What Sample Size Do You Need for a Survey? A Guide

"How many people do I need to survey?" is one of the most common and most misunderstood questions in research. Survey too few people and your results are shaky; survey far more than necessary and you waste time and budget. The encouraging truth is that you don't need to be a statistician to make a sound decision. This guide builds the intuition with plain language and one simple example.

Population vs. Sample

Your population is everyone you want to draw conclusions about, for example all customers of your store or all students at a university. Your sample is the subset who actually respond. The whole point of sampling is to learn about the large population by talking to a manageable number of people, as long as that sample fairly represents the whole.

A key insight surprises many people: beyond a certain point, what matters is the size of your sample, not the percentage of the population it represents. A well-chosen sample of around a thousand can describe a city or an entire country with similar precision, because precision depends mostly on the count of responses, not the size of the population behind them.

Two Ideas That Drive Sample Size

Confidence Level

The confidence level expresses how sure you want to be that your sample reflects reality. A 95% confidence level is the common default. Loosely, it means that if you repeated the survey many times, the results would land close to the true value in about 95 out of 100 of those surveys. Higher confidence requires a larger sample.

Margin of Error

The margin of error is the cushion around your result. If 60% of respondents prefer option A with a margin of error of plus or minus 4%, the true figure for the whole population is likely between 56% and 64%. A smaller margin of error means more precision, and that also demands a larger sample.

These two ideas work together. A typical professional setup is 95% confidence with a margin of error around 3% to 5%. Tightening either one increases the number of responses you need.

Practical Rules of Thumb

You don't have to calculate from scratch every time. These approximate guides assume a 95% confidence level and a fairly large population:

• A margin of error of about plus or minus 5% needs roughly 385 responses.
• A tighter margin of about plus or minus 3% needs roughly 1,000 responses.
• A margin of about plus or minus 10%, suitable for a rough read, needs only around 100 responses.

Notice the pattern: cutting the margin of error in half requires roughly four times as many responses, so precision gets expensive quickly. There is also a point of diminishing returns, where adding hundreds more responses barely narrows the margin.

A Simple Worked Example

Imagine you run an online shop with 50,000 customers and you want to estimate satisfaction at 95% confidence with a margin of error of plus or minus 5%. The rules of thumb point to roughly 385 completed responses.

Now here's the part people forget. If you expect only about a quarter of invited people to actually respond, you must invite far more than 385. To collect 385 completed responses at a 25% response rate, you would invite around 1,540 customers. Always plan your invitation list around your expected response rate, not just your target sample.

Smaller Populations and Subgroups

If your population is genuinely small, say 200 employees, you don't need a huge sample; you can survey a large share of them and your required number drops. On the other hand, if you plan to compare subgroups, for example responses by department or by age band, you need enough responses within each subgroup, not just overall. Slicing your data thins each group, so plan for that up front.

Don't Forget Sample Quality

A large sample that is skewed beats nothing, but it can still mislead. If only your happiest customers reply, no sample size fixes that bias. Aim for responses that represent your population across the dimensions you care about, and be honest in your reporting about who actually answered.

Bringing It Together

Pick a confidence level, usually 95%, decide how much margin of error you can tolerate, use a rule of thumb to set a target, and then inflate your invitation list to account for non-response. That sequence will serve most surveys well without any heavy mathematics.

With MSN Forms, real-time analytics and charts let you watch responses accumulate, so you can see when you've reached your target sample and stop collecting with confidence.