Start with the decision, not a number
There is no sample size that fits every survey. Before calculating a target, write down the decision the results will inform. Are you checking whether customers understand a new feature, estimating the share of an audience with a particular need, or comparing responses across groups? Each goal asks something different of the sample.
For an exploratory survey, the priority may be to uncover common themes and improve the questions you would ask next. For an estimate intended to represent a defined population, precision and recruitment matter much more. A useful planning statement is: “We need to estimate or compare this measure, for this population, closely enough to make this decision.” That statement gives the sample-size discussion a purpose.
Define who counts and how you will reach them
Specify the population before deciding how many responses to collect. “Customers” might mean everyone who has ever bought from you, people who purchased recently, or active users of one product. These groups can give different answers. Your survey invitation, eligibility questions and reporting should all use the same definition.
Next, consider how respondents will be selected. A randomly selected sample from a suitable list supports statistical estimates more readily than an open link shared on social media. A large convenience sample can still be useful for gathering feedback, but its size does not remove selection bias. People who see the invitation or choose to respond may differ from those you hope to describe. Sample-size formulas address random variation; they do not correct a mismatch between the respondents and the target population.
Choose a level of precision that matches the goal
When estimating a percentage from a probability sample, researchers often describe precision using a margin of error at a stated confidence level. As a rough planning example, estimating a percentage near 50% with a margin of error of about plus or minus five percentage points at 95% confidence calls for roughly 385 completed responses from a large population, under simple random sampling assumptions. Seeking about plus or minus three points raises that target to roughly 1,070. These are planning figures, not promises about the accuracy of any survey.
A smaller margin of error requires substantially more responses, so ask whether the extra precision would change your decision. If the aim is to choose between two clearly different options, a broad estimate may be enough. If a small difference would drive a costly change, you may need greater precision or a different research design. Also remember that percentages near 50% generally require larger samples for a given margin of error than percentages near the extremes.
For a small, clearly defined population, the number of completed surveys needed for a given margin of error may be lower than the large-population examples suggest. Even then, nonresponse deserves attention: the people who answer may not resemble those who do not. Treat any calculated target as one part of a quality plan, alongside question wording and recruitment.
Plan for segments and comparisons
An overall sample can look ample while leaving too few responses in the groups you care about. If the decision depends on differences between new and long-standing customers, for example, plan for enough usable responses in each group. The relevant number is not just the total; it is the count within each segment being reported.
Decide which comparisons matter before fielding the survey. Adding many segments after data collection can produce thin groups and tempting but unstable differences. When important groups are rare, you may need to recruit more of them deliberately. If you do, document the approach and be careful about interpreting the combined results as though everyone had the same chance of selection. Comparisons between groups also require more caution than a single overall estimate: a small gap in reported percentages is not automatically a meaningful difference.
Turn the target into a recruitment plan
Set a target for completed, eligible responses, then work backward to the number of invitations or contacts needed. Some invited people will not respond, and some responses may be incomplete or ineligible. A small pilot can help you check whether the survey is understandable and give you a better basis for planning recruitment, though its response rate may not carry over exactly to the full survey.
If time or access is limited, it is better to narrow the research question than to present a weak sample as definitive. You might focus on one audience, report broad patterns rather than fine-grained percentages, or use interviews to explore why people answer as they do. Record your target population, recruitment method, planned comparisons and final response counts. Those details help readers judge what the survey can support—and what remains uncertain.
