Part to Whole
Part-to-whole figures show how a total divides into named categories. Use this section when your main question is about composition: what share of the whole each category represents, and whether that composition should be shown for one whole or compared across several groups.
This branch covers part-to-whole displays, not precise category-to-category comparison, statistical testing, or open-ended custom chart design.
If you need to compare values accurately across categories or groups, go to Distribution and Group Comparison. If you need to choose a test for proportions or contingency tables, go to Categorical and Association methods.
Start here
First decide whether you are showing one whole or comparing composition across multiple groups. Single-whole composition usually starts with pie or donut; across-group composition starts with stacked bar.
| If your goal is to... | Start with... |
|---|---|
| show the simplest view of one whole divided into categories | Pie Chart |
show one whole and use the center for Total n, total value, or a short summary | Donut Chart |
| compare how composition changes across several groups | Stacked Bar Chart |
| compare category values precisely rather than show rough shares | Distribution and Group Comparison |
| decide which inferential method to use for proportions or contingency tables | Categorical and Association methods |
When part-to-whole framing is the right choice
Use a part-to-whole figure when all of the following are true:
- The main question is "what share of the total does each category represent?"
- The audience needs a quick impression of composition, not a precise reading
- For a single whole, pie and donut work best when you have five or fewer categories
- For a single whole, slice differences should be large enough to read from angle alone (a rough guide: adjacent slices should differ by at least 10 percentage points)
- When composition varies across multiple groups, use stacked bar rather than applying the pie/donut constraints to the entire part-to-whole family
- Your choice is about showing composition, not about selecting a statistical test or making a precise value comparison across categories
Do not use a pie or donut chart when:
- You need readers to compare values accurately across categories
- Categories are similar in size; small angular differences are unreliable [1]
- You have more than five categories; cognitive load rises sharply with each additional slice [5]
- The underlying question is really a group comparison (use Distribution and Group Comparison figures instead)
- Composition varies across multiple groups and the comparison across groups is part of the message
Choosing between pie and donut
| Situation | Recommended |
|---|---|
| Simple composition, no additional summary needed | Pie Chart |
| Center space available for total n or a key summary value | Donut Chart |
| Five or fewer categories, rough shares only | Either |
| Multiple categories, composition varies across groups | Stacked Bar Chart |
| Accurate category comparison required | Neither; use a bar or strip plot |
The center opening in a donut chart does not improve comparison accuracy [3]. Choose donut when the center space adds a useful label, not because it looks more modern or fixes the limitations of the pie format.
In the current product, donut center labels can show Total n, Total value, Largest category, or custom text. Both pie and donut charts also support multiple slice-label modes: Name + %, Name + value, Name + value + %, % only, and Value only.
Charts in this section
A full circle divided into labeled slices. Use for the simplest part-to-whole message with no additional summary in the figure.
The same chart with a center opening. Use when the center can carry a meaningful label such as total sample size, total value, or the dominant category.
Bars divided into colored segments across multiple x-axis categories. Use when the question is how composition changes from group to group, especially when a pie or donut chart would need to be repeated several times.
Design Rationale & References
Licklider's design choices
Licklider treats pie and donut charts as use-with-caution figures because the visual encoding, angle and area, is systematically less accurate than the length-based encoding used in bar and dot plots [1, 2]. Readers can reliably detect rough proportional differences (for example, roughly half vs roughly a quarter) but struggle to compare slices of similar size, especially non-adjacent ones [4, 5]. For this reason, Licklider recommends these charts only when the part-to-whole message is the primary point and category count is small. When comparison between categories is the real goal, bar-like or point-based figures provide more reliable communication [1, 2].
Methodological foundations
Cleveland, W. S., & McGill, R. (1984). Graphical perception: Theory, experimentation, and application to the development of graphical methods. Journal of the American Statistical Association, 79(387), 531-554. Establishes the hierarchy of visual encodings: position along a common scale (bar charts) is decoded more accurately than angle or area (pie charts), the foundational basis for Licklider's preference for bar-like figures in comparison tasks.
Heer, J., & Bostock, M. (2010). Crowdsourcing graphical perception: Using Mechanical Turk to assess visualization design. CHI '10: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 203-212. Large-scale replication confirming that pie charts produce significantly larger reading errors than bar charts in modern digital environments.
Hollands, J. G., & Spence, I. (1992). Judgments of change and proportion in graphical communication. Applied Cognitive Psychology, 6(3), 235-258. Demonstrates that pie charts are effective for part-to-whole proportion judgments but ineffective for comparison between elements, directly motivating Licklider's narrow use-case definition.
Known limitations
Simkin, D., & Hastie, R. (1987). An information-processing analysis of graph perception. Journal of the American Statistical Association, 82(398), 454-465. Shows that comparing non-adjacent slices in a pie chart requires mental rotation, increasing cognitive load, the basis for Licklider's warning about large category counts.
Skau, D., & Kosara, R. (2016). Arcs, angles, or areas: Individual data encodings in pie and donut charts. Computer Graphics Forum, 35(3), 121-130. Finds that readers use arc length as well as angle when reading pie-style charts; donut charts do not reduce accuracy relative to pie charts, suggesting the choice between them should be based on layout needs rather than perceptual performance.
Paradigm shifts worth knowing
- Batch, A., et al. (2020). There is no spoon: The effects of hole size on the perception of donut charts. CHI '20: Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. Hole size in donut charts has minimal effect on reading accuracy, and placing a label in the center does not harm performance, supporting Licklider's center-label feature for donut charts.
See also
- Distribution and Group Comparison for comparing values across groups rather than showing composition
- Categorical and Association methods for chi-square and proportion testing