Open-ended question

What open questions are good for

Three jobs, and they are all discovery rather than measurement.

• Finding the option you did not think of. Run the open version first, read the answers, and build your closed question from what people actually said.
• Explaining a number. A score tells you the temperature. The free text next to it tells you why, and it is the part that gets quoted in the report and read by the people who can change something.
• Giving people somewhere to put the thing that is bothering them. A single "anything else?" box at the end takes the edge off the sense of being processed, and occasionally contains the only important finding in the survey.

The cost

Open text is expensive at both ends. For the respondent, typing costs far more effort than tapping, and the first free-text box is one of the most reliable drop-off points in any survey — see completion rate. For you, 800 free-text answers have to be read and coded into themes before they can be counted, and coding is slow, subjective work that needs at least a sample checked by a second person.

Budget accordingly. Two well-placed open questions in a survey you will actually read are worth more than eight you will skim.

A worked example

Weak: "Do you have any comments about our services?" — attracts either nothing or an unrelated complaint, because it gives the respondent no frame.

Better: "You said you were dissatisfied with bin collections. What went wrong?" — narrow, anchored to something they just said, and only shown to the people it applies to. Built with conditional logic to trigger on the dissatisfied answers, and piping to name the service, it asks perhaps 15% of respondents one extra question and returns the most useful text in the survey.

Common mistakes

• Using open text because you have not decided what you are asking. "Tell us your thoughts on the proposal" outsources your analysis problem to 2,000 strangers.
• Making it mandatory. A required free-text box produces "n/a", ".", and abandonment. Required open questions are almost never justified.
• Counting words instead of reading them. A word cloud is not analysis. It reports which nouns are common, which is usually the subject of your own question.
• Quoting the vivid one. The most quotable comment is rarely the most representative. Code first, quote second.
• Forgetting it is personal data. People put names, addresses and medical detail in free-text boxes regardless of what you asked for. That has consequences for anonymity and for what you can publish.

In NumoForms

Two question types cover open text: Short text for a single line and Paragraph for multi-line. Both support a minimum and maximum length with a live character counter, which is a gentler way of signalling "a sentence, not an essay" than a rejection message after the fact.

A middle option worth knowing: "Other (please specify)" on a choice question. It keeps the structure of a closed question and gives the minority who do not fit somewhere honest to go. The free text is stored separately from the option label and gets its own column in the CSV export, so it never contaminates your counts. It is also kept only while the question is visible and the option is still selected — change your mind and the stray text goes with it.

Results show live charts per question and let you browse individual responses one at a time, which is how free text is best read. There is no automated analysis of open text in the product.