Random sampling uses a chance mechanism to select units from a defined population or sampling frame. It replaces a researcher’s choice of convenient or interesting cases with a specified selection process. That process matters because it provides a basis for statistical inference when the sampling design and subsequent analysis are handled appropriately.

In simple random sampling, every possible sample of the chosen size has the same chance of selection. Other probability designs use different arrangements, and selection probabilities need not always be equal. “We asked people at random” is therefore an incomplete description unless the method explains how selection actually happened.

The list determines who has a chance

Suppose a fictional software company randomly selects accounts from its current customer list. The draw can be properly random while excluding former customers, people using an unregistered version and prospective buyers. It represents selection from that list, not from everyone who might use the product.

Check the frame before drawing the sample. Remove duplicates appropriately, establish eligibility and identify coverage gaps. Decide whether the unit is a person, an account, a household or something else. Selecting accounts and then treating every user within them as independently sampled can misrepresent the design.

The Australian Bureau of Statistics explains probability sampling in terms of known selection chances. That requirement is what separates a probability sample from a haphazard collection of available respondents.

Choose a design suited to the population

Simple random sampling can work with a suitable list. Stratified sampling divides the population into defined groups and samples within them, which can support subgroup coverage and precision when designed well. Cluster or multistage designs sample groups or stages of units and require analysis that accounts for those relationships.

Systematic sampling selects at regular intervals after a random start. Its suitability depends partly on how the list is ordered; periodic patterns can interact with the interval. These methods are related, but their uncertainty calculations should not be treated as interchangeable.

In the software example, the team might sample separately within plan tiers to obtain enough information about a small enterprise group. If selection rates differ, estimating an overall customer result requires appropriate treatment of those rates rather than simply averaging all replies together.

Separate selection from participation and assignment

A randomly selected invitation list does not guarantee a randomly responding group. Non-response bias remains possible when willingness or ability to participate relates to the outcome. Track participation and examine relevant differences where data permit.

Random sampling is also distinct from random assignment in an A/B test. Sampling concerns who enters the study; assignment concerns which condition they receive. An experiment can assign a convenience sample randomly without thereby representing an entire market.

For exploratory interviews, purposive sampling may be more useful because the question requires particular experience or variation. Random sampling is a method with a specific inferential role, not a general badge of rigour that every research activity must wear.