How many survey responses do you need before you can trust the results? For an organisation planning customer, tenant or member research, that question affects the budget, the timetable and what the findings can support.
There is no single survey sample size that suits every project. For a large population, around 385 completed responses is a common starting point for estimating a percentage with a margin of error of about ±5 percentage points at 95% confidence. That calculation assumes a simple random sample and an expected proportion of 50%. Your sampling method, reporting needs and audience may call for a different target.
What does survey sample size mean?
When planning fieldwork, distinguish the number of people you invite from the number whose answers you can analyse. You might invite 1,000 customers and receive 250 usable completed questionnaires. Your achieved sample for those results is 250. A question shown only to people who made a complaint could have a much smaller base, so always check how many people answered the particular question being reported.
Start by defining whose experience you want to understand. Customers who contacted you last month form a different population from everyone with an account. Write that definition into the brief before selecting records or calculating a response target.
What is a good sample size for a survey?
A useful starting point is to decide how precise the main result needs to be. For a large population, simple random sampling and a percentage near 50%, the approximate margins of error at 95% confidence are:
100 completed responses — ±9.8 percentage points.
200 completed responses — ±6.9 percentage points.
400 completed responses — ±4.9 percentage points.
These are illustrations of sampling uncertainty. The usual calculation changes for smaller populations, and weighting or a more complex sample design can change precision. A sample size calculator can help you explore assumptions, but a final target should reflect the actual research design. SurveyMonkey’s sample size guide explains the standard calculation and its inputs.
Plan for the comparisons you need to make
Suppose you want to compare five service areas. A total of 400 responses might leave one area with only 25 answers. Its percentage would be much less stable than the headline result, particularly if you then divide it again by age or contact method.
List the groups you need to report before fieldwork begins. Decide which comparisons matter enough to require their own response targets. If you need to detect a small change between years, the sample should be planned for that purpose too. Our guide to comparing survey results over time covers the other checks that make a comparison meaningful.
Check who your survey is reaching
A large response total can still leave important experiences missing. An email survey of recent customers, for example, may underrepresent people whose contact details are out of date. Repeating the same approach could bring in more answers while leaving that gap in place.
Review the contact list, the invitation process and who has responded. Consider whether another method could help people take part. A link shared openly on social media creates a different kind of sample from randomly selecting people from a customer list. As Pew Research Center explains, margins of error for opt-in samples depend on modelling assumptions, and sampling uncertainty sits alongside potential bias from nonresponse and coverage. Our article on reliable survey results explores these wider quality checks.
Work backwards to an invitation target
Once you have a response target, estimate how many eligible people you will need to contact. If you need 300 completed responses and expect one in four invited people to provide a usable completion, you would plan to invite about 1,200 people. That is a planning assumption, so use evidence from similar previous projects and review progress during fieldwork.
Check whether the list is large enough, whether everyone is eligible and whether contact details are usable. Allow time to reach groups with lower participation. Our practical tips for increasing survey response rates can help you prepare the invitation and questionnaire.
Can a small survey still be useful?
Yes, provided the conclusions fit the evidence. A small survey can highlight a confusing letter or an appointment problem to investigate. Report the actual response numbers and describe the limits of wider comparisons. For a small, clearly defined group, inviting everyone may be practical, although the people who reply may still differ from those who do not.
If the main aim is to test whether people understand the questions, a pilot survey serves a different purpose from estimating satisfaction across your whole customer base. Set the target around what that stage needs to establish.
Set a response target your research can support
Before launch, record the audience, the main measures, the groups you need to compare and the assumptions behind your target. Give someone responsibility for reviewing participation while there is still time to adjust fieldwork.
Viewpoint helps organisations plan and deliver research that supports decisions about their services. Explore our research services or contact the team to discuss the audience and response target for your next survey.