How to Determine Research Sample Size
Many factors determine the reliability of a study, but the most fundamental is the sample. How many people you ask, how you select them, and how well they represent your target population directly shapes the quality of your results. If your sample is too small, it becomes difficult to trust the results. If it's too large, budget and time are wasted. In this article, we explain the variables that affect sample size, how calculations work in practice, and common mistakes in the process.
What is a sample?
A sample is a subset selected to represent the entire target population (universe) you want to reach in your research.
The universe could be all consumers in a country, people who purchase a specific category, or a much narrower demographic segment.
Whatever the case, measuring the entire universe is rarely possible or necessary. A properly constructed sample can represent the universe with sufficient confidence, and this representational power enables the research to produce actionable results.
What determines sample size?
Although sample size is calculated with a single formula, the variables within that formula change from study to study. Each one pulls the number of participants you need in a different direction.
Confidence level
Expresses how reliable you want your results to be. Standard practice uses a 95% confidence level. Raising it to 99% is possible but requires more participants.
Margin of error
Shows how much results can deviate from the true value. In general market research, a 5% margin of error is common. For high-risk decisions like pricing or pre-launch tests, reducing it to 3% may be more appropriate.
Population size
The total size of your target audience. If the population is very large, its impact on sample size diminishes. In small populations, the sample-to-population ratio needs to be higher.
Expected variance
An estimate of how similar responses will be on the topic you're measuring. Without reference data from a previous study, maximum variance is typically assumed (p=0.5).
Subgroup analysis needs
If you want to evaluate results across demographic or behavioral breakdowns, each subgroup needs sufficient sample size on its own. As subgroup count increases, total sample need rises accordingly.
How is sample size calculated?
There's a formula side to this, and honestly it's not the most exciting formula:
n = (Z² × p × (1-p) × N) / (e² × (N-1) + Z² × p × (1-p))
- n = Required sample size
- Z = Z-value for confidence level (1.96 for 95%)
- p = Expected proportion (0.5 if unknown)
- N = Population size
- e = Margin of error
Most researchers in practice rely on online calculators or platform tools.
Still, knowing the formula isn't useless—understanding why the sample grows or shrinks when you change a parameter enables conscious decisions in research design.
Quick reference table
The table below shows minimum sample sizes for different population sizes, assuming 95% confidence level and 5% margin of error.
| Population size | Required sample |
|---|---|
| 500 | 217 |
| 1,000 | 278 |
| 5,000 | 357 |
| 10,000 | 370 |
| 50,000 | 381 |
| 100,000 | 383 |
| 1,000,000+ | 384 |
The most striking point: as the population grows, the required sample stabilizes. The difference between 100,000 and 1 million is just 1 participant. In practice, 600 participants is a more realistic starting point for most consumer research with breakdown analysis.
Sampling method selection
How you reach people matters as much as how many you reach. A sample of 1,000 collected with the wrong method can be less reliable than 300 collected with the right method.
Probability sampling
Approaches where every individual in the population has an equal or known chance of being selected. Simple random sampling, stratified sampling, and cluster sampling fall into this group. Ideal for statistical generalization.
Non-probability sampling
Convenience sampling, snowball sampling, and quota sampling fall into this category. The most commonly used method in online consumer panels is quota sampling, where demographic quotas are predetermined and data is collected until each quota is filled.
Quota sampling isn't statistically "random," but when properly designed and fed from a sufficiently large panel pool, it produces reliable results in practice. The critical factors: realistic quota definitions and sufficient participant diversity in the panel.
Common mistakes
The most common mistakes in sample planning and how to avoid them:
Planning sample only as total size
Going to field with 400 because the formula said 384 seems tempting, but with gender and 3 age groups you end up with 60-70 per cell. Start from 'how many in the smallest subgroup.'
Ignoring margin of error
'We asked 300 people' means nothing on its own. The real question is what the margin of error is and whether it's acceptable for your decision.
Not accounting for response rate
Assuming all will complete the survey isn't realistic. Due to screening questions, drop-offs, and quality filters, keep the starting pool larger than your target.
The 'more is always better' assumption
After a certain threshold, the marginal contribution of each added participant drops dramatically. Spending that budget on better targeting or question design is almost always more efficient.
Neglecting quota structure
Without quotas, most responses come from the easiest-to-reach demographic group. Quota structure is as critical a design decision as sample size and both need to be planned together.
How Sorbunu manages sampling
If you think you need to calculate and balance all these variables separately—good news: Sorbunu handles most of this for you.
When creating research on Sorbunu, sample size, target audience definition, screening questions, and quota structure are determined in a single flow. Price and estimated field time update instantly based on your selected parameters.
The real difference emerges after fieldwork begins. You can track quota progress live, and if a specific demographic group falls behind target, update quotas in real-time.
A pool of over 4 million verified consumers makes it easy to reach sufficient sample even in niche target audiences. Beyond standard demographic breakdowns, you can directly reach behavioral segments.
Frequently Asked Questions
The theoretical minimum for large populations with 95% confidence and 5% margin of error is 384 participants. In practice, most research requires analysis across demographic or behavioral breakdowns, and losses occur from screening and quality filters. For research with breakdown analysis, 600 participants is a more realistic starting point.
For seeing general trends within a single homogeneous group, 100 participants can be a starting point. However, it's insufficient for statistically powerful comparisons or low margin of error. For exploratory research or initial hypothesis tests, 100 may be reasonable; decision-making research generally requires higher numbers.
Generally directly proportional—more participants means higher field costs. However, this relationship isn't always linear. Reaching niche audiences can be more expensive than broad audiences. Screening questions also affect cost, as higher screening rates require contacting more people to reach your target.
Quota structure determines how the total sample is distributed across subgroups, and as the number of quotas increases, total need may also increase. For example, with 3 age groups and 2 genders, you have 6 cells. If you want at least 50 participants per cell, total sample shouldn't be less than 300.
This depends entirely on the panel's structure. Unverified and open-access panels may have data quality issues. However, panels with identity verification, behavioral cross-checks, participation frequency management, and real-time response quality control offer high reliability.
In traditional research models, this is difficult after fieldwork begins. In real-time platforms, adjusting quota and sample parameters while the field continues is possible. Checking confidence level adequacy from interim data and expanding or reducing the sample makes the research process more flexible and efficient.
Related Solutions
Product & Concept Testing
Test new product, packaging, or concept alternatives with consumer feedback.
Brand Awareness & Perception Research
Measure your brand's awareness and perception levels among your target audience.
Pricing Research
Determine consumers' price perception and willingness-to-pay thresholds.
Start Your Research with the Right Sample
Define your target audience on Sorbunu, automatically calculate your sample size, and go to field in minutes.