When to Use a Pie Chart (And When Not To)
Use a pie chart only for parts of one whole, at one moment, with five or fewer slices. Rules, design fixes, better alternatives and a decision table.
https://www.youtube.com/watch?v=o7F-tbBl_hA
published
Outrank AI
when to use a pie chart, data visualization, dashboard design, chart types, business intelligence
0afc3f1d-cafd-40c6-bdd8-232529ea0ffd
Use a pie chart only when you are showing the parts of a single, meaningful whole, at a single point in time, across five or fewer categories that add up to 100%. If any one of those three conditions fails, a bar chart, stacked bar chart or line chart will tell the story more accurately. That is the whole rule, and almost every bad pie chart you have ever seen breaks at least one part of it.
Pie charts are not inherently bad. They are narrow. They answer one question extremely well - "what share of the total is this?" - and every other question badly. This guide covers the three conditions in detail, the design choices that make a valid pie chart readable, the alternatives that beat it, and how to decide in about ten seconds.
The One Question a Pie Chart Answers
A pie chart communicates compositional data: individual pieces that combine into one complete total. Last month's website sessions split into organic search, paid social, referral and direct. This quarter's marketing budget split across channels. Today's open support tickets split by priority.
The circle itself is the message. People see a closed shape and instantly understand they are looking at all of something, not a sample of it. That familiarity is why pie charts survive in executive summaries: the reader does not have to learn the chart before reading the data.
The same shape is why they fail everywhere else. A pie chart is a photograph, not a film. It has no axis for time, no baseline for comparison, and no way to show a category that grew. Ask it to do any of that and you get a chart your audience has to decode instead of read.
The Three Rules That Decide It
Rule 1: The Slices Must Form a Complete Whole
Your categories have to be mutually exclusive (every data point belongs to exactly one slice) and collectively exhaustive (together they account for everything). A pie chart is a promise to the reader that these are all the slices and that together they are the entire thing.
This is where survey data usually disqualifies itself. If respondents could select multiple answers, the percentages sum past 100% and the "whole" is fiction. The same applies if you quietly drop a small "Other" bucket to tidy the chart - the moment the remainder disappears, every remaining slice is overstated.
Rule 2: Two to Five Slices, Six at the Absolute Limit
Human perception is poor at comparing angles and areas and good at comparing lengths. Past five or six slices, the wedges become slivers, the labels collide, and the reader is left estimating. Four sales regions is a fine pie chart. Fifteen sales territories is a bar chart.
If you genuinely have more categories, group the tail into a single, honestly labelled "Other" slice - and keep it small. When "Other" is one of your largest slices, the chart has stopped being informative.
Rule 3: Compare Parts to the Whole, Not to Each Other
Use a pie chart to say "organic search is roughly half of all traffic." Do not use it to say "referral traffic beat email traffic." At 12% versus 11%, those two wedges are visually identical, and your reader will either misread them or give up and read the labels - at which point you should have shipped a sorted bar chart.
A pie chart answers "what share of the whole is this?" A bar chart answers "which is bigger, and by how much?" Picking the chart is really picking which of those questions matters to your reader.
Pie Chart vs. the Alternatives: A Decision Table
This is the table we use internally when reviewing dashboards. Find the job in the left column, then use the chart in the middle column.
What you need to show | Best chart | Why the pie chart loses |
|---|---|---|
Share of one total, 2-5 categories, one moment in time | Pie or donut chart | Nothing loses - this is the pie chart's job |
Ranking categories against each other | Sorted bar chart | Lengths on a shared baseline beat angles for precision |
How a mix changes month over month | Stacked bar chart (or 100% stacked) | A row of pies forces the reader to compare across circles |
A trend or trajectory in a single metric | Line chart | A pie chart has no time dimension at all |
Composition of several different groups side by side | Grouped or stacked bar chart | Cross-circle comparison has very high cognitive load |
Many categories, some nested inside others | Treemap | A flat pie chart cannot express hierarchy |
Distribution of a single continuous variable | Histogram | There is no meaningful part-to-whole relationship |
If you want the longer version of this decision, our guide to choosing the right charts and the reference on types of graphs and when to use them both go deeper. For the specific part-to-whole case, see comparing one piece of data to the whole.
Designing a Pie Chart People Actually Read
Once your data passes the three rules, four design decisions determine whether the chart lands.
Label directly. Put the category name and percentage on or beside each slice. A separate legend makes the reader's eyes bounce between the key and the chart for every single slice.
Order the slices. Largest slice starting at 12 o'clock, then descending clockwise. This one convention removes most of the effort of reading a pie chart.
Use colour as a spotlight, not decoration. One saturated colour for the slice that matters, neutral greys for the rest. Then check the palette against colour-vision deficiency.
Never use 3D or exploded slices. Tilting the plane makes foreground wedges look larger than identical background wedges. An exploded slice screams importance regardless of its value. Both break the honest relationship between area and number.
These are the same instincts that separate a clear dashboard from a cluttered one - our write-up on data visualization best practices covers them across every chart type, and the collection of bad data visualization examples shows what happens when they are ignored.
Where Pie Charts Earn Their Place on a Dashboard
On a business dashboard, a pie chart works as a glanceable summary tile, not as an analysis surface. Three patterns hold up in practice:
Lead source mix. A marketing lead sees at a glance that organic search is the dominant channel this quarter and paid social is second. The follow-up question - is organic growing? - belongs to a line chart next to it.
Expense allocation. A finance lead reviewing quarterly spend across four or five buckets gets the shape of the budget instantly, which is usually enough to decide what to interrogate further.
Support queue composition. An operations lead checking the split of open tickets by priority gets a workload temperature check without reading a table.
The pattern is consistent: the pie chart states the situation, and something else explains it. A good dashboard makes that next step cheap - click the slice, and the rest of the view filters to it. At that point the pie chart stops being a static picture and becomes navigation. The principles are the same on small screens, though space pressure is harsher; see our notes on designing dashboards for mobile users.
Getting the Chart Right Without a Rebuild
Most bad pie charts are not authoring mistakes - they are maintenance debt. Someone picked a chart type once, the underlying question changed, and nobody went back. That is a workflow problem more than a design one.
In Querio, every answer is produced as real SQL and Python inside a reactive notebook: the cells are connected, so when the query changes, the chart updates with it rather than going stale. Charts are built on Vega-Lite via Altair, and dashboards ("boards") are assembled directly from those notebooks, with the layout stored as code the agent can read and edit. So "make this a sorted bar chart instead" is a request, not a rebuild - and because the SQL behind the chart is inspectable, you can check that the slices really do add up to the whole before you publish it.
Frequently Asked Questions
Can I use a pie chart to show change over time?
No. A pie chart has no time dimension, and placing several side by side asks the reader to compare wedge angles across separate circles - a task people are demonstrably bad at. Use a line chart for a trend, or a 100% stacked bar chart when you need the mix to stay visible as it shifts.
What is the maximum number of slices?
There is no formal limit, but the practical ceiling is six, and the comfortable range is two to five. Beyond that, adjacent slices become visually indistinguishable. Either group the tail into a single "Other" category or switch to a sorted bar chart, which handles many categories without any loss of clarity.
Is a donut chart better than a pie chart?
A donut chart is a pie chart with the centre removed. The hollow middle is useful real estate for the total, and some designers find arc lengths marginally easier to judge than wedge angles. The limitations are identical, so apply exactly the same three rules.
What about a pie chart where the slices do not sum to 100%?
That is not a pie chart, it is a misleading graphic. If your categories overlap - as with multi-select survey answers - use a bar chart of the individual response rates instead, and label the axis clearly so nobody assumes the values are shares of a whole.
Should percentages or raw values be labelled?
Show percentages, since that is what the chart is for, and include the total somewhere on the chart - in the donut hole, the title or a subtitle. Percentages alone hide sample size, and "45% of new leads" reads very differently at 40 leads than at 4,000.
