To calculate survey sample size, decide your confidence level (usually 95%), your acceptable margin of error (often ±5%), and your population size, then plug them into the sample-size formula. For most studies, this lands somewhere between roughly 350 and 400 completes for a large population at 95% confidence and a ±5% margin of error.

The three inputs you need

  • Confidence level — how sure you want to be that the true value falls within your margin of error. 95% is standard (z-score 1.96); 90% (z = 1.645) and 99% (z = 2.576) are also common.
  • Margin of error — the ± range around your result you are willing to tolerate, expressed as a proportion (±5% = 0.05). Smaller margins need bigger samples.
  • Population size — how many people are in the group you are studying. Above ~20,000 the population barely affects the answer.

A fourth input, the response proportion p, is set to 0.5 by default because that assumption produces the largest (most conservative) required sample.

The formula

First compute the sample size for an effectively infinite population:

n₀ = (z² × p × (1 − p)) ÷ e²

where z is the z-score for your confidence level, p is the expected proportion (use 0.5 if unknown), and e is the margin of error as a decimal. If your population N is finite, apply the correction:

n = n₀ ÷ (1 + ((n₀ − 1) ÷ N))

Worked example

You want 95% confidence (z = 1.96), a ±5% margin of error (e = 0.05), and you assume p = 0.5.

Step 1 — n₀ = (1.96² × 0.5 × 0.5) ÷ 0.05² = (3.8416 × 0.25) ÷ 0.0025 = 0.9604 ÷ 0.0025 = 384 (rounded up).

Step 2 — if your population is 10,000, apply the correction: n = 384 ÷ (1 + ((384 − 1) ÷ 10,000)) = 384 ÷ 1.0383 ≈ 370.

So you would aim for about 370-384 completed responses.

Rules of thumb

  • ~384 completes gives you ±5% at 95% confidence for any large population.
  • To halve your margin of error you must roughly quadruple your sample (±2.5% needs ~1,500).
  • For segment-level analysis, each subgroup you want to read on its own needs its own adequate sample, so plan for ~100+ per key segment.
  • Below ~30 responses in a group, report findings as directional only.
  • Always field a few more completes than the formula requires to absorb dropouts and quality removals.

How to do this in SegmentOS

Once you know your target sample size, use the SegmentOS audience panel to buy that many completes. In the audience calculator you set the number of responses you need and your targeting (region, age, demographics), and SegmentOS shows a transparent cost per complete, a total budget, and a feasibility estimate for whether your target is reachable. If you are segmenting heavily, increase your buy so each subgroup you plan to analyze with segmentation has enough responses to be reliable.

Frequently asked questions

What sample size do I need for 95% confidence and a 5% margin of error?

About 384 completed responses for a large population. If your population is small (say a few thousand), the finite-population correction lowers that number somewhat, but 384 is the safe default to plan around.

Does population size matter for sample size?

Only for small populations. Above roughly 20,000 people the required sample barely changes, which is why a national survey and a global survey need similar sample sizes. For small populations, the finite-population correction reduces the number you need.

Why is p set to 0.5 in the formula?

The term p × (1 − p) is largest when p = 0.5, so using 0.5 yields the most conservative (largest) sample size. If you have a strong prior that the proportion is far from 50%, a smaller sample can achieve the same margin of error.

How big a sample do I need to compare segments?

Each segment you want to analyze on its own needs its own adequate sample, not just the total. A common target is at least 100 responses per key segment, and you should treat any group under about 30 responses as directional only.