Forms · Glossary

Should a survey include a not applicable option?

Not applicable means the question does not apply to this respondent. Do not know means it applies but they cannot judge. Prefer not to say means they could answer and choose not to. All three are legitimate answers, they are not interchangeable, and each one is coded differently when the results are analysed.

These three options are usually treated as one escape hatch and dropped into a question list without much thought. Getting them right is the difference between a clean denominator and an analysis built on guesses.

Indunil Asanka · Co-founder

5 min read · Published 12 September 2026

Three escape options compared
	Not applicable	Do not know	Prefer not to say
Means	The question does not apply to me	It applies but I cannot judge	I could answer and choose not to
Use on	Items that only apply to some respondents	Knowledge or observation questions	Sensitive or personal questions
Coded as	Excluded from the base for that item	Reported separately, not as a midpoint	Excluded, and the count reported
If omitted	People guess or skip	The middle of the scale absorbs them	People abandon or answer falsely
Placement	After the scale, visually separated	After the scale	Last option on the question
Report	State the base for each item	A high rate is a finding in itself	A high rate signals a trust problem

The denominator problem

The reason these options matter is arithmetic. If a rating question was answered by everybody, the percentages describe everybody. If a fifth of respondents had no basis to answer and were forced to pick a point anyway, the percentages describe a mixture of judgements and guesses, and nothing in the data marks which is which. Offering not applicable moves those people out of the base for that item, so the remaining numbers describe people who actually had a view. The cost is that the base varies between items, which has to be stated in the report. That is a small price, and a report that gives the base per item is more credible than one with a single sample size at the top.

Do not know is a finding, not a nuisance

On knowledge and observation questions, a high rate of do not know answers is information rather than missing data. In a 360 feedback process it usually means the reviewer does not see that behaviour, which says something about visibility rather than performance. In a customer survey it often means a feature exists that nobody has encountered. Treating those answers as a midpoint hides the finding and distorts the average, which is why the option should be separate from the scale rather than sitting at its centre. Report the rate alongside the results, and watch it across waves, since a rising rate usually means something has become less visible rather than less liked.

Prefer not to say, and when to offer it

This one belongs on sensitive questions, particularly demographics, income, health and anything about identity. Offering it costs a small amount of data and buys two things: respondents who would otherwise abandon the survey stay, and the answers you do get are more likely to be true. A high rate on a particular question is itself a signal, usually that the question felt intrusive or that people do not trust how the answer will be used. In a workplace survey it is worth reading alongside the overall response rate, since both move together when trust is low. Never make a sensitive question required without this option.

Where the escape option becomes a problem

Adding all three to every question is the opposite failure. Too many escape routes give people an easy way to finish a survey without engaging, and the answer counts drop to the point where nothing can be said. Put not applicable only on items that genuinely do not apply to everybody, do not know only on questions requiring knowledge or observation, and prefer not to say only on sensitive items. On a straightforward satisfaction question about something the respondent has just used, none of the three is needed. Place whatever you do offer at the end of the option list, visually separated, so it does not compete with the scale.

Building it in a form

On a radio field, the escape option is simply the last option in the list. In a matrix the usual build is an extra column at the right, remembering that a matrix is rows by columns with radio style allowing one answer per row. Because dropdown, radio and checkbox fields have no working other option, an answer that needs explaining takes a separate short text field after the question. A matrix exports as one column with each statement and its answer joined as Statement: Answer, so a not applicable answer arrives as text and has to be excluded from the base when the column is split.

Cleaning the data afterwards

Whatever escape options you offer, the analysis has to handle them explicitly rather than letting a spreadsheet decide. The common accident is that a text answer such as Not applicable is read as a category and quietly included in an average, or excluded in one chart and included in another, so two numbers in the same report disagree. Set the rule before the first chart is built: which options are excluded from the base, which are reported separately, and what the minimum base is for showing a percentage. Write the rule into the report itself so a reader can see it. Where the export puts a matrix into one column with each statement and its answer joined, the split step is where these values are easiest to mishandle, so check a handful of rows by hand after splitting rather than trusting the first pass.

Questions people ask

Should not applicable count as a neutral answer?

No. Coding it as the midpoint pulls every average towards the middle and hides the fact that the item did not apply. Exclude it from the base for that item and report how many people chose it, which tells the reader how much of the sample the result actually describes.

Where should the option sit in the list?

At the end, after the scale, and visually separated so it does not read as another scale point. Putting it first or embedding it in the middle of a scale invites people to select it by accident and destroys the ordering that makes the scale interpretable.

Does offering an escape option lower data quality?

Offered selectively, it improves quality by removing guesses. Offered on every question, it lowers engagement and shrinks the usable sample. The test is whether a real respondent could honestly be unable to answer the item, which is true far less often than survey drafts assume.

How do I report an item with a small base?

State the base next to the result and suppress the percentage entirely below a threshold you set in advance, commonly around thirty responses for a percentage and higher for a breakdown. Deciding the rule before seeing the data stops it looking like a convenient way to hide an unwelcome number.

Is prefer not to say the same as leaving it blank?

No, and the difference is useful. A blank could mean the person skipped the question, lost interest or missed it. Prefer not to say is a deliberate choice, and recording it separately tells you that the question was read and declined, which is a much clearer signal about the question.

Should these options appear in a matrix?

They can, as an extra column at the right, and it works well where several statements might not apply to everyone. The caution is width: an extra column on a phone makes a grid harder to read, so check the block on a small screen and consider splitting the matrix if it no longer fits.

Make one with forms

The button opens the generator with this use case already described. Change the wording to match your own.

Create a form with OneCraft

Sources

Written and checked by the OneCraft team. Last checked 12 September 2026.