The sample size you choose affects how reliable your results are, how quickly your survey fills, and how confidently you can analyse subgroups within your data.

Attest includes a built-in sample size calculator to help you make this decision based on statistical principles, while also taking practical constraints into account.

How to think about sample size

Before using the calculator, it’s worth considering three practical questions.

What subgroups do you need to analyse?

Even if your reporting focuses on the overall sample, you may want to analyse specific groups, such as age bands, genders, brand users, or regions. Your total sample needs to be large enough that these subgroups are meaningful.

Is your audience feasible?

Highly constrained audiences, niche targets, or smaller markets may not support very large sample sizes. Feasibility should always be checked alongside statistical ambition.

What is the purpose of the research?

Different research types require different levels of robustness. A brand tracker typically needs a larger sample than a quick concept test.

How the sample size calculator works

The calculator is available on the Audience page, at the top right of the page.

It uses three inputs:

Population size
Predefined population sizes are available for many markets. If your market is not listed, or if you have a more precise estimate, you can select Custom.
Margin of error
This defines how far your results may differ from the true population value. Smaller margins of error require larger sample sizes.
Confidence level
This indicates how confident you want to be that your results reflect the wider population. Higher confidence levels increase the required sample size.

Once you are happy with the calculated size, select Apply to survey to update your survey directly.

Understanding the statistics behind the calculator

The calculator follows standard market research conventions, using the formula below. Note that the value in the calculator is the finite size, calculated with the equations below and rounded up to the nearest whole number which is typical for this statistical calculation.

A z-score is the number of standard deviations (e.g. the variation) the value is from the mean. To find the right z-score, you can use the below figures for inputting to the formula:

These statistics form the basis of the calculator input behind the scenes, so you can easily see the output.

Sample size by research type

Sample size is not one-size-fits-all. Below are general guidelines rather than fixed rules.

Brand tracking

Larger samples are recommended to reduce natural variation over time.
A common starting point is around n=1,000, depending on feasibility.
Keep sample size consistent across waves where possible.

Concept and creative testing

Sample size depends on whether you use monadic or sequential monadic testing.
Where possible, monadic testing is recommended for more reliable comparisons.
A common guideline is around n=250 per cell, assuming feasibility allows.

Price testing

Sample size depends heavily on the analysis method being used.
Consider feasibility early, especially for niche audiences.

Consumer profiling

Larger samples are generally recommended, as useful subgroups often emerge during analysis rather than being predefined.
Robust overall sample size is more important than tight targeting.

Balancing robustness and feasibility

Larger samples increase statistical confidence, but they must be achievable.

Factors that can limit feasible sample size include:

narrow demographic targeting
multiple quotas
qualifying questions that screen out many respondents
smaller or less active markets

In these cases, it may be better to:

reduce sample size expectations
simplify targeting
focus analysis on higher-level trends rather than small subgroups

FAQ

Is a larger sample always better?

Not necessarily. A larger sample improves robustness, but only if it is feasible. Overly ambitious sample sizes can result in slow fill times or incomplete surveys.

Can I change the sample size after launching?

In many cases, you can increase sample size after launch. Reducing sample size is more limited and depends on survey status.

What is margin of error?

Margin of error describes how much your survey results may differ from the true value in the wider population.

For example, a margin of error of ±5 percent means that if 50 percent of respondents select an answer, the true population value is likely between 45 and 55 percent. Smaller margins of error require larger sample sizes.

What is confidence level?

Confidence level indicates how certain you want to be that your results reflect the wider population.

A 95 percent confidence level means that if you ran the same survey multiple times, you would expect the results to fall within the margin of error in 95 out of 100 cases. Higher confidence levels require larger sample sizes.