If you want to report a percentage, like "62% of readers open every issue", you need about 385 responses for a margin of error of 5 points, and fewer if your whole audience is small. If you want to hear what people say in their own words, you need far fewer: a few dozen written answers usually show you the main things people say.
Most surveys run by a small team are the second kind, and get sized like the first. Here's how to tell which one you're running, and what number to aim for.
Two different jobs, two different numbers
A survey can do two jobs, and they need different sample sizes.
Measuring. You want a number you can trust: what share of people prefer A over B, or whether satisfaction went up since last year. Here the sample size decides how far your number could be from the truth.
Hearing. You want to know what people are dealing with, what they'd change, why they left. Here the question isn't how precise a percentage is. It's whether you've heard the main things people have to say.
Most small surveys mix both. Size the survey for the job you'll make decisions from.
If you're measuring: the 385 rule
The standard formula for a percentage is n = z² × p × (1 − p) ÷ e². At 95% confidence (z = 1.96), assuming a 50/50 split (p = 0.5, the safest guess) and a 5-point margin of error (e = 0.05), it comes to 384.16, which rounds up to 385 (calculate.co.nz).
That number assumes a very large group. If your whole audience is small, a finite population correction brings it down, because once you've heard from a big share of everyone, there's less left to be wrong about (MeasuringU). The corrected size is n₀ ÷ (1 + (n₀ − 1) ÷ N), where N is the size of your whole audience.
Worked out with that formula, at 95% confidence and a 5-point margin:
| People in your audience | Responses needed |
|---|---|
| 100 | 80 |
| 500 | 218 |
| 1,000 | 278 |
| 5,000 | 357 |
| 10,000 | 370 |
| 100,000 or more | 383 to 385 |
Two things stand out. Past a few thousand people, the number barely moves. And for a small audience, you need most of them. A community of 100 people needs 80 replies, so a survey there only measures well if almost everyone answers.
If you'll accept a 10-point margin, the base number drops to 97. For a rough read on a yes-or-no question, that's often fine. Just say "give or take 10 points" when you share it.
If you're hearing: far fewer
Written answers work differently. You're not estimating a percentage. You're collecting the different things people say, and the common ones show up early.
The best evidence comes from research on open questions in interviews. Weller and colleagues looked at how many people it takes before new answers stop turning up. When people were asked to list everything they could think of, 10 people were enough to produce 95% of the ideas that came up most often. Hearing nearly every possible answer took much longer: a median of 75 people, ranging from 15 to 194 (Weller et al., 2018).
That maps well onto surveys. The big, repeated answers arrive in the first few dozen replies. The long tail of one-off answers keeps growing for a long time, and most of it won't change what you do.
There's a practical ceiling too. Wikimedia's survey guidance warns that when most of your questions ask for written sentences, "100 respondents may actually be more data than you can realistically synthesize" (Wikimedia). Past that point, more replies only help if something is counting them for you.
A sensible target for a hearing survey is 30 to 50 written answers to your main open question. If the last 10 replies didn't add anything new, you have enough.
Plan for fewer written answers than finished surveys
Not everyone who finishes your survey writes in the open box. On Pew Research Center's panel, people skip closed questions 1% to 2% of the time, and open questions around 18% of the time on average. Questions that ask for several sentences get skipped more than ones that ask for a word or a phrase (Pew).
So if you need 40 written answers, aim for about 50 finished surveys.
Working backwards to how many people to ask
You don't control how many people answer. You control how many you ask. Divide the responses you need by the share of people you expect to reply.
Response rates vary a lot. Wikimedia's guidance says it's "often between 10% and 20% and could even be less" (Wikimedia). For newsletters, Inbox Collective suggests aiming for 2% to 5% of your audience (Inbox Collective).
The numbers in this example are invented to show the arithmetic. Say you write a newsletter with 4,000 subscribers and want to know what they'd like more of.
- At a 3% response rate, you'd get about 120 replies.
- That's far short of the 351 you'd need to report precise percentages for 4,000 people.
- It's plenty for hearing: about 100 written answers after skips, more than double the 40 you need.
So the honest way to run it is as a hearing survey. Ask one or two open questions, count what people said, and share it as "the most common thing you asked for was…" rather than "37% of readers want…".
A quick way to decide
Ask yourself what sentence you'll write when the survey closes.
- "Most of you want X" or "X came up more than anything else": hearing. Aim for 30 to 50 written answers.
- "62% of you want X": measuring. Use the table above.
- "More of you want X than last year": measuring, twice. You need the full number both years, asked the same way.
If you're not sure, run it as a hearing survey first. Pew Research Center does something similar: researchers sometimes run open questions in a pilot to find the most common answers, then turn those into the choices for a closed question later (Pew).
Where Enform fits
Enform is built for the hearing kind. It reads every reply, counts what people said, and every count opens to the replies behind it, so the long tail of one-off answers doesn't bury the big ones. Our guide on how to analyze open-ended survey responses walks through the counting by hand, and pricing is here. If you're still picking a tool, see survey tools for startups.
FAQ
Is 100 survey responses enough?
For hearing what people say in writing, usually yes. For reporting percentages, 100 replies gives you a margin of error of about 10 points from a large audience. That's fine for a rough read, and too loose to compare small differences.
How many responses do I need for a small business survey?
If a few hundred people buy from you, ask all of them. For measuring, you'd need most of them to reply (218 of 500). For hearing, 30 to 50 written answers is enough to show the main things people say.
What response rate should I expect?
It varies widely. Wikimedia's guidance says 10% to 20% is common and it can be lower. Inbox Collective suggests aiming for 2% to 5% of a newsletter audience.
How do I know when I have enough open-ended answers?
When the last 10 or so replies only repeat things you've already seen. Keep a running list of what people said and note when new entries stop appearing.
Sources
- Sample Size Calculator, calculate.co.nz
- Finite Population Correction, MeasuringU
- Open-ended interview questions and saturation, Weller et al., PLOS ONE (2018)
- Why do some open-ended survey questions result in higher item nonresponse rates than others?, Pew Research Center (2021)
- Three Questions Everyone Should Ask in a Survey, Inbox Collective (2023)
- Learning patterns/Who should we survey?, Wikimedia Meta-Wiki
- Writing Survey Questions, Pew Research Center