Quantitative Sample Size Calculator UX
Calculate sample size for surveys, A/B tests, usability testing and card sorting with academically validated formulas
Calculate your sample size
1. Select methodology
Choose the research methodology you will use
Cochran's Formula: Finite Population
Cochran's formula (Cochran, 1977) is the statistical standard for calculating sample size when you know the exact size of your population. It includes a finite population correction factor (FPC) that reduces the required sample, optimizing resources without losing precision.
n = (N * Z^2 * p * q) / ((N-1) * E^2 + Z^2 * p * q)When to use Cochran's formula?
- You know the exact size of your population (employees, registered users, clients)
- The population is less than 100,000 individuals
- Your sample represents more than 5% of the total population (5% rule)
- You need statistical rigor for publications or critical decisions
Cochran's formula parameters
- N
- Total population size
- Z
- Z-value for confidence level (1.96 for 95%)
- p
- Expected proportion (0.5 if unknown)
- q
- 1 - p (complement of the proportion)
- E
- Acceptable margin of error (e.g., 0.05 for +/-5%)
Practical example: finite population
You have 2,000 registered users and want to survey them with 95% confidence and +/-5% margin of error.
With Cochran's formula: n = (2,000 * 3.84 * 0.25) / (1,999 * 0.0025 + 3.84 * 0.25) = 323 participants. Without the finite correction you'd need 385.
Simplified Formula: Infinite Population
When the population is unknown or exceeds 100,000 individuals, the finite correction factor has less than 1% impact. In these cases, the simplified formula is used, which does not require knowing the population size.
n = (Z^2 * p * q) / E^2When to use the infinite population formula?
- You don't know the exact size of your population
- The population exceeds 100,000 individuals
- Your sample represents less than 5% of the total population
- Market research with broad or public audiences
Infinite population formula parameters
- Z
- Z-value for confidence level (1.96 for 95%)
- p
- Expected proportion (0.5 if unknown)
- q
- 1 - p (complement of the proportion)
- E
- Acceptable margin of error (e.g., 0.05 for +/-5%)
Practical example: infinite population
You want to survey users of an app with millions of downloads, with 95% confidence and +/-5% margin.
With the simplified formula: n = (3.84 * 0.25) / 0.0025 = 385 participants. This result is independent of the population size.
Power Formula: A/B Testing
An A/B test does not estimate a value in a population: it contrasts two variants to decide whether the observed difference is real or noise. That is why it uses neither Cochran's formula nor the simplified one — it uses power analysis (Kohavi et al., 2009; Cohen, 1988). Population size does not enter the calculation, and the number it returns is per variant: the full experiment needs twice as much.
n = 2 * (Za + Zb)^2 * p * (1-p) / D^2When to use the A/B testing power formula?
- You are comparing two or more variants and need to know whether the difference between them is real
- You know the baseline conversion rate of your control variant
- You can define the minimum effect that would justify making the change (MDE)
- Traffic accumulates over time: there is no closed population capping the sample
A/B testing formula parameters
- n
- Sample per variant (the full experiment requires 2n)
- Za
- Z-value for the confidence level (1.96 for 95%)
- Zb
- Z-value for statistical power (0.84 for 80%; 1.28 for 90%)
- p
- Baseline conversion rate, as a proportion (e.g., 0.05 for 5%)
- D
- Minimum detectable effect (MDE) as a proportion (e.g., 0.005 for +0.5 points)
Practical example: conversion A/B test
Your checkout converts at 5% and you want to detect a 10% relative improvement (that is, reaching 5.5%), with 95% confidence and 80% power.
The absolute MDE is 0.5 points = 0.005. With the power formula: n = 2 * (1.96 + 0.84)^2 * 0.05 * 0.95 / 0.005^2 = 29,792 per variant, 59,584 in total. At 2,000 daily visits, the test takes about 30 days.
When to use each formula?
First decide which question you are answering: estimating a value in a population (surveys) or detecting a difference between variants (A/B testing). Only within the survey path does the choice between finite and infinite depend on whether you know your population size and what proportion of it your sample will represent.
- Are you comparing two variants to know whether the difference between them is real? If yes -> use the A/B testing power formula; population size does not enter the calculation and the result is per variant.
- Do you know the exact size of your population? If not -> use the infinite formula.
- Does the population exceed 100,000? If yes -> use the infinite formula (the result is practically identical).
- Will your sample be more than 5% of the population? If yes -> use the finite formula (Cochran) to optimize resources.
- When in doubt -> use the infinite formula. It always gives an equal or larger sample, which is more conservative.
Quick comparison
| Criterion | Finite Population | Infinite Population | A/B Testing |
|---|---|---|---|
| Question it answers | What is the value of X in my population? | What is the value of X in my population? | Is the difference between A and B real? |
| What sets the size | Margin of error and confidence | Margin of error and confidence | MDE, confidence and power |
| Known population | Yes, required | Not necessary | Not applicable |
| Population size | < 100,000 | > 100,000 or unknown | Irrelevant: traffic accumulates |
| Sample vs population | > 5% of population | < 5% of population | Not applicable |
| Correction factor | Yes (reduces sample) | Not applicable | Not applicable |
| Typical result | Smaller sample (optimized) | Conservative sample | Sample per variant (total is 2n) |
| Use case | Companies, closed communities | General market, mass apps | Conversion experiments on web or product |
References
- Cochran, W. G. (1977). Sampling Techniques (3rd ed.). New York: John Wiley & Sons. wiley.com
- Kohavi, R., Longbotham, R., Sommerfield, D., & Henne, R. M. (2009). Controlled Experiments on the Web: Survey and Practical Guide. Data Mining and Knowledge Discovery, 18(1), 140–181. link.springer.com
- Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates. routledge.com
Detailed Guides by Methodology
Learn more about each methodology with examples, benchmarks, and best practices
Finite Population Survey
For known populations < 100,000 users
Infinite Population Survey
For large populations > 100,000 users
Problem Discovery
How many sessions to see a problem at least once
Quantitative Usability
Quantify usability problems (TSR, SUS)
Conversion A/B Testing
Sample per variant to detect a real difference
Quantitative Card Sorting
Statistical analysis of groupings and mental models
Need to design the complete study?
The calculator gives you the number; defining the methodology, recruiting well and holding the statistical rigor is the rest of the work. That part we do with you.