A questionnaire looks like the easy part of a dissertation. You write some questions, send them out, and the data comes back ready to analyse. In practice, a questionnaire is one of the easiest places to lose marks, because every flaw in its design is frozen into the data before a single response arrives. A confusing question does not produce a confusing answer you can tidy up later; it produces an answer you cannot trust and cannot interpret. A missing response option does not create a small gap; it creates a systematic bias in everything that follows. That is why supervisors scrutinise questionnaires so closely, and why time spent designing yours properly is repaid in a methodology chapter you can defend and results you can actually use.
This guide walks through designing a questionnaire for a dissertation step by step: defining what you need to measure, choosing question types, writing items that are clear and unbiased, building a scale that works, piloting the instrument, and handling ethics and consent. It is written for primary research at undergraduate and taught-postgraduate level, where questionnaires are most commonly used, but the principles apply wherever survey data feeds into a dissertation.
Start with what you need to measure, not with questions
The most common questionnaire mistake is starting with a list of interesting questions. Start instead with your research questions and objectives, and work out what each one requires you to measure. If your research question asks how students experience online feedback, you need measures of experience, of specific feedback practices, and perhaps of satisfaction and perceived usefulness. Each measure then becomes a small section of the questionnaire. If a proposed question does not serve any research question, cut it. Long questionnaires do not look thorough; they produce fatigue, rushed answers, and drop-outs, and markers notice when a results chapter reports data that answers nothing the dissertation asked.
Validity and reliability: the two tests your questionnaire must pass
Validity asks whether your questionnaire measures what it claims to measure. A question about “satisfaction with feedback” that actually measures politeness towards the tutor is not valid, however neatly the responses chart. Reliability asks whether it would produce consistent results if repeated under similar conditions. A question that different respondents interpret in different ways is not reliable, because its answers reflect interpretation rather than the thing being measured.
Three practical habits protect validity. First, use established wording where it exists: if a validated scale measures your concept, adapting it with acknowledgement is usually stronger than inventing items from scratch, and your methodology chapter can cite its prior use. Second, match each item to one concept only. A question that asks about two things at once — “How satisfied are you with the quality and speed of feedback?” — forces a respondent who liked the quality but hated the speed to give a single meaningless answer. Third, ask about behaviour and experience before opinion. Concrete questions (“How often did you receive feedback within the stated timeframe?”) anchor responses in something real; abstract opinion questions on their own invite guesswork.
Choosing question types
Closed questions
Closed questions offer fixed response options: yes or no, a list of categories, a multiple-choice set, or a scale. They are quick to answer, easy to analyse, and comparable across respondents, which is why they carry most dissertation questionnaires. Their weakness is that they constrain what respondents can say, so they work best where you already know the range of likely answers. Always check that your options are exhaustive and mutually exclusive: every respondent should find an option that fits, and no respondent should fit two at once. An “Other (please specify)” option is a useful safety net, not a substitute for thinking through the categories properly.
Open questions
Open questions let respondents answer in their own words. They are valuable for exploring experiences you cannot predict, for following up a closed question (“Please explain your answer”), and for capturing language that enriches your discussion. They are slower to answer and much slower to analyse, because free text must be coded before it can be summarised. Use them sparingly and purposefully: one or two well-placed open questions at the end of a section, or a single open question at the end of the questionnaire, usually yields more usable material than a questionnaire full of blank boxes that respondents leave empty.
Demographic and filter questions
Designing Likert scales that actually work
The Likert scale is the workhorse of dissertation questionnaires, and it is also the format most often built badly. A Likert item presents a statement and asks respondents to rate their agreement, typically on five points: strongly disagree, disagree, neither agree nor disagree, agree, strongly agree. Five points suit most dissertation purposes; seven points give finer discrimination but demand more of respondents and are harder to justify with a small sample.
Keep the scale consistent across a section: do not switch from agreement to frequency to satisfaction halfway through, and do not reverse the direction of the scale without a clear reason. If you do include reverse-worded items to guard against automatic agreement, flag them clearly in your analysis plan, because they must be reverse-scored before any totals are calculated, and forgetting that step is a classic analysis error. Label every point, not just the ends: respondents treat an unlabelled middle point inconsistently, and your methodology chapter should be able to state exactly what each point meant.
Writing bias-free questions
Bias in a questionnaire rarely announces itself. It creeps in through wording that pushes respondents towards one answer, and markers are trained to spot it. The most common forms are worth learning by name, because naming them in your methodology chapter — and showing you avoided them — earns marks.
Leading questions suggest the answer: “How much did you enjoy the excellent teaching on this module?” Rewrite it neutrally: “How would you rate the teaching on this module?” Loaded questions carry an assumption the respondent may not share: “Why do you think students fail to attend lectures?” assumes non-attendance is the student’s failure; a neutral version asks about attendance and its reasons separately. Double-barrelled questions ask two things at once, as discussed above, and must be split. Ambiguous wording includes vague quantifiers (“often”, “regularly”, “recently”) that mean different things to different people; replace them with defined periods or frequencies (“in the last four weeks”, “at least once a week”). Social desirability bias is the tendency to give the respectable answer, and it is strongest on sensitive topics; neutral wording, anonymity, and indirect phrasing (“Some students find it difficult to…”) all reduce it.
Structuring the questionnaire
Order matters. Begin with a short introduction that states the purpose of the study, who is conducting it, how long the questionnaire takes, and that participation is voluntary and anonymous where applicable. Follow with easy, non-sensitive questions to build confidence, place the most important questions in the middle where attention is highest, and save sensitive or personal questions for the end, when respondents are already invested. Group related questions under clear section headings, and keep the flow logical: jumping between topics confuses respondents and increases the chance of careless answers.
Piloting: the step you must not skip
Piloting means testing the questionnaire on a small group similar to your target sample before you launch it properly. Ten to twenty participants is a reasonable pilot for most undergraduate dissertations, though your supervisor may advise a different number. The pilot is not a dress rehearsal for data collection; it is a diagnostic exercise, and you should treat it as one.
Ask pilot participants to do three things: complete the questionnaire, note any question they found confusing, ambiguous, or difficult to answer, and tell you how long it took. Then examine the pilot data itself. Are there questions everyone answered the same way, which may mean the item does not discriminate? Are there questions with many missing answers, which may mean the question is unclear, sensitive, or badly placed? Did any response option go unused, or did many respondents choose “Other”, which may mean a category is missing? Are the open responses usable, or did respondents misunderstand what was being asked?
Revise on the basis of the pilot and say so in your methodology chapter. Markers reward a pilot that led to real changes — reworded items, a dropped question, a revised scale — because it shows the instrument was tested rather than merely assembled. A pilot that changed nothing is either a sign of an unusually good first draft or, more often, a sign that the pilot was not examined closely enough. If your changes are substantial, pilot the revised version again with a few fresh participants before launching.
Ethics, consent, and data handling
No questionnaire goes out before ethics approval where your university requires it, and most universities require it for any research involving human participants. Your ethics application will normally need the questionnaire itself, the participant information sheet, and the consent process, so build these early rather than treating ethics as a form to complete after the questionnaire is finished.
Common mark-losing mistakes
Questions that do not map to research questions. Every item should earn its place by serving a stated objective. If you cannot say which objective an item serves, remove it.
Double-barrelled and leading items. These are the two flaws markers find first, and they undermine validity in a way that no analysis can repair.
Scales used inconsistently. Changing scale labels or direction mid-questionnaire, or leaving points unlabelled, produces data that cannot be compared across items.
No pilot, or a pilot reported as a formality. A methodology chapter that says “the questionnaire was piloted” without saying what changed is a missed opportunity to demonstrate rigour.
Ethics treated as an afterthought. Collecting data before approval, or collecting identifiable data without a stated reason, can invalidate a whole study, however good the questionnaire itself is.
Frequently asked questions
How many questions should a dissertation questionnaire have? There is no fixed number. Aim for the smallest set that answers your research questions properly, usually completable in 10 to 15 minutes. Quality and focus matter more than length.
Should I use a validated scale or write my own questions? Use a validated scale where one fits your concept, with acknowledgement and any necessary permission. Write your own items only where no suitable scale exists, and pilot them carefully.
How many people should I pilot with? Ten to twenty participants similar to your target sample is reasonable for most undergraduate dissertations, but follow your supervisor’s advice and your ethics approval.
Five-point or seven-point Likert scale? Five points suit most dissertation purposes and are easier for respondents. Seven points give finer discrimination but need a larger sample to justify, and every point should be labelled either way.
Do I need ethics approval for an anonymous questionnaire? In most UK universities, yes. Any research involving human participants normally requires ethics approval before data collection begins, even when responses are anonymous. Check your institution’s rules early.
What do I do if my pilot shows a question is broken? Fix it, and report the change in your methodology chapter. That is exactly what piloting is for, and a documented revision strengthens your work rather than weakening it.
Closing guidance
A good questionnaire is designed backwards: from research questions, to concepts, to items, to a pilot, to a launch. Work in that order and you will produce data you can analyse with confidence and a methodology chapter you can defend without apology. Work in the opposite order — writing questions first and hoping the analysis sorts itself out — and you will discover the flaws only when it is too late to fix them. Draft your questionnaire, map every item to an objective, pilot it properly, and revise it honestly. Your results chapter will be far easier to write, and your marker will be able to see, at every stage, that the instrument was built with care. If you would like structured guidance on your questionnaire design, from mapping items to objectives through to a pilot-ready draft, our dissertation support team can help you work through it step by step.