SpaceForms / Tools / Margin of Error Calculator
Margin of Error Calculator (Free, 2026)
Enter your sample size, confidence level and expected response proportion. The calculator instantly returns the ± margin of error and tells you whether your survey is precise enough to publish, decide on, or treat as directional only.
Total completed responses
Z = 1.96
Use 50% if unknown
Margin of error at 95% confidence
±4.9%
Good — standard for decision-making
Based on n = 400 and p = 50%
Formula: ME = Z × √(p(1−p)/n) × 100. Z is the z-score for your confidence level, p is the expected response proportion, n is your completed sample size. The result is the ± half-width of the confidence interval.
Margin of error at 95% confidence by sample size
Quick lookup for the most common confidence level (95%) assuming the conservative p = 50%. Use this to sanity-check whether your survey is large enough before you spend time fielding it.
| Sample size (n) | Margin of error | Use case |
|---|---|---|
| 50 | ±13.9% | Pilot test only |
| 100 | ±9.8% | Internal directional |
| 200 | ±6.9% | Team / segment read |
| 400 | ±4.9% | Standard product decision |
| 600 | ±4.0% | Customer base survey |
| 1,000 | ±3.1% | Publishable benchmark |
| 2,000 | ±2.2% | National poll-grade |
| 5,000 | ±1.4% | Tight subgroup analysis |
Assumes simple random sampling, p = 0.5. For small finite populations apply the Finite Population Correction.
How margin of error works (the intuition)
Margin of error quantifies sampling uncertainty. If 60% of your 400 respondents say they prefer Option A, the true population preference is very unlikely to be exactly 60%. With a ±4.9% margin at 95% confidence, the realistic range is 55.1% to 64.9% — and there's still a 5% chance the true value lies outside even that range.
Two facts surprise most teams. First, margin of error has almost nothing to do with population size once the population is above ~20,000. A national poll of the U.S. (population 335M) and a survey of a 50,000-person customer base both need roughly the same sample (~1,000) for a ±3.1% margin. Second, margin of error shrinks with the square root of n. Doubling the sample doesn't halve the margin — you need 4× the sample to cut margin in half.
That's why polling firms cluster around n = 800–1,200. It hits ±3% precision (the publishable standard) without the diminishing-returns cost of going to n = 5,000. The same logic should govern your in-app or email surveys.
When margin of error misleads you
- Non-probability samples. Email lists, in-app prompts and panel providers are not random samples. The formula gives you a number, but selection bias from who chooses to respond often dwarfs sampling error. Per Refiner's 2026 in-app benchmark, only 27.5% of in-app prompts get a response — the other 72.5% are not a random subset.
- Subgroup analysis. A ±5% margin on the full sample becomes ±15% or worse when you slice by region or persona. If your decision depends on a subgroup, calculate margin of error for that subgroup's n, not the whole survey.
- Comparing two proportions. Margin of error for a single proportion is not the right test for "is A better than B?". Use a statistical significance calculator for that — two-sample z-tests have different math.
- Likert and rating scales. Classical margin of error is built for proportions (yes/no, support/oppose). For mean scores on Likert scales the right precision metric is the standard error of the mean, not this formula.
Worked example
Scenario
You ran a CSAT survey to 8,000 customers and got 412 completed responses (a 5.2% response rate — below the Refiner 2026 average of 27.5%, but typical for cold email). 73% said they were "satisfied or very satisfied." Can you publish "73% of customers are satisfied"?
Calculation
- • n = 412, p = 0.73, confidence = 95% (Z = 1.96)
- • ME = 1.96 × √(0.73 × 0.27 / 412) × 100
- • ME = 1.96 × √(0.000478) × 100
- • ME = 1.96 × 0.02188 × 100 = ±4.29%
Interpretation
The true CSAT is somewhere between 68.7% and 77.3% with 95% confidence. You can publish "roughly 7 in 10 customers are satisfied" — but claiming a precise 73% would oversell the data. If you need ±2% precision, you'd need to roughly quadruple the sample to ~1,650 responses.
Field a survey with the right sample size
SpaceForms lets you build, distribute and track responses for free. Pre-built CSAT, NPS and custom templates — unlimited responses on the free tier so you can actually hit the sample size you need. No credit card.
FAQs
What's an acceptable margin of error for a survey?
±3–5% at 95% confidence is the standard for public-facing or decision-grade research. ±5–8% is acceptable for directional or internal insights. Above ±10% the data is generally too noisy to act on.
Why does the calculator default to 50% expected proportion?
p = 0.5 produces the widest possible margin of error for a given sample size. That makes it the safest (most conservative) default when you don't yet know how respondents will split. Lower or higher proportions always produce a narrower margin.
Do I need to correct for a small population?
Only if your population is small relative to your sample (e.g., surveying 200 of 500 employees). Apply the Finite Population Correction: ME_corrected = ME × √((N−n)/(N−1)). For populations above ~20,000 the correction is negligible.
Is margin of error the same as confidence interval?
The confidence interval is the full range (e.g., 68.7% to 77.3%). The margin of error is the ± half-width of that range (±4.3%). They describe the same uncertainty in different formats.
How does response rate affect margin of error?
It doesn't — directly. Only completed-response sample size enters the formula. But low response rates introduce non-response bias that the formula can't capture. A 5% response rate with n = 400 has the same calculated margin as a 50% response rate with n = 400, but the first is far less trustworthy.