A population often contains distinct subgroups. Stratified sampling divides the population into those subgroups (strata) and then draws a sample from each. The method ensures that every important subgroup is represented, which can improve precision and allow separate analysis of each stratum.

Strata are defined by characteristics relevant to the research question—age, region, customer segment, or product line. Sample sizes within strata may be proportional to the size of the stratum or may be equal if comparison across strata is the priority. Within each stratum, selection is usually random.

Stratified sampling is more complex to administer than simple random sampling, but the gain in representativeness and efficiency often justifies the effort. It is widely used in opinion research, market research, and quality control when the population is known to be heterogeneous. The key is choosing strata that matter for the variables being studied.