How to choose the right sample size for your survey

One of the most common questions before launching a survey is "how many responses do I need?" The right answer depends on how large the population you're measuring is and how much margin of error you're willing to accept. It's not a fixed number that applies everywhere.

Population size matters less than it seems

It's counterintuitive, but the sample size needed for a given margin of error barely changes between a population of 10,000 people and one of 10 million. What changes a lot is the margin of error between 50 and 500 responses. That's why, for most business surveys, the relevant question isn't "what percentage of my base responded" but "how many total responses do I have."

Quick reference numbers

At a 95% confidence level, the usual standard:

• ~100 responses: margin of error of ±10 percentage points. Fine for very general trends, not for fine-grained decisions.
• ~385 responses: margin of error of ±5 points. The most common standard for satisfaction and product surveys.
• ~1,065 responses: margin of error of ±3 points. Recommended if you're making important decisions or publishing the result externally.

When you plan to segment, you need more than you think

If you plan to cross-tab results by segment (plan, region, tenure), what matters is the sample size of each individual segment, not the total. 400 total responses sound solid, but if that's split across 4 segments of 100, each segment carries the margin of error of a 100-person sample, not a 400-person one.

When it's not worth worrying so much about

If the survey's goal is exploratory (understanding general themes, generating hypotheses, finding recurring pain points in open-ended comments), statistical precision matters less. For that kind of survey, 30 to 50 well-chosen responses can give you more value than 500 rushed, unfocused ones.