Is Other, please specify a partially open format?

Yes, when it combines listed choices with a genuine free-response opportunity.

Must every write-in become a new category?

No; some fit existing categories or remain ambiguous.

Can recoding change the headline counts?

Yes, which is why both the rule and the original responses should be preserved.

AI Glossary / Statistics and probability / Intermediate

Partially open question

Definition

A partially open question is a survey question that combines structured response options with an opportunity to answer in the respondent's own words. A common form offers predefined choices plus Other, please specify. It aims to retain comparable categories while allowing responses that the question designer did not anticipate.

What it means

A fully closed question restricts respondents to listed answers. A fully open question asks them to formulate their own response. A partially open format sits between these approaches by providing a framework and a route beyond it.
The text option can reveal missing categories, misunderstood wording, or experiences that do not fit the list. It does not remove the influence of the listed options. Respondents may still be prompted by those choices and may be less likely to formulate a different answer.

How it works

Define the concept and decide whether respondents should choose one answer or several. Use clear instructions, mutually exclusive choices where appropriate, and a response list based on prior research or piloting.
Provide an explicit text field associated with the alternative response. Keep Other distinct from don't know, not applicable, and prefer not to answer when these meanings matter. A respondent who selects Other but leaves the text blank has still supplied information about the inadequacy of the listed choices.
Test the format with intended respondents. Check whether the text field is visible, usable on different devices, and interpreted as intended. A hidden or burdensome write-in option may function poorly despite being technically available.
Create a coding plan before interpreting the text. Some write-ins may fit existing categories, while others justify new categories or remain unclear. Preserve the original selection and verbatim response alongside any recoded value.
Document coding decisions and allow uncertain cases to be reviewed. AI can suggest labels or clusters, but its classifications should be checked against a clear codebook. Do not invent a missing explanation or silently replace respondents' meanings.
Report denominators carefully. The share selecting Other, the share providing usable text, and the share assigned to a newly identified theme describe different quantities. For multiple-response questions, percentages can exceed 100% across categories.
If the response list changes in later surveys, assess comparability. Adding a newly discovered category can increase its selection by making it visible, so differences over time may partly reflect the changed measurement instrument.

Three practical examples

Product feedback: a survey lists common reasons for not using an AI feature and includes an option to explain another reason.
Service research: customers select their main contact purpose, with a write-in field for purposes outside the current taxonomy.
Workforce planning: employees choose a preferred training format and can describe an alternative arrangement that the organizers did not list.

Think of it this way

Think of a form with a set of labeled drawers and an extra space for items that do not fit. The extra space can reveal that a drawer is missing, but the existing labels still influence where people place their items.

Worked example

Suppose a fictional single-choice survey asks 100 respondents for their main barrier to using an AI tool. Initial selections are cost 40, training 25, access 15, and Other 20. These counts sum to 100.
Of the 20 Other selections, 15 include usable text and 5 leave the explanation blank. Reviewers find that 3 usable responses clearly describe cost, 2 describe training, and 10 identify an unlisted integration problem.
Under a documented recoding rule, the revised analytical counts are cost 43, training 27, access 15, integration 10, and Other with missing detail 5. They still sum to 100. The original selections and text are retained separately.
Integration represents 10/100 = 10% of all respondents in this coded dataset. It represents 10/15, approximately 66.67%, of usable Other explanations. These denominators answer different questions.
The calculation does not show that only 10% of all users have integration problems. The question asked for one main barrier, and the survey's sampling and response process would determine any population interpretation. The five blank explanations are not evidence of no barrier.

Illustrative case study

Illustrative fictional case: Fern Support includes an Other field in a customer survey but finds that many respondents use it to describe problems already listed. Cognitive interviews reveal that the original category labels use internal terminology unfamiliar to customers.
The team rewrites the choices, retains a write-in option, and tests the revised question. It flags the wording change when comparing later results with the original survey. This fictional case illustrates how free text can diagnose a measurement problem rather than simply supply additional categories.

Common misconceptions and mistakes

Do not assume an Other option makes the question fully open or unbiased. Do not treat blank write-ins as no opinion or no problem. Avoid merging don't know, refusal, and unlisted substantive answers into one undifferentiated category.
Do not silently overwrite raw responses during coding. Do not compare a prompted category in one survey with an unprompted write-in theme in another as though the measurement conditions were identical. AI-generated summaries also need checks for omitted or misclassified responses.

Business use

Partially open questions help businesses balance standardized reporting with discovery. They can improve category design, reveal emerging needs, and make surveys less restrictive. The benefit depends on reviewing the text and maintaining a transparent coding process, rather than collecting comments that are never examined or treating automatic labels as unquestionable facts.

Check your understanding

How many respondents selected Other initially?

Answer: Twenty.

How many supplied usable explanations?

Answer: Fifteen.

Why are 10% and 66.67% both correct for integration?

Answer: One uses all respondents as the denominator, while the other uses usable Other explanations.

Related terms

Open-ended question
closed-ended question
survey design
response option
coding frame
item nonresponse
cognitive interviewing
questionnaire testing

Learn next

Next study coding reliability and question pretesting. Then examine how prompted choices, response order, missing text, and changes in category wording affect interpretation and comparison over time.

Sources

NIST/SEMATECH, Engineering Statistics Handbook glossary, partially open question: https://www.itl.nist.gov/div898/handbook/glossary.htm — supplied hybrid-format definition. Pew Research Center, Writing Survey Questions: https://www.pewresearch.org/writing-survey-questions/ — primary survey-method guidance on response formats, offered choices, wording, and pretesting. The recoding example and fictional case are original.

Previous: Partialing Next: Particle