statistics

Sample Size Calculator

Determining how many participants or observations your study requires is fundamental to valid research design. Too few data points and your results lack credibility; too many wastes resources. This calculator finds the optimal sample size by considering three inputs: your acceptable margin of error, desired confidence level, and expected proportion. Researchers, market analysts, and quality engineers use this approach to plan surveys, experiments, and audits with statistical rigor.

Last updated: April 19, 2026

Creators Anna Szczepanek, PhD Statistics

Reviewers Wojciech Sas and Davide Borchia

5,525 people find this calculator helpful

Understanding Sample Size in Research

Sample size represents the number of individual observations or responses you collect from a larger population. Getting this number right is essential because it directly affects whether your findings reflect genuine population characteristics or merely random noise.

The relationship between sample size and research quality works in both directions. A sample that is too small introduces excessive sampling error—your results may diverge wildly from the true population value. Conversely, collecting far more data than necessary squanders time, money, and participant goodwill without meaningfully improving accuracy beyond a certain threshold.

Three core statistical parameters drive sample size calculation:

Margin of Error: The acceptable range around your estimated result. A 5% margin means your true value could lie 5 percentage points higher or lower than what your sample shows.
Confidence Level: How certain you want to be that your results fall within the margin of error. 95% confidence is the research standard; 99% confidence requires larger samples.
Proportion Estimate: Your best prior guess about the characteristic you're measuring. If you have no prior information, 50% is the conservative default.

The Sample Size Formula

The foundational equation calculates the minimum sample size needed for a proportion-based study:

n = Z² × p × (1 − p) / ME²

For finite populations, apply the correction:

n_corrected = n / (1 + n / N)

n — Required sample size
Z — Z-score corresponding to your confidence level (1.96 for 95%, 2.576 for 99%)
p — Estimated proportion (between 0 and 1; use 0.5 if unknown
ME — Margin of error as a decimal (0.05 for ±5%)
N — Total population size (only needed if population is limited)

Practical Calculation Example

Suppose you're surveying university students about campus dining preferences. Your goals are a 95% confidence level, ±3% margin of error, and you estimate 60% of students regularly use on-campus facilities.

Applying the formula:

Z-score for 95% confidence = 1.96
p = 0.60, so (1 − p) = 0.40
ME = 0.03
n = (1.96)² × 0.60 × 0.40 / (0.03)² = 3.8416 × 0.24 / 0.0009 ≈ 1,025 students

You would need approximately 1,025 responses. If your university has 8,000 total students, applying the finite population correction gives a slightly smaller required sample. If it has 200,000 students, the correction is negligible—the sample remains close to 1,025.

Common Pitfalls When Determining Sample Size

Avoid these mistakes when planning your data collection:

Assuming 0.5 as your proportion estimate when you have prior data — Using 50% as your default is mathematically conservative but inefficient if past surveys, pilot studies, or external benchmarks suggest otherwise. If you have credible preliminary information that 70% exhibit the trait, use 0.70 instead—you'll reduce required sample size without sacrificing accuracy.
Forgetting the finite population correction for small populations — When your population is under 10,000, ignoring the correction overstates how many responses you actually need. A calculated sample of 600 from a population of 800 becomes roughly 400 after correction. This distinction grows more dramatic as population shrinks relative to sample size.
Conflating margin of error with confidence level — These are independent parameters. A 99% confidence level doesn't automatically provide a narrower margin of error—it requires a larger sample. You can have high confidence with a wide margin (relaxed precision) or lower confidence with tight precision. Match both to your study's practical needs.
Ignoring dropout and non-response rates — In practice, not everyone who agrees to participate completes the study. If you expect 20% attrition, multiply your calculated sample by 1.25. Surveys often see 30–50% non-response, so your actual recruitment target may be 1.5–2 times the theoretical minimum.

When Sample Size Matters Most

Sample size calculations are non-negotiable in certain high-stakes contexts. Clinical trials, where adverse events can affect human health, demand rigorous power analysis. Market research informing million-dollar product launches depends on defensible sample sizes. Quality control in manufacturing uses statistical sampling to ensure consistency across production batches.

Conversely, some exploratory research—qualitative interviews, usability testing, or preliminary concept validation—intentionally uses smaller samples to generate hypotheses rather than test them definitively. Recognizing which context you're in determines whether you need the calculator's precision or can work more flexibly.

For most academic research and commercial surveys, the standard benchmark is a 95% confidence level with a ±5% margin of error, yielding roughly 384 respondents for large populations. This balance has become conventional because it provides meaningful statistical rigor without requiring prohibitively large sample sizes.

Frequently Asked Questions

What's the minimum sample size to avoid unreliable results?

Samples below 30 observations are generally considered too small to produce trustworthy findings, particularly when studying large populations. With such a limited dataset, sampling variability dominates—your results could easily swing 10 percentage points or more from the true population value. However, 'minimum' depends on context. Qualitative research might intentionally use smaller samples; quantitative studies demand the calculator's approach. The threshold of 30 comes from statistical theory, but formal sample size calculations are always preferable to rough rules of thumb.

Why does using 50% as the proportion estimate give the largest sample size?

The proportion 0.5 (or 50%) represents maximum uncertainty. Mathematically, p × (1 − p) reaches its peak when p = 0.5, yielding 0.25. Any departure from 50% reduces this product—for example, 0.6 × 0.4 = 0.24, or 0.8 × 0.2 = 0.16. Since this term directly multiplies the numerator, larger products demand larger denominators (bigger samples) to achieve the same margin of error. Using 0.5 when your true estimate is, say, 0.8 'wastes' sample size. Conversely, if you're genuinely uncertain, 0.5 is the most defensible conservative choice.

How does increasing confidence level from 95% to 99% affect sample size?

The Z-score jumps from 1.96 to 2.576, and since Z is squared in the formula, sample size increases by a factor of (2.576 / 1.96)² ≈ 1.73. In practical terms, moving from 95% to 99% confidence requires roughly 73% more respondents. This dramatic increase is why 95% has become the standard: it offers strong confidence without unrealistic sample demands. Some fields like pharmaceutical research do use 99% confidence, but they budget accordingly. Recognize this trade-off when your stakeholders demand higher confidence—the cost in resources is substantial.

What is finite population correction and when should I apply it?

The finite population correction (FPC) adjusts your calculated sample when the population is limited enough that sampling a large proportion of it would reduce variability. The correction formula is n_adj = n / (1 + n/N), where N is total population. Apply it when your sample size exceeds 5% of the population. For example, if you calculated a need for 300 responses from a population of 2,000, you're sampling 15%—definitely apply FPC, reducing your requirement to roughly 230. For populations over 10,000, the correction becomes negligible and can be ignored. This prevents oversampling small populations.

How does sample size relate to confidence intervals?

Sample size and confidence interval width are inversely related: doubling your sample size narrows your confidence interval by approximately 30%. A small sample yields a wide interval, reflecting high uncertainty about where the true population value lies. A large sample yields a narrow interval, pinpointing the estimate more precisely. Confidence level sets the interval's coverage probability (95% or 99%), while sample size controls its width. You're using this calculator to choose a sample size that achieves both a desired confidence level and an acceptable interval width (margin of error).

Why might I need a larger sample than the calculator suggests in the real world?

The calculator assumes perfect data collection: every selected participant responds, all answers are accurate, and no data is lost. Reality introduces friction. Response rates for surveys typically range from 20% to 50%; phone surveys might see 10% completion. Participants may provide incomplete or inconsistent answers. Online panels experience dropout. Clinical trials lose patients to side effects or life changes. A pragmatic approach multiplies the calculated sample by an adjustment factor—1.5 to 2.0 for low-response scenarios, or 1.2 for controlled settings like in-person interviews. This ensures you still reach your target effective sample size despite inevitable losses.

More statistics calculators (see all)

Quartic Regression Calculator Poisson Distribution Calculator Accuracy Calculator Two Envelopes Paradox Calculator Median Absolute Deviation Calculator Coefficient of Variation Calculator Rayleigh Distribution Calculator False Positive Paradox Calculator