Sample Size Calculator
How many survey responses do you need? Pick what you want to learn, and the calculator asks only for what matters.
Calculate your sample size: how many responses you actually need
How many people do you need to survey before the results mean anything? The honest answer: it depends on what you want to find out. Which is why this calculator starts with your goal, not with formulas.
Three goals, three calculations: estimate a share (what percentage of your customers would recommend you?), compare two results (is version A really more popular than B, has the score moved since last year?), or explore opinions (open questions, interviews). Pick the goal and only the inputs that matter for it appear.
From statistics to practice
The statistical minimum is only half the answer. Not everyone you invite responds: at a 20% response rate, 385 responses take nearly 2,000 invitations. And if you want to read results by segment (new vs. returning customers, DE vs. EN), you need the sample per segment, not in total. The calculator does both conversions for you.
Small population? If your audience is only a few hundred people, the required sample shrinks considerably, and at some point the tool simply tells you: just ask everyone.
Frequently asked questions
Why doesn't the calculator ask how many questions my survey has?
Because sample size doesn't depend on it. It depends on the goal: estimating a share, proving a difference, or collecting opinions. If your survey has several goals, calculate per goal and take the largest number.
What does margin of error mean?
The range around your result within which the true value sits, at your chosen confidence. If you measure 60% agreement at ±5 points, the real value is very probably between 55 and 65%. Halve the margin and the required sample roughly quadruples.
Aren't 100 responses always enough?
For a rough impression, maybe. For a claim at ±5 points and 95% confidence you need around 385 responses with a large audience, and comparing two groups quickly takes 400 per group. Making small effects visible is expensive, and that is exactly what the tool works out honestly.
What about interviews and open questions?
There it's not statistics but saturation: the point where new conversations stop bringing new insights. As a rule of thumb that's 15 to 30 people per segment you want to understand.