Your research question is settled and your advisor has signed off on an empirical chapter. Then comes the part that methods courses rarely walk through in practice: how does that question turn into a questionnaire whose answers you can actually analyze?
You run a survey for your bachelor’s or master’s thesis in six steps: derive measurable variables from your research question, choose between a standardized and an open approach, write the questionnaire and its consent page, test both on a few people, collect answers over a fixed data collection period and analyze them afterward. At the end you have a dataset and the five details that belong in your methods section.
📌 At a glance
- Your research question decides the method, not your deadline.
- 385 usable responses cover 95 percent confidence, plus/minus 5 points.
- An IRB can waive the signed consent form.
- The consent page names purpose, length and data handling.
- Collection dates, sample and drop-outs go in the methods section.
Create a survey for free
With empirio.ai you can create a modern online survey in minutes — free to start.
- AI-built survey
- Adjust by drag & drop
- Real-time analysis
When a survey fits your thesis
A survey fits your thesis when your research question asks about attitudes, judgments or self-reported behavior, and when those answers cannot be pulled out of existing statistics, records or documents. If your work aims at numbers the U.S. Census Bureau or the Bureau of Labor Statistics already publish, or at datasets ICPSR already archives, your own survey is effort without return.
The second decision is about the form of the survey and your research question governs that too. If you want to know how widespread something is, or whether two variables hang together, you need many comparable answers and therefore a standardized survey. If you want to understand why people act the way they do, you need a few long conversations. Doing both is possible and costs twice the time.
The table sorts the three cases that come up most often in student work.
| Your research question aims at | Suitable approach | What you need for it |
|---|---|---|
| Prevalence and relationships | standardized online survey | closed questions and a target number |
| Reasons, meanings, experiences | semi-structured interview | a few interviews and an analysis method |
| Numbers already collected | secondary analysis | access to the dataset, no data collection |
What your paper is called depends on the degree. The Cornell Graduate School uses thesis for master’s candidates and dissertation for doctoral candidates, while undergraduate work usually goes by senior thesis or capstone project. The requirements for the empirical part do not change with the name.
How the two approaches differ otherwise is set out in the comparison of qualitative and quantitative surveys. And whether your paper needs an empirical chapter at all is the question before that: the guide to empirical research in a final paper works through it.
The process: six steps and a timeline built backward
A survey for a thesis follows six steps and the order is not a matter of taste. Each step settles what is even possible in the next one. Anyone who writes the questionnaire before the research question notices it at the analysis stage at the latest.
- Sharpen the research question. From it you derive the variables you want to measure.
- Turn variables into questions. Each variable gets one question and one answer format. How to go about it is set out in the guide to questionnaire design.
- Write the questionnaire and the consent page. Both belong together, because the consent page decides your response rate.
- Run a pilot. Five people from your target group fill it out and tell you where they got stuck.
- Fix the collection period and gather answers. Start and end dates are set in advance and written down.
- Analyze and describe. First the numbers, then the text in your methods and results sections.
Build the timeline backward, because data collection depends on other people and cannot be compressed. Take your submission date, subtract the weeks for analysis and writing and put the collection period in front of that. For a term paper, three to four weeks for the whole collection is a realistic allowance, for a thesis with a larger target sample closer to six.
Approval belongs in the timeline as well. Your advisor normally reads the questionnaire before it goes out, and a review by the Institutional Review Board (IRB) can come on top. The federal rule, 45 CFR 46, covers research that federal agencies support, and many universities require IRB review regardless. Whether your project needs it is decided by the IRB at your own school, not by you and not by this article.
💡 Tip
Student surveys often fit an exempt category. 45 CFR 46.104(d)(2) covers survey procedures when the information is recorded so that the identity of the participants cannot readily be ascertained. Exempt is a determination, though, and not a shortcut you take yourself: ask the IRB office to confirm it in writing before you collect a single response.
How many participants does your survey need?
The number of participants you need hangs on the population and on the precision you want to reach. At a 95 percent confidence level and a margin of error of plus or minus 5 percentage points, 385 usable responses are enough once the population reaches one million or more. For 10,000 people the number is 370, for 1,000 people 278.
| Population | Usable responses needed | What follows from that |
|---|---|---|
| 1,000 people | 278 | workable if you can reach the whole group |
| 10,000 people | 370 | realistic only through an institution |
| 1 million and above | 385 | the number barely rises beyond this |
The numbers in the table hold for a random sample. Almost nobody writing a thesis has one, because participants decide for themselves whether to take part. A large sample then makes your analysis more precise, but not representative. The difference matters enough for a piece of its own: when a survey counts as representative works it through, and the guide to the sample explains the selection methods behind it.
In qualitative work the number counts differently. Saturation decides there: you keep interviewing until new interviews stop adding anything. How many that takes cannot be fixed in advance and belongs in the methods section as a justification.
Plan for drop-outs and for half-finished questionnaires. If you need 278 usable cases, you do not send the link to 278 people but to a good deal more.
The consent page at the start of your survey
The consent page of your survey names five things: who you are, what the study is about, how long it takes to fill out, what happens to the answers and that taking part is voluntary. It sits on the first page of the questionnaire and it is the point at which most people decide whether to go on.
You often read that every participant has to sign a consent form. That is not what the Common Rule says. 45 CFR 46.116(a)(1) requires that “Before involving a human subject in research covered by this policy, an investigator shall obtain the legally effective informed consent of the subject”. Consent and the documentation of consent are two different things: under 45 CFR 46.117(c) an IRB may waive the signed form where the research “presents no more than minimal risk of harm to subjects and involves no procedures for which written consent is normally required outside of the research context”. An online consent page with an “I agree” button is the usual answer, and your IRB decides whether it is enough.
The second common misunderstanding is about anonymity. If your survey records no identifiers, the exempt category in 45 CFR 46.104(d)(2) is normally the one that applies, because the identity of the participants cannot readily be ascertained. What matters is that the combination of age, major and class year does not point at anyone either. What belongs to that is set out in the piece on the anonymous survey and in the guide to data privacy in surveys.
A consent page for your survey: an example to copy
The example below covers the five points in the order in which they are read. Replace the square brackets with your own details and cut whatever does not apply to your project. Write the short pitch you later post in groups and forums separately, because this text sits inside the questionnaire itself.
Thank you for taking the time. For my master’s thesis at [university] I am studying how students use learning platforms when they study for exams. The questionnaire has 14 questions and takes about four minutes. Your answers are collected anonymously, no names, email addresses or IP addresses are stored, and the analysis only ever looks at all responses together. Taking part is voluntary and you can stop at any time. If you have questions about the study you can reach me at [email address]; questions about your rights as a participant go to the [university] IRB at [contact]. The survey runs until [date].
The time estimate in that text is the point where most is guessed and least is measured. Fill the questionnaire out once yourself with a stopwatch and give the number that comes out. An estimate that is too low costs you half-finished questionnaires later on.
⚠️ Note
As soon as you record identifiers, the consent page changes and so does the review. Names, email addresses and course rosters take your study out of the exempt category, and the IRB then looks at storage, access and retention. The normal case in student surveys is anonymous collection and the exception is the identified one. Your IRB decides which one you are in. This article is not legal advice.
Create a survey for free
With empirio.ai you can create a modern online survey in minutes — free to start.
- AI-built survey
- Adjust by drag & drop
- Real-time analysis
Which survey tool is good enough for a thesis
A survey tool is good enough for a thesis if you can download the raw data in a format you can analyze, if the response limit sits above your target sample and if you know where the provider stores the answers. Everything else is comfort, and those three points decide whether your data can be used at all.
The export is where it fails most often. If you only ever see the answers as charts on a web page, you cannot work out a mean, cannot run a cross-tabulation and cannot document anything in the appendix. So check before you start whether you get a CSV or Excel file, and download it once as a test while there are only three practice responses in the dataset.
| Criterion | What to look for | Why it counts in your thesis |
|---|---|---|
| Raw data export | CSV, Excel or SPSS | no raw data, no analysis and no appendix |
| Response limit | cap per month and per survey | a limit mid-collection costs you cases |
| Storage location | where the servers stand | the IRB application asks for it |
| Written agreement | offered by the provider or not | campus IT reviews it before approval |
Plan on being asked where the responses are stored and who can see them. Whether your school keeps a list of approved tools, or sends a new one through a security review first, is worth asking your advisor or the IRB office before you set the survey up. The provider does not settle that question and neither does this article.
At empirio.ai, an online survey tool from Germany, EU hosting and the data processing agreement belong to the free plan, as do unlimited surveys and questions and 50 participants a month. Downloading the raw data, by contrast, starts with the Lite plan at 19 euros a month (as of September 2026). How the common tools differ otherwise is set out in the comparison of survey tools.
Analyzing the survey and writing the methods section
Five details about the survey belong in the methods section of your thesis: the instrument, the mode of collection, the collection period with start and end dates, the sample size and the sampling method. Drop-outs and incomplete cases come on top, because they are a result too and not a blemish.
Write those five details down while the survey is still running. Anyone who reconstructs them afterward no longer finds the start date and ends up estimating the number of drop-outs. A short note each day is enough for that.
Checklist for the methods section
The list below covers what a thesis committee expects at this point. If a point is missing, what is missing is not a line but the traceability of your data collection. That is exactly what an empirical paper is graded on.
- The questionnaire appears in full in the appendix.
- The mode is named: online, on paper or in person.
- The collection period appears with start and end dates.
- People invited and usable cases are reported separately.
- The sampling method is named, together with its limits.
- Drop-outs are counted and their handling is justified.
The analysis then runs in two stages. First you describe the data with frequencies, means and measures of spread, then you test relationships and differences. Which procedures come into question hangs on the level of measurement of your questions and is set out in the piece on data analysis methods. The difference between the two stages is explained in the guide to descriptive and inferential statistics.
Common mistakes in a survey for a thesis
Most problems with a student survey do not arise while the questions are being written, but in the decisions before and after. Four of them come up so regularly that they deserve a section of their own.
What these four mistakes have in common is that they can barely be repaired afterward. Knowing them in advance saves you awkward questions when your committee reads the methods section.
The data collection starts before the research question
A questionnaire written without a sharpened research question asks a great deal and rarely hits the right thing. The variables are not derived from the question but from instinct, and at the analysis stage exactly the variable that would show the relationship is missing. So formulate the question and your hypotheses first and the items in the questionnaire after that.
The questionnaire promises anonymity and asks for an email address
A survey stops being anonymous the moment the same questionnaire asks for an email address for a gift card drawing. Every answer then hangs on an address, whatever the consent page says, and the study leaves the exempt category on the spot. Separating the two makes it clean: the thank-you page carries a second link to a form of its own in which only the address lands. The two datasets have no connection to each other afterward.
Answers that do not fit the picture disappear from the dataset
Deleting answers afterward because they contradict your hypothesis is research misconduct. The Federal Policy on Research Misconduct defines falsification as “manipulating research materials, equipment, or processes, or changing or omitting data or results such that the research is not accurately represented in the research record”, and the Office of Research Integrity adds that “the omission of data is considered falsification when it misleads the reader”. Something else is allowed: excluding cases that were clearly not answered seriously, where someone clicked a whole matrix identically or went through the questionnaire in 40 seconds. The difference lies in the criterion. It is set before you look at the results, applies to every case alike and appears with its justification in the methods section.
The collection period is not documented
Without start and end dates the data collection cannot be placed in context, and that is exactly what you will be asked about. A survey on exam preparation that ran over winter break produces different answers from the same survey in finals week. So note both dates on the day they fall and write down which part of the semester the collection sat in.
Conclusion
A survey for a thesis stands or falls on two decisions and both are taken before the first question is written: whether the research question calls for a survey at all and how many usable cases you need. Everything after that is craft and can be planned. Build the timeline backward from your submission date and write the details of the data collection down while it is running.
What binds you in the end are the requirements of your own program, not general advice. Departments differ on whether your advisor has to approve the questionnaire, which IRB route applies and how the dataset is documented in the appendix. Settle that before the collection period starts.
Where to go next
- Looking for the right question types? Open and closed questions compared
- Still short of participants? How to find survey participants
- Want to know what a survey costs? What a survey costs
Questionnaire ready, collection period set?
Unlimited questions, EU hosting and the data processing agreement belong to the free plan at empirio.ai.