Table of Contents
If you are staring at a survey and wondering how many responses you actually need, you are asking the right question, and the answer is simpler than most stats textbooks make it sound. Survey sample size comes down to three plain ideas, a couple of reference numbers to memorize, and one honest truth about what matters more than hitting a magic total.
This guide is written for the stressed student version of you: the one with a deadline, a supervisor who wants "an adequate sample," and no desire to wade through formulas. We will keep the math light, give you numbers you can quote, and point you to a calculator when you need an exact figure.
Here is the one-line version if you only read this far. For a large population and normal confidence levels, aim for around 385 responses for a tight result, or around 100 if you can accept a looser one. Everything below explains where those numbers come from and how to adjust them for your study.
The two words that decide your sample size
Two ideas set your number: confidence level and margin of error. Learn these two and the rest is lookup.
Confidence level is how sure you want to be that your result is not a fluke. The standard in almost all student research is 95%. It means that if you repeated the study many times, about 95 out of 100 versions would land close to the true answer. You can go higher (99%), but it costs you a lot more responses for a small gain, so 95% is the sensible default.
Margin of error is the wiggle room around your result. If 60% of your sample prefers option A with a 5% margin of error, the real figure is somewhere between 55% and 65%. A smaller margin means a more precise answer, and it needs more responses. Most surveys use a 5% margin; a 10% margin is acceptable for exploratory or early-stage work.
How to read it in one sentence
"At 95% confidence with a 5% margin of error" means: I am 95% sure the true answer is within 5 points of what my sample showed. That is the phrasing your supervisor wants to see in your methods section.
The reference numbers to memorize
For a large population (roughly, more than about 20,000 people, or any population so big it barely affects the math) at 95% confidence, the required sample size is fixed and worth memorizing:
- 385 responses for a 5% margin of error
- 97 responses for a 10% margin of error
- 1,067 responses for a 3% margin of error
Notice the pattern: cutting your margin of error roughly in half (from 10% to 5%) roughly quadruples the responses you need, and tightening it further to 3% costs a lot more again. Precision is expensive. That is exactly why you should not chase a 3% margin for a class project when 5% or even 10% answers your research question.
For the exact figure for your own study, plug your numbers into the sample size calculator. It handles the confidence level, margin of error, and population size for you, so you never have to touch the formula.
Sample size by population size
The reference numbers above assume a big population. If your population is small (all 300 students in your department, all 80 employees at one company), you need fewer responses because a "finite population correction" kicks in. Here is a standard lookup for 95% confidence and a 5% margin of error.
Required sample size at 95% confidence, 5% margin of error
| Population size | Responses needed |
|---|---|
| 100 | 80 |
| 200 | 132 |
| 300 | 169 |
| 500 | 217 |
| 1,000 | 278 |
| 2,000 | 322 |
| 5,000 | 357 |
| 10,000 | 370 |
| 50,000 | 381 |
| 100,000+ | 385 |
Two things jump out. First, for small populations you need a surprisingly large share of them (80 out of 100 people). Second, past about 20,000 the number barely moves, which is why "385" is the famous figure. If your target population is huge, do not panic at the size; the requirement flattens out.
Rules of thumb for common student analyses
Sometimes the question is not "how precise is my percentage" but "do I have enough data to run my statistical test." For those, researchers lean on rules of thumb. These are heuristics, not laws. They are widely taught conventions, and your specific method or supervisor may set a different bar, so treat them as a floor to discuss, not a guarantee.
These are heuristics, not hard rules
The numbers below are common conventions for planning, not precise requirements. For a rigorous justification of the sample you need to detect a specific effect, run a proper power analysis instead. Always check what your own method and supervisor expect.
- Regression: a common heuristic is at least 10 to 15 participants per predictor variable. Five predictors would suggest 50 to 75 respondents as a rough minimum, more if you expect small effects.
- t-tests and group comparisons: around 30 participants per group is a rough convention for reasonably stable results. Small or subtle differences need substantially more.
- Correlations: many methods texts suggest something in the ballpark of 30 or more to get a stable estimate, with larger samples for weaker expected relationships.
- Factor analysis and surveys with many items: conventions vary widely, often stated as several participants per item or a few hundred total. Check the guidance for your specific technique.
When you are designing an experiment and want to move past rules of thumb, a power analysis tells you the sample size needed to detect an effect of a given size. Our statistical power calculator does this for you, and it is the more defensible way to justify your n in a dissertation.
What matters more than hitting a magic number
Here is the part textbooks bury. A representative sample of 200 beats a biased sample of 2,000 every time. Sample size controls precision; it does nothing for accuracy. If the people who answered are systematically different from the people you care about, a bigger sample just gives you a more confident wrong answer.
So spend at least as much energy on two things. First, representativeness: does your sample actually reflect the population you want to speak about? A survey only shared in one Facebook group speaks for that group, not for "students" in general. Second, response quality: rushed, inattentive, or contradictory responses pollute your data no matter how many you collect. Add an attention check, watch for people who finish impossibly fast, and read our guide on collecting high-quality survey data.
A bigger sample will not fix bias
If your recruitment method reaches the wrong people, collecting more of them makes the problem worse, not better. Fix who you are asking before you worry about how many.
What to do when you cannot reach your target
Real life gets in the way. The deadline arrives and you have 140 responses instead of the 385 you planned. You have good options, and none of them is faking data.
Report it honestly as a limitation. A methods section that states your achieved sample, the resulting wider margin of error, and how it constrains your conclusions is stronger, not weaker, than one that pretends everything was perfect. Supervisors respect honesty about limitations.
Adjust your analysis to match your data. With a smaller sample, report your findings as tentative, avoid slicing into tiny subgroups, and lean on descriptive statistics rather than tests that assume a lot of data. Do not run a ten-predictor regression on 60 people and present it as solid.
Or get more respondents. Often the fix is simply better distribution. A survey exchange like SurveySwap's free exchange lets you earn real responses by answering other people's surveys, which is a genuinely useful lifeline in the final week before a deadline. Our guide to finding dissertation survey respondents has more targeted tactics for academic audiences.
The takeaway: pick a confidence level (95%) and a margin of error you can live with, use the reference numbers or the calculator to set a target, and then spend your effort on reaching the right people and keeping their answers clean. That combination, not a big round number, is what makes your survey sample size defensible.
Frequently asked questions
How many survey respondents do I need for a dissertation?
It depends on your population and how precise you need to be. For a large population at 95% confidence, aim for around 385 for a 5% margin of error, or fewer if your target population is small (the lookup table above shows the exact figures). For hypothesis tests, a power analysis is the most defensible way to justify your number.
What is a good sample size for quantitative research?
There is no single "good" number; it flows from your confidence level, margin of error, and population size. As a common minimum, many student studies target around 100 to 385 responses. Use the sample size calculator for your exact figure, and remember that representativeness matters more than raw size.
Is 30 respondents enough for a survey?
For rough group comparisons, about 30 per group is a widely cited rule of thumb, but it is a heuristic, not a guarantee. Thirty is usually too few to estimate percentages precisely (the margin of error is large) or to detect small effects. Treat 30 as an absolute floor for a single small comparison, not a target for a full survey.
Does a bigger sample always give better results?
No. A bigger sample improves precision, but it does nothing to fix a biased sample. A representative sample of 200 is more trustworthy than a skewed sample of 2,000. Focus first on reaching the right people and collecting quality responses, then on hitting your target size.