Why the percentages add up to more than 100%
A question such as “Which of these features do you use?” may allow each respondent to select several answers. Each option is then counted separately. If 100 people answer and 60 select search, 45 select saved lists, and 30 select alerts, the percentages are 60%, 45%, and 30% of respondents. Their sum is 135% because some people selected more than one feature. That total is not an error.
The most useful first step is to state exactly what a percentage represents. In this example, 60% means 60 of the 100 people who answered the question selected search. It does not mean search accounted for 60% of all choices made. Those are different measures: percentage of respondents describes how common an answer is among people, while percentage of selections describes its share of all marked answers.
Choose a denominator before making the chart
For most questions about how many people use, want, or experience something, use the number of eligible respondents who answered the question as the denominator for each option. Divide the number who selected an option by that respondent base. This lets a reader interpret every bar as a proportion of people, even when the bars collectively exceed 100%.
Define the base with care. If someone skipped the question, decide whether the chart describes all survey participants or only those who answered it, and label it accordingly. If the questionnaire used a “none of these” answer, check whether it was intended to be exclusive of other options. Do not quietly treat a blank response as “none”: skipping a question and actively choosing none convey different information.
Percentage of selections can be useful when the question is about the composition of all choices made. In the hypothetical example, the three options received 135 selections altogether, so search represents 60 of 135 selections. If you show that measure, call it “share of selections,” not “percentage of respondents.” Avoid switching between the two denominators across charts without a clear explanation.
Use bars to show options independently
A horizontal bar chart is often a clear choice for multiple-choice results. Put one answer option on each row, use a consistent scale starting at zero, and label the axis “Percentage of respondents selecting each option.” Long answer labels usually fit more comfortably beside horizontal bars than beneath vertical ones. Sorting bars from highest to lowest can make the most frequently selected options easy to find; retain the questionnaire order when that sequence carries meaning.
Include the respondent base near the chart, such as “Base: 100 respondents who answered this question.” A short note—“Respondents could select more than one option; percentages may total more than 100%”—resolves the apparent contradiction without asking readers to infer the survey format. Where space permits, show counts alongside percentages. Counts make small bases and rounding easier to judge.
A pie chart or a single 100% stacked bar is usually a poor fit for respondent percentages here. Both imply that the displayed categories are parts of one whole, yet one person may belong to several categories. Separate bars preserve that overlap instead of hiding it.
Compare groups without losing the base
Multiple-choice charts become more delicate when comparing audiences. Suppose you want to compare selections by new and returning customers. Calculate each option’s percentage using the number of people who answered the question within that group. Label both bases, because a difference of a few percentage points is harder to interpret when one group contains far fewer respondents than the other.
Grouped bars or a compact table can show the same options side by side. Keep option labels and scales consistent across groups. If the chart becomes crowded, focus on the options most relevant to the decision and make clear that the display is a selection of results. Do not describe a higher bar as proof of a group’s underlying preference: the survey wording, who responded, and the available answer choices can all affect the pattern.
Consider whether an option’s wording might overlap with another. If “email updates” and “newsletters” were both offered, some respondents may have selected both for similar reasons. The chart reports what people marked; interpreting why they marked it may require looking back at the question or using follow-up research.
Give readers the context they need to act
A useful chart supports a specific question, not just a ranking. Before publishing, ask what decision the display is meant to inform. If a team is deciding which features to explain in onboarding, the percentage of respondents selecting each feature may be more actionable than the share of all selections. The chart can identify commonly selected options, while comments or interviews may help explain what people meant by their choices.
Check that the title, labels, base, and note tell a consistent story. “Features selected by respondents” is more precise than “Feature preferences” if the question asked about use rather than preference. If respondents could choose up to three options, say so; that limit affects how the results should be read. Finally, use readable text labels rather than relying on color alone, so the chart remains understandable in grayscale or to readers who distinguish colors differently.
Totals above 100% need no special correction. They need an honest denominator and a visual form that treats answers as overlapping selections. With those pieces in place, readers can see both the most common answers and the limits of what the survey establishes.
