An open-ended question is only as good as the answers it produces, and most of what goes wrong in a verbatim file was decided when the question was written. A vague question gives you "good" and "ok" in four hundred variations; a question that asks for two things gives you answers you have to split before you can count them.
This guide is about the writing side: how to word an open-ended question so the answers can be coded later, which kinds earn their place in the questionnaire, and what the person coding will thank you for.
What an open-ended question is, and when to use one
An open-ended question has no answer options: the respondent types or says what they think. Use one when you do not know the answer list in advance (reasons, problems, unprompted brand mentions), or when a closed list would put words in their mouth. If you already know the six possible answers, a closed question is cheaper to field and to analyze.
Wording that gives you codable answers
- Ask one thing. "What did you like and what should we improve?" produces answers that mix both. Split it into two questions, or accept that every response will need two codes.
- Anchor it to an experience. "What was the main reason for your score?" beats "Any comments?" because it tells the respondent what kind of answer you want.
- Avoid leading words. "What did you love about our fast delivery?" has already told them the answer.
- Name the time frame. "In your last purchase" gets specifics; "in general" gets generalities.
- Say how much to write. A short instruction ("a sentence or two is fine") reduces both one-word answers and essays.
Examples: weak and stronger versions
| Weak | Stronger | Why |
|---|---|---|
| Any comments? | What is the main reason for the score you gave? | Tells the respondent what to talk about |
| What do you think of the product? | What, if anything, would make you stop using this product? | Asks for one concrete thing |
| What did you like and dislike? | What is the one thing we should improve first? | One idea per answer |
| Which brands come to mind? | When you think of a quick lunch, which brand names come to mind? (list as many as you like) | Defines the context for an unprompted mention |
How many open-ended questions to include
Fewer than you want. Each one costs the respondent effort and the analyst time, and the quality of answers drops as the questionnaire gets longer. A common pattern is one open end after the key closed metric (the "why" behind an NPS or satisfaction score) and one or two at the end for what was missed. Put them where the respondent already has the topic in mind.
What the coder needs from the question
Coding is where wording decisions show up. Three things make it easier:
- The question text travels with the data. A codebook built from the responses alone cannot tell "price" as a complaint from "price" as a reason to buy. The question tells it which one it is. In Survey Coder Pro the question text is what the codebook generation starts from.
- Decide single or multiple codes up front. If a response can legitimately carry two ideas, the codebook has to allow multiple codes per response, and your tables will then add up to more than 100%. See our guide to cross-tabulation for how to read that.
- Write down the rules you already know. If "delivery" and "shipping cost" should be separate codes in this study, say so before coding. In Survey Coder Pro that goes in the question's AI instructions field, and it is injected into the coding prompt.
What happens to the answers
A typical workflow: generate a codebook from the real responses, review and edit it, code every response against it, and then review what the model was unsure about or could not classify. Responses that do not fit any code are a signal about the question as much as about the answers — if a third of them are off-topic, the wording is probably too open. For the full process see the survey coding guide, and for what each code entry should contain see our codebook example.
Frequently asked questions
What is an example of an open-ended survey question?
"What is the main reason for the score you gave us?" asked right after a 0-10 recommendation question. It is specific, it asks for one thing, and it is tied to something the respondent just did.
How many open-ended questions should a survey have?
As few as the objective allows. One "why" after the main metric is the common minimum; more than three or four in a general-population survey usually lowers answer quality.
Should open-ended questions be required?
Usually not. Forcing an answer produces filler ("n/a", "nothing", random characters) that you pay to code. Make them optional, or required only where the answer is the point of the survey.
How do you analyze open-ended survey questions?
By coding: assigning each response to one or more categories in a codebook, then counting. The count is what lets you compare groups and report percentages instead of quotes.
Next steps
If you have a file of open-ended answers and want to see how they code, the free coding tool takes up to 100 responses, proposes a codebook from the actual text and returns an Excel, with no account needed.