5:1 for normal text labels and 3:1 for graphical parts needed to understand the chart. Do not rely on color alone: direct labels, patterns or a table help readers with color vision deficiency.
AFFiNE is not a chart maker, so build the chart in Excel, Google Sheets or a charting library. Where AFFiNE helps is the review step: on an Edgeless whiteboard you can place the exported pie, donut and bar versions side by side, add notes about grouping and labels, and keep the source table and the final choice in the same doc for your team.
It is not better in general. A donut chart is about as accurate as a pie chart, and its center can show a total or key number. Choose it when that center label adds meaning; otherwise a pie is simpler.
Like a pie, it is hard to compare similar values, it gets cluttered past five or six categories, and it cannot show change over time. Very thin rings also make small slices hard to see.
Use a bar chart when readers need to rank or compare exact values, when values are close, when there are more than five or six categories, or when you compare several groups or time periods.
A donut chart is a pie chart with the center cut out. Readers judge a donut mostly by arc length instead of angle and area, and the empty center can hold a summary number.
It is usually called a semi-donut, half doughnut or gauge chart. It shows one value or a few parts on a 180-degree arc and fits well in dashboard cards. Print the numbers below the arc, because angles on a half circle are hard to judge.
Avoid it. Matching slices across separate circles is hard and error-prone. Use a grouped or stacked bar chart to compare composition across groups or years.
On October 30, 2026, we added a section on donut variations (progress rings, half donuts, nested donuts and sunbursts) and dashboard use, added waffle and 100% stacked bar charts to the decision table, added the 1991 Spence and Lewandowsky study for balance, linked the perception studies we cite, corrected the Tableau steps (Tableau has no built-in donut), and added an FAQ on half donut charts.
On October 29, 2026, we replaced the text-only examples with labeled charts that show the same fictional data as a pie, a donut and a bar chart, added an answer-first comparison table and a decision flowchart, cut the article roughly in half, checked the perception research and tool steps against their sources, and removed links that did not match the topic. AI tools assisted with drafting, and the content was reviewed and edited by the AFFiNE team.