Skip to content
Free public beta until December 4, 2026: every course is open, no card needed. After that, the first chapter of each course stays free and full access is a one-time $39 ($19 for up to 100 founding members chosen for helpful feedback). Your progress is kept. How the beta works

Which sampling design do I need?

The decision path from a study's constraints to the right sampling method.

Reach for this whenever a study needs to choose HOW to draw its sample -- ask what's known about the population's structure and what actually constrains the study (money, time, travel), in that order, before defaulting to whichever method sounds familiar.

Step 1 — must every subgroup be guaranteed representation?

See section →
A sample drawn by taking a separate SRS from within every stratum of the population.

Example

Guaranteeing every one of 50 states appears in a national survey.

Reach for this when a subgroup's membership is already known and its representation must be guaranteed.

Step 2 — is travel/logistics cost the binding constraint?

See section →
A sample drawn by randomly selecting a few whole groups and including every unit inside them.

Example

Surveying every household in 6 of 300 randomly chosen villages.

Reach for this when visiting every unit in a few whole groups is far cheaper than visiting a few units everywhere.

Step 3 — is there already an ordered list with no periodic pattern?

See section →
A sample drawn by taking every k-th unit after one random starting point.

Example

Pulling every 40th customer from a numbered list of 2,000.

Reach for this when a numbered list already exists and no hidden period matches the step size.

Step 4 — none of the above apply

See section →
A sample drawn so every possible group of that size has an equal chance of selection.

Example

Drawing 30 numbers at random from a numbered list of 200 employees.

The fair default when no grouping, travel cost, or existing order pushes toward another method.

Last resort, not a real choice

See section →
A sample of whichever units are easiest to reach, with no randomness.

Only when no genuine random design is achievable -- report its limits honestly.

See section →
A sample where units opt themselves in, like an online poll.

Self-selected respondents skew toward strong opinions, regardless of sample size.

Graph settings

Calculator