Survey Question Examples: Segment, Buyer, and Usage Research
A survey question is a translation: it turns a research construct into words and response options a respondent can actually understand and answer. Examples are useful when they preserve that logic, not when they are copied as ready-made truth. Segment attributes, buying process, observed usage, and explanation need different questions, and no polished wording can rescue the wrong population or an undefined construct.
There is no formula that makes a survey question valid. Measurement quality depends on the construct, sample, wording, recall period, options, order, mode, and pretest. Arithmetic cannot repair an ambiguous, leading, double-barrelled, or poorly sampled item.
A survey question is one item. A questionnaire is the ordered instrument: stems, response options, instructions, routing, and surrounding context. Segment questions describe grouping variables; buyer questions examine roles and process; usage questions ask about behavior in a bounded period. They can live in one questionnaire, but they do not become one construct.
Before the wording, settle what the answer must represent
Before writing, complete this sentence: “We will use the answers to decide ___ for ___ population.” Then name the evidence unit. Is the respondent answering about themselves, their organization, one buying process, one product account, or one recent task?
That choice prevents a common B2B error: asking an individual to provide authoritative organization-level facts they may not know. “How much did your company spend on this category?” may collect guesses. “Which of these roles did you personally perform in the most recent evaluation?” stays within the respondent’s possible knowledge.
Pew Research Center’s question-design guidance explains that open versus closed format, available options, wording, and order can materially change answers. Closed options must be exhaustive enough for the population and should not overlap.
Examples are starting points, not validated questions. Replace bracketed terms, then test the wording, response options, and recall task with people from the intended population before fielding.
Segment question examples
Segment variables should be chosen because they support a decision, not because they are easy to ask. Prefer observable operating differences over labels that merely sound strategic.
| Research need | Adaptable question | Response design note |
|---|---|---|
| Role context | “Which of the following best describes your role in [process]?” | Use mutually exclusive primary-role options plus “another role” and “not involved.” |
| Organization context | “Approximately how many people currently work at your organization?” | Use non-overlapping bands that match the business decision; include “not sure.” |
| Operating model | “Which option best describes how your team currently completes [task]?” | Derive options from prior interviews or observed workflows, not internal product tiers. |
| Maturity evidence | “Which of these activities has your team completed in the past [bounded period]?” | Ask about observable actions; allow multiple selection only when actions can coexist. |
| Outcome constraint | “Which one factor most limited your team’s ability to complete [task] in the past [period]?” | Randomize nominal options where appropriate and include an open “another factor.” |
Avoid treating a self-selected label such as “advanced” as equivalent to behavior. If maturity matters, ask about the practices, records, or decisions that define it. Keep demographic or sensitive questions only when they are necessary, permitted, and accompanied by an appropriate privacy and use explanation.
Buyer-research question examples
Buyer research is more reliable when it anchors answers to a specific, recent decision rather than a hypothetical future purchase.
| Research need | Adaptable question | Why it is bounded |
|---|---|---|
| Trigger | “Thinking about the most recent time your organization evaluated [category], what event started the evaluation?” | Anchors recall to one process rather than a generic opinion. |
| Personal role | “Which activities did you personally perform during that evaluation?” | Separates respondent behavior from the whole buying group. |
| Participants | “Which functions participated before the decision was made?” | Asks about observed participation, not assumed authority. |
| Considered options | “Which approaches did the team seriously consider?” | Include internal build, process change, postponement, and “not sure” where relevant—not only vendors. |
| Decision criteria | “Which one criterion had the greatest influence on the final decision?” | Forces priority; follow with an open reason rather than a double-barrelled rating. |
| Outcome | “What was the status of that evaluation at the time you last had direct knowledge of it?” | Provides terminal and unknown options without assuming a purchase occurred. |
| Explanation | “What made that factor important in this evaluation?” | Open text captures respondent language after the bounded choice. |
Question order matters. Asking respondents to rate a product’s features before asking what shaped their decision can prime those features. Put unaided recall and broad process questions before detailed lists when discovery is the goal.
Usage-research question examples
Usage questions need a defined product, actor, action, and time window. The UK Government Analysis Function’s questionnaire guidance warns about recall error and recommends clear period boundaries.
| Research need | Adaptable question | Avoid |
|---|---|---|
| Recency | “When did you personally last use [feature]?” | “Do you use [feature] regularly?” |
| Frequency | “During the past [period], on how many days did you personally use [feature]?” | Vague labels without a defined period |
| Task | “The last time you used [feature], what were you trying to complete?” | Assuming the intended use case |
| Completion | “Were you able to complete that task during the same session?” | Combining completion and satisfaction |
| Friction | “What, if anything, prevented you from completing it?” | “Why was the feature difficult?” |
| Alternative | “What did you do next?” | Assuming the user stayed in the product |
“How easy and effective was the workflow?” is double-barrelled: ease and effectiveness can diverge. Ask them separately. “How much do you love the new workflow?” is leading and presupposes use. A neutral item first establishes exposure, then asks for an evaluation only from respondents who can answer.
UK Government Analysis Function notes that Pew and UK government guidance both treat wording and context as measurement issues; the UK guidance specifically separates double-barrelled constructs and bounds recall periods.
Open and closed formats produce different evidence
Closed-ended questions make comparisons and routing easier, but the option list cues respondents. Open-ended questions preserve respondent language and can reveal missing categories, but they increase response burden and require a coding plan.
A productive pattern during development is to begin with interviews or open-ended pilot items, build a response set from the observed language, and then test the closed version. Pew describes this as one way to develop answer choices grounded in how respondents frame an issue. Keep an “other” route when the listed options cannot safely claim exhaustiveness, and review those responses after fielding.
Preflight the questionnaire as an instrument
- Check the construct — For every item, name the single fact, behavior, evaluation, or reason it is meant to measure and the decision that consumes it.
- Check answerability — Confirm the respondent could reasonably know the answer. Add “not sure,” “not applicable,” or a routing exit when uncertainty is legitimate.
- Check wording and options — Remove assumptions, leading language, double barrels, overlap, unexplained terms, and vague recall periods. Preserve an ordered scale’s logical sequence.
- Check order and routing — Put eligibility and unaided recall before prompts that could prime an answer. Test skip logic, back navigation, and treatment of partial responses.
- Test with the population — The Census Bureau lists cognitive testing, focus groups, expert review, behavior coding, and split-ballot experiments among its evaluation methods. Use the smallest appropriate set that exposes interpretation and mode failures before launch.
Report what the instrument can support
A well-designed item still describes the respondents who were reached and chose to answer. Population inference depends on sampling, coverage, nonresponse, weighting, and mode—not wording alone. Preserve the exact question text, option order, routing, field dates, population, and analysis exclusions so another reviewer can understand what changed.
Write each survey question for one construct, one evidence unit, and one declared decision; then test how real respondents interpret it before treating the answers as data.
Frequently asked questions
How should a Likert scale be written?
For a single Likert-style item, ask about one construct and provide ordered, balanced response labels with a neutral midpoint only when neutrality is a legitimate state. Keep the direction and number of points consistent across comparable items; the UK Government Analysis Function illustrates a five-point sequence from “very unlikely” to “very likely,” which is clearer than leaving respondents to interpret unlabeled numbers.
What is the difference between a cognitive interview and a pilot survey?
A cognitive interview is a one-to-one qualitative test that probes how a participant understands a question and reaches an answer; a pilot runs the near-final instrument under conditions closer to fieldwork to expose routing, format, duration, and response-pattern problems. The UK questionnaire-design guidance treats them as distinct stages, so resolve interpretation failures through cognitive interviews before using a pilot to test the whole collection process.
Is an anonymous survey the same as a confidential survey?
An anonymous survey gives the research team no practical way to link a response to a person, while a confidential survey may retain that link under access and disclosure controls. University of Massachusetts Amherst’s survey guidance notes that IP addresses, email invitations, ZIP codes, and combinations of demographic answers can make responses identifiable, so inspect the platform’s metadata collection and deletion settings before promising anonymity.