Start by separating reasons an answer is absent
An empty survey cell does not always mean the same thing. A respondent may have skipped a question, selected a visible “prefer not to answer” option, or been routed past it because an earlier answer made it irrelevant. A survey may also contain a “not applicable” or “don’t know” response. These cases can affect interpretation differently, so avoid combining them into a single “missing” category before checking what the questionnaire actually recorded.
Begin with a simple inventory for each question: who was eligible to see it, how many people provided one of its substantive answers, and how many fell into each other recorded category. If the survey did not capture why a question was left blank, label the absence as an unanswered question rather than guessing at the respondent’s intent. This groundwork matters more than the chart type: it determines what the displayed percentages can honestly mean.
Choose a denominator readers can understand
Suppose a question asks recent purchasers which feature influenced their choice. Showing each feature as a percentage of all survey respondents would answer a different question from showing it as a percentage of recent purchasers who answered. Neither denominator is automatically wrong, but the chart must identify which population it describes.
A useful default is to calculate response percentages among people who were eligible and gave a valid answer, then state that base beside the chart. When nonresponse itself is important to the story, show it separately rather than silently removing it. For example, a chart about awareness of a product might include “don’t know” as a visible response if that was an offered option. A question shown only to purchasers, however, should not treat nonpurchasers who never saw it as if they chose “not applicable.”
Be especially careful when comparing groups. If one group has many more unanswered questions than another, charts based only on completed answers may conceal a meaningful difference in participation. A brief note can show both the eligible number and the number who answered, letting readers judge what the percentages cover.
Give distinct response types distinct visual treatment
Use chart categories for answers respondents could actually select, including “don’t know” or “prefer not to answer” when those options are relevant to the question. Give those categories clear labels and a restrained color so they remain visible without being mistaken for a substantive preference. If “not applicable” was an offered answer, consider whether readers need to see its share or whether a note beside the chart is clearer.
For skipped questions and routing exclusions, a separate line beneath the chart is often easier to read than adding more bars or slices. It might state that the chart includes only people asked the question who provided an answer, followed by the counts of unanswered and not-asked cases. Keep this information close to the visual; a distant methodology note is easy to miss when a chart is shared on its own.
Avoid using the same visual treatment for “zero selected this option” and “no usable answer was recorded.” Zero is a result; missing is a limit on what can be reported. If a subgroup has too few completed answers for a useful comparison, consider leaving its chart value unplotted and explaining why, rather than drawing a zero-height bar.
Check the chart against the questionnaire
Before publishing, trace each plotted category back to the survey wording and routing. Confirm whether “other” was an offered choice, whether respondents could select more than one answer, and whether a blank was possible even when the question appeared. This check can reveal misleading labels: “all customers” may be inaccurate if the chart includes only customers who completed a follow-up question.
Then test the chart with a reader who has not seen the data table. Can they tell who was asked, who answered, and what happened to everyone else? Can they distinguish a respondent’s “not applicable” choice from someone who was never shown the question? If not, revise the title, base note, or category labels before adding more visual detail. A clear chart does not need to display every data-processing decision, but it should make the decisions that change its meaning easy to find.
