If your final year project is survey-based, the questionnaire is your data-collection instrument, and no analysis in Chapter 4 can rescue data that was collected badly. Vague wording, the wrong sample or an uneven response scale show up as unusable data weeks after distribution, when going back is expensive.
This guide walks you through designing one that holds up in front of your supervisor and your defence panel, from objectives to pilot test.
A questionnaire needs a defensible design before its results become a chapter. UniDraft helps organise the instrument details, actual methods and findings you confirm into an editable project document. Start your project document when that material is ready; you remain responsible for designing the questionnaire, recruiting participants and analysing genuine responses.
The short answer
- Start from your objectives. Every item must trace to a research objective or hypothesis in Chapter 1. If it does not, cut it.
- Choose item formats and order for the question, rather than forcing every construct onto one agreement scale.
- Check burden. Use the items your question and measurement rationale need, and pilot the actual form.
- Size the sample with a stated method appropriate to your design and show the working.
- Get it reviewed, then pilot it, before it reaches your real respondents.
- Ask one clear thing per question. Check leading wording and double-barrelled items.
Start from your research objectives, not from a template
Every question on your questionnaire should trace back to a specific research objective or hypothesis in your Chapter 1. If you cannot say which objective a question answers, cut it. This is the fastest way to avoid the bloated, copy-pasted questionnaires that supervisors flag on sight. If your objectives are still loose, fix them first with how to write strong research questions and objectives, because a vague objective produces vague questions. If you have not settled on a topic yet, the free topic generator can give you starting ideas.
A practical structure:
- Section A: respondent demographics. Age, gender, level or department, or whatever variables you will actually use in your analysis. Collect only demographic information justified by the study and appropriate data-handling process; cross-tabulation is not its only possible use.
- Section B onward: one section per objective. Group items by the objective or hypothesis they test, and label each section accordingly.
Build an objective-to-item map first
Before writing a single item, make a small table. It is the quickest way to prove coverage to a supervisor and to spot objectives with too few items.
| Objective (Chapter 1) | Variable to measure | Items | Scale |
|---|---|---|---|
| 1. Determine how often students use the library's e-resources | Frequency of use | B1-B4 | Never to Very often |
| 2. Assess how adequate those resources are | Perceived adequacy | C1-C6 | Strongly disagree to Strongly agree |
| 3. Identify barriers to use | Barriers present | D1-D5 | Strongly disagree to Strongly agree |
| 4. Compare use by level | Level (100-400) | A3 | Category |
Illustrative example only. Replace it with your own objectives. Notice that objective 4 needs no new items; it reuses a demographic item. The number of items depends on the construct and measurement rationale. A single factual item can be appropriate; a complex latent construct may require a supported multi-item measure.
Choosing your question types
Closed-ended questions (multiple choice, yes/no, rating scales) are faster to analyse statistically and are what most quantitative final year projects need. Open-ended questions give richer detail but are harder to code and analyse at scale. Include open items only where they serve the question and you have a plan to analyse them; no universal one-comment-item limit is established here.
| Type | Best for | Watch out for |
|---|---|---|
| Multiple choice (single answer) | Facts: level, department, frequency | Overlapping or missing options; add "Other (specify)" when the list may be incomplete |
| Multiple response ("tick all that apply") | Which of several things apply | Harder to analyse; report each option as its own percentage |
| Yes / No | Simple factual screening | Loses nuance; avoid for opinions |
| Likert scale | Attitudes, opinions, perceptions | Keep direction and number of points consistent |
| Ranking | Priorities | Gets tiring with long lists |
| Open-ended | Explanations, suggestions | Needs manual coding; use sparingly |
An agreement item pairs a statement with ordered response categories. A set of items may form a scale when its measurement and scoring are justified. The software name does not determine whether means, category counts or another analysis is appropriate. Two rules matter more than the number of points:
- Keep the scale consistent within a section. Do not switch from five points to four halfway through.
- Keep the direction consistent, with agree always scoring higher, unless you deliberately include reverse-worded items and reverse-score them before analysis. Decide this before distribution, not while you are cleaning data.
Whether to include a middle "Neutral" option is a design choice. Distinguish a neutral opinion from not knowing, not having experience or not wishing to answer where those responses are relevant. Ask your supervisor which convention your department follows and state your choice in Chapter 3.
How many questions is enough
Use the number needed for the constructs, question and confirmed brief. Inspect duplicates and unnecessary items regardless of the total. Pilot the actual form to check comprehension and burden; no universal item count or completion time is established here.
Questionnaire or interview?
Use a questionnaire when you need measurable data from a reasonably large number of respondents, to compare groups, test a hypothesis, or report percentages and means. Use an interview when your objective needs depth: opinions, reasoning or context that a tick-box cannot capture, and your sample is small enough to speak to individually. Projects that need both breadth and depth sometimes combine a questionnaire with a few interviews to explain unexpected results, but check what your department expects first, since some permit only one instrument type. Your choice belongs in your methodology chapter, justified against your objectives.
Determining your sample size (with a worked example)
If your target population is known and finite, one simplified calculation discussed in the cited IFAS guide is:
n = N / (1 + N·e²)
Where n is the sample size, N is the population size and e is the margin of error (commonly 0.05 for 5%). The formula is a simplified version of a standard sample-size equation and, as the University of Florida's IFAS Extension guide on sample size notes, it assumes a 95% confidence level and P = 0.5 (the most variable case).
Fictional arithmetic exercise, not a sample-size recommendation. Suppose a justified simple-random survey plan has a population of 500 and uses e = 0.05 under the formula’s stated assumptions:
n = 500 / (1 + 500 × 0.05²)
n = 500 / (1 + 500 × 0.0025)
n = 500 / (1 + 1.25)
n = 500 / 2.25
n = 222.2 → 223 (round up)
The arithmetic gives 223 after rounding up under those assumptions; this does not settle your study’s sample size. Design, intended analysis, effect or precision requirements and non-response may call for a different calculation. State the actual rationale in your methodology. "A sample was drawn using Yamane's formula" without the numbers is exactly the vague claim that draws a query at defence.
Here is how the same formula behaves for other population sizes (rounded up, e = 0.05):
| Population (N) | Sample (n) |
|---|---|
| 100 | 80 |
| 200 | 134 |
| 300 | 172 |
| 1,000 | 286 |
| 2,000 | 334 |
| 5,000 | 371 |
Two things follow. First, the sample grows slowly once the population is large, so a fixed percentage of everyone is not a general sample-size rule. Second, in a small population the required sample is a big share of it, so a census (asking everybody) may be more sensible.
What the formula does not do. Under an appropriate design and assumptions, it provides a planning calculation for usable responses, not a universal requirement or the number of invitations. If you expect some people not to return theirs, distribute more and record how many came back. It also says nothing about who to ask: you still need a sampling technique (simple random, stratified by level, and so on) that your supervisor accepts. The full logic, including when Yamane is not the best fit, is in how to calculate sample size for a final year project.
Validity and reliability: what your supervisor is checking
Validity asks whether your questionnaire measures what it claims to measure:
- Face validity: does the instrument look, on the surface, like it measures the right thing? Usually confirmed by your supervisor or a knowledgeable colleague before distribution.
- Content validity: does the full set of items cover every aspect of what you are studying, not just part of it? Your objective-to-item map is the evidence here. If you adapt items from a published scale, cite the original with the citation finder and say what you changed.
Reliability asks whether the instrument produces consistent results. For a group of items meant to measure one construct, your department may ask for an internal-consistency statistic such as Cronbach's alpha. Reviews of the statistic (see Tavakol and Dennick in the sources) note that reports of acceptable values range from about 0.70 to 0.95, that alpha depends on the number of items and on whether they measure one dimension, and that a very high value can mean items are redundant. They also make the key point: an instrument cannot be valid unless it is reliable, but a reliable one is not automatically valid.
So do not compute one alpha across unrelated questions, and do not present a single cutoff as proof your instrument is good. Compute alpha per construct, and only if your supervisor expects it. The full review, pilot and reporting workflow is in the dedicated validity and reliability guide.
Pilot testing before you distribute
When your design allows a pilot, test the draft with people similar to your target respondents before full distribution. Agree the number and recruitment approach with your supervisor. Then:
- Ask participants to complete it while you watch, and note where they hesitate or ask what something means.
- Ask them afterwards what each confusing item made them think. If someone says "well, it depends which part you mean", you may have a double-barrelled question to split.
- Record completion time and burden. Revise unnecessary or confusing items while preserving the measurement requirements.
- Fix, then have your supervisor sign off the final version.
Keep pilot responses out of your final sample if your approved method requires it, and report the pilot briefly in Chapter 3: who took part, and what you changed.
Question order, layout and length
Order changes answers. Earlier questions can colour how people answer later ones, an effect documented by survey researchers at Pew Research Center, so put general questions before specific ones and keep sensitive items away from the start. A workable flow:
- A short introduction: who you are, your department, the purpose, that participation is voluntary and the actual privacy arrangements, accurately distinguishing anonymity from confidentiality.
- Easy demographic items.
- Objective sections, ordered logically, each with a heading and a one-line instruction.
- An optional open comment at the end.
Use plain language your respondents actually use, not academic vocabulary, and let the literature review guide which constructs and wording are standard in your field. If you are surveying students or the public, avoid unexplained abbreviations. Give every option a clear label rather than only numbers, and make sure choice lists do not overlap (for age: 18-22, 23-27, not 18-22, 22-27).
Consent, privacy and ethics
State the purpose and voluntary nature of participation, and confirm the required review and data-handling process before collection. Avoid names, phone numbers or matric numbers that the study does not need. The research-ethics preparation checklist separates institutional review, access permission, participant consent and sharing arrangements. Consult the relevant institutional process and regulator guidance for applicable requirements; do not treat a generic consent statement as complete legal or ethical clearance. Describe only what you actually did in Chapter 3.
Distributing and getting enough responses
- Paper on campus and online invitations can reach different people; this guide establishes no response-rate advantage for either. Count how many you issue and how many come back.
- Online forms can simplify collection and export, but they do not clean or validate the data for you. Check access, coding and whether your sampling reaches the population you defined, not just your friends.
- Plan for non-response. Issue more than your calculated sample, follow up politely, and report the response rate.
- Check as you go. Look at the first batch of returns for blanks and patterns like everyone ticking the same option, which may need investigation under pre-agreed cleaning rules. A repeated response pattern alone does not prove inattention and should not trigger automatic deletion.
Separate questionnaire design, delivery and analysis
A valid-looking form can still collect the wrong information. Check the approved instrument first, pilot the actual delivery next, and preserve the raw responses before analysis. If collecting online, the Google Forms project guide covers respondent access, privacy settings, test submissions and exports. It does not make a convenience sample representative.
Before calculating scores, use the coding and cleaning guide to document response labels, missing values and justified corrections. Do not treat a blank as neutral, delete an inconvenient participant or assume automatic charts have resolved your analysis decisions. Keep a record of what changed between the raw export and the file used for Chapter 4.
Mistakes that quietly ruin a questionnaire
- Leading questions that nudge an answer ("Don't you agree that online learning is more effective?" instead of "How effective do you find online learning?").
- Double-barrelled questions that ask two things at once and force one answer to cover both.
- Jargon your respondents may not share.
- Unclear response scales: vague options like "sometimes" without defined meaning produce data you cannot analyse.
- Inconsistent scale direction: flipping between "agree = 5" and "agree = 1" across sections corrupts scoring.
- Missing or overlapping options in multiple-choice items, so a respondent has no honest answer.
- Items with no objective, collected "just in case" and never analysed.
A before/after example
Before (leading and double-barrelled): "Don't you think the library's poor internet and outdated books make research difficult for students?"
This is leading (it assumes a negative view) and double-barrelled (internet and books are two separate issues in one answer).
After (neutral, single concept, scaled): "The library's internet access is adequate for my research needs." Strongly Disagree / Disagree / Neutral / Agree / Strongly Agree
"The library's book collection is adequate for my research needs." Strongly Disagree / Disagree / Neutral / Agree / Strongly Agree
Splitting the issues avoids forcing one answer to cover two potentially different views; the resulting items still require review and piloting. Pew Research's guidance on writing survey questions makes the same points: one issue per question, and wording that does not signal the "right" answer.
Design the questionnaire with the analysis in mind
Before you print, ask how each item will appear in Chapter 4. Possible outputs include category counts, appropriate scale summaries or a justified comparison. An individual ordered item does not automatically justify treating its categories as equal numerical intervals. If you cannot picture the table, the item is probably not needed. Also decide how you will code responses (for example, Strongly Disagree = 1 to Strongly Agree = 5) and how you will interpret mean scores; justify any scoring or classification rule and explain its limitations; a numerical midpoint alone does not validate a substantive interpretation. The choice of test comes next, and which statistical test fits your project depends on how you measured each variable. The tables themselves are covered in writing the results chapter.
Pre-distribution checklist
- Every item maps to an objective in your map.
- Demographic items are only those you will analyse.
- One idea per question; no leading or loaded wording.
- Scale points, labels and direction are consistent.
- Multiple-choice options do not overlap and include "Other" where needed.
- Introduction states purpose, voluntariness and confidentiality.
- Sample size, population and method are written down with the working.
- Supervisor has reviewed it; pilot done and changes noted.
- You know how many copies to issue and how you will follow up.
Where this fits in your project
A well-designed questionnaire is what your methodology chapter describes: the instrument, the sample and how you sized it, and your validity and reliability checks. If you want the whole path from topic to submission, how to write a final year project step by step puts this stage in context.
Where UniDraft fits in
UniDraft does not design or distribute your questionnaire, and it never invents survey data or results; the instrument and its responses have to come from you. It helps organise the written document around your confirmed topic, methods and real results; you verify the instrument description, calculation, evidence and consistency yourself. You can upload your results, and if your supervisor asks for changes, check current included correction rounds before checkout. Before submission, the free formatting checker tests your .docx against your university's rules. Check current document pricing before checkout.
Sources
- Determining Sample Size, University of Florida IFAS Extension: gives Yamane's simplified formula and its 95% confidence, P = 0.5 assumptions.
- Making sense of Cronbach's alpha, Tavakol and Dennick (International Journal of Medical Education, via PubMed Central): acceptable-value ranges, dependence on item count, and the reliability-validity relationship.
- Writing Survey Questions, Pew Research Center: double-barrelled questions, biased wording and question-order effects.
- Our Data Privacy Policy, Nigeria Data Protection Commission: the regulator's statement on the Nigeria Data Protection Act 2023 and the role of consent.