There is no single number of respondents that is sufficient for every survey in a bachelor's or master's thesis. Sample size should reflect the aim of the study, the recruitment method and the planned analysis. Collecting 100 responses alone does not mean that you can draw conclusions about all students, teachers or employees.
Have you prepared your questionnaire and want to know when to stop collecting responses? Start with what you want to establish from the data. Describing the experiences of people on one course calls for a different sample from comparing several groups. Below, you will find a way to organise these decisions and an example showing why the number of submitted questionnaires can be misleading.
Are 30, 50 or 100 survey responses enough?
None of these numbers is a universal threshold for a valid study. A lecturer's requirement may set a practical minimum for a particular assignment. It does not replace a methodological justification. If you are comparing groups, the number of people in each group matters too, not just the total number of responses.
Imagine two surveys with 100 respondents each. In the first, only members of a student research society respond; in the second, participants have been randomly selected from a list of a defined population. The numbers are equal, but the sampling methods and possible limitations on inference differ. In his paper on large samples, Mirosław Szreder (2022) explains why increasing the sample size does not automatically remove bias arising from non-probability sampling.
First, decide who you actually want to study
The population is the group to which the research question relates. The sample consists of the people included in the study. “Students” is usually too broad a term if the questionnaire only reaches friends on one degree programme.
An example of narrowing the question
Instead of: “How do students assess their work placements?”
More precisely: “How do third-year education students at the university being studied assess the organisation of their placements in the 2025/2026 academic year?”
The second version specifies the people, context and period. It is an original example of study planning, not a description of a study that has been carried out.
Write down the eligibility criteria before recruiting participants. These might include completing a particular stage of a degree, undertaking a placement or having experience in a specific job role. Then check whether your chosen recruitment channel allows you to invite those people. If your topic or aim is still unclear, first clarify the foundations of your methodology chapter.
Sampling: random selection and an open link are different situations
Szreder (2010) links the choice of sampling technique to factors including knowledge of the population, the intention to generalise findings and the practical resources available for the study. You should therefore define the recruitment method before looking for a sample size calculator.
Probability sampling
This uses a defined random selection mechanism. You need to establish the population you are sampling from and the probability of inclusion for its members. A chance encounter with someone in a corridor is not, in itself, a random sampling procedure.
Non-probability sampling
This may involve, for example, inviting people who are available or selecting participants who meet specific criteria. An open link also involves self-selection: the people who respond are those who saw the invitation and wanted to take part.
Describe what you actually did. If the survey was shared in online groups, do not call the sample random simply because you do not know the respondents. Nor should you claim that it is representative solely on the basis of the number of responses.
What do you want to calculate? This determines your sample size plan
Describing responses within the group you have surveyed
You can summarise how the people in your study responded. State the denominator used for percentages, any missing data and the limits of generalisation. “60% of study participants selected…” is a narrower claim than “60% of all students choose…”. The latter requires an appropriate basis for drawing conclusions about the population.
Estimating a parameter with a specified level of precision
Planning requires you to specify, among other things, the parameter to be estimated, the desired precision, the confidence level and the sampling design. A standard calculator for a population proportion does not automatically tell you the sample size needed for regression or a comparison of means. Its result is meaningful only alongside its assumptions.
Comparing groups or examining an association
Planning statistical power requires you to specify the intended test or model, a substantively meaningful effect, the chosen significance level and the target power. The division into groups and the data structure also matter. Do not choose these values solely to justify the number of people you have already recruited.
If your study will involve hypothesis testing, you can discuss the plan as part of the service “Statistical analysis for bachelor’s and master’s theses”. It is worth reviewing your questionnaire alongside the choice of calculations, before recruitment begins.
120 questionnaires, but only 6 people in one group: an example
A teaching example using fictional data. The form received 120 submissions. Eight did not meet the agreed eligibility criterion, and another five were duplicates. Excluding them left 107 responses.
The study was intended to compare three types of placement. The groups contained 78, 23 and 6 people respectively. The figure of 107 therefore does not reflect the conditions for each planned comparison. In addition, only 102 people answered one of the questions.
What does this mean? You need to distinguish the number of submissions, the sample after exclusions and the number of valid responses to a particular question. Exclusion criteria should be justified, documented and applied consistently. You must not remove responses because they make the expected result less favourable.
In this example, the feasibility of the comparisons needs to be reassessed. Possible options include further, justified recruitment or narrowing the scope of the analysis. Combining categories makes sense only where there is a substantive justification. It should not be done solely to obtain a favourable test result.
What should you do if you have too few responses?
- Identify where the problem lies. Too few people overall, a missing group and many skipped questions require different responses.
- Revisit your recruitment channels. Assess whether the invitation reaches people who meet the criteria. Do not change the population without documenting that decision.
- Check the questionnaire. Unclear questions or an overly long questionnaire can make participation difficult. Changing the instrument during the study requires an assessment of whether the data remain comparable.
- Set a realistic scope for your conclusions. A limited sample may call for a more cautious account and for some planned analyses to be dropped.
If you are still preparing your instrument, see the guide to designing a questionnaire for your thesis. If you have already collected the responses, keep the source file and record what actually happened during recruitment. There is no need to pretend that the study followed the original plan.
How should you describe your sample in the methodology chapter?
A good description allows the reader to trace the path from the research question to the data. Include the population, eligibility criteria, recruitment method and period, sample size justification, number of submissions, exclusions and final sample. Distinguish the plan from what actually happened if they differed.
Instead of stating “A sample of 100 people is sufficient”, explain what it is meant to be sufficient for and the basis for your assessment. If you did not formally plan the sample size, do not claim afterwards that you did. An honest account of a limitation tells the reader more than a justification that only appears sound.
Questions about sample size and collected data
How many people are needed for statistical analysis in a master's thesis?
There is no single number suitable for every thesis. The planned test, the number and size of groups, the expected effect size, and the required power and precision of the analysis are important. A small sample limits inference. It is best to discuss the sample size before starting data collection.
What if I already have research results?
Send the tool used, the assumptions, and information about the collected data. First, we will check what questions the data can answer. The methodology description should reflect the actual course of the research; missing information cannot always be supplemented without additional data collection.
Sources and scope of this guide
The Polish publications below underpin the distinction between sampling methods and the need for caution with large sample sizes. The study organisation examples and checklist are original teaching material.
- Szreder, M. (2010). Losowe i nielosowe próby w badaniach statystycznych. Przegląd Statystyczny, 57(4), 168-174. Full text: ps.stat.gov.pl/PS/2010/4/2010_57_4_168-174.pdf.
- Szreder, M. (2022). Szanse i iluzje dotyczące korzystania z dużych prób we wnioskowaniu statystycznym. Wiadomości Statystyczne. The Polish Statistician, 67(8), 1-16. DOI: 10.5604/01.3001.0015.9704.
