Start with the comparison you want readers to make
A chart is most useful when its layout matches the question behind it. Suppose a survey asks respondents in three age groups how satisfied they are with a service. You might want to show the overall mix of answers within each age group. Or you might want to compare the share who said “very satisfied” across age groups. Those are different reading tasks, even though they use the same results.
Stacked bars emphasize composition: each bar represents a group, and segments within it show how that group’s responses are divided. Grouped bars emphasize comparison of individual categories: bars for the same response option sit alongside bars for other groups. Decide which relationship matters most before choosing a chart. If readers need to do both tasks, two small charts may work better than asking one chart to carry every comparison.
Use stacked bars to show the response mix
A stacked bar works well when response options form a meaningful whole, such as one answer per respondent to a satisfaction question. A 100% stacked bar gives every group the same total length, making it easier to compare proportions when groups have different numbers of respondents. Place response options in a consistent order in every bar, from the most negative to the most positive, for example.
The trade-off is that only the segments at the ends of the bars share a common baseline. Middle segments start in different positions, so small differences between groups can be hard to judge. A stacked chart can reveal that one group has a larger combined share of positive answers, but it is less effective when the main question is whether a particular middle response, such as “neither satisfied nor dissatisfied,” differs across groups.
Ordinary stacked bars, which show counts rather than percentages, answer a different question. A taller bar may simply reflect a larger survey group. Use counts when the number of responses is itself important, and make that purpose clear. If your aim is to compare attitudes across groups of unequal size, percentages are usually easier to interpret.
Use grouped bars for category-by-category comparisons
In a grouped bar chart, each response option has a bar for each audience group, arranged side by side. Because those bars begin at the same baseline, readers can more readily compare the share choosing a specific option. This is useful when a finding centers on one answer, such as which groups are most likely to select “very satisfied.”
Grouped bars become crowded as the number of audience groups or response options grows. Long labels, a large legend and many similar colors can turn a simple comparison into a search task. Consider showing only the categories relevant to the question, while explaining what has been omitted. If the full response scale matters, a 100% stacked chart or a set of small charts may be clearer. Avoid treating a grouped chart of selected answers as though it displays the entire distribution.
Check the survey numbers before drawing either chart
First, define the denominator behind every percentage. Is it all respondents in an audience group, only those who answered the question, or a narrower subset? Apply the same rule across groups and state it near the chart. If some respondents skipped the question, decide whether to show “no answer” as a category or exclude it from the percentage calculation; either choice should be visible to readers.
Next, check whether respondents could choose more than one answer. With a multiple-choice question that permits several selections, percentages across options may add up to more than 100%. Those answers do not form parts of a single whole, so a 100% stacked bar would be misleading. Grouped bars can show the share of each group selecting each option, provided the labels make clear that selections overlap.
Keep group sizes available alongside percentages. A difference that looks prominent on a chart may rest on very few responses in one group. When groups are small, describe apparent differences cautiously rather than implying precision the survey cannot support.
Make the final comparison easy to read
Give the chart a title that names the measure and the groups being compared. Label the percentage axis, include response options in a predictable order, and use the same color for each option throughout. Choose colors that remain distinguishable without relying on color alone; direct labels can help when space allows. In grouped bars, start a percentage axis at zero so bar lengths do not exaggerate differences.
Finally, ask someone unfamiliar with the results to describe what they notice first. If they focus on group size when you meant to show response mix, consider 100% stacked bars. If they struggle to compare one option across groups, try grouped bars. The better chart is not the one that holds the most data; it is the one that lets readers answer the intended question without guessing how the numbers were assembled.
