Process
How to run a stratified sample in six steps
A six-step process for running a stratified study: choose a stratifying variable that drives different answers, define strata that don't overlap, set the total sample size, allocate it proportionately or disproportionately, draw at random within each stratum and weight oversampled groups at analysis. It suits product and UX researchers sampling across plan tiers, regions or roles. Koji screener questions can put each participant in the right segment.
Where this appears
What's in this process
- Choose a variable that matters: Stratify by what is likely to drive different answers: plan tier, region, role or usage level.
- Define clean strata: Every person fits exactly one stratum, and together the strata cover the whole population.
- Set the total sample size: Base it on the confidence level and margin of error you need.
- Allocate the sample: Proportionate for the whole population, disproportionate for subgroup comparisons.
- Sample at random in each stratum: The random draw inside each group is what keeps it a probability sample.
- Weight oversampled strata: If you oversampled a small group, weight the results to get accurate whole-population figures.
Use this image
Free to use in your docs, decks and posts under CC BY 4.0, as long as you credit Koji with a link to this page. To embed it, paste this into your page:
html
<a href="https://www.koji.so/visuals/how-to-run-a-stratified-sample-steps"><img src="https://www.koji.so/visuals/how-to-run-a-stratified-sample-steps/how-to-run-a-stratified-sample-steps.png" alt="Process showing six steps for stratified sampling, from choosing a stratifying variable and defining strata to random draws and weighting." width="1080" height="1080" /></a>
<p>Source: <a href="https://www.koji.so/visuals/how-to-run-a-stratified-sample-steps">How to run a stratified sample in six steps</a>, Koji</p>
