Stratified sampling

Stratified sampling is a method for selecting a representative sample from a population. It works by splitting the population into distinct, non-overlapping strata, or groups, based on shared characteristics. A random sample is then drawn from each stratum. This helps the final sample reflect the full population more accurately.

Example

A financial analyst studying company performance across industries might use stratified sampling to make sure the sample reflects the broader market. They would first divide companies into strata by industry sector, such as technology, healthcare, and finance. Then they would randomly select companies from each stratum for the sample. This creates a more balanced sample and supports more reliable conclusions about overall market performance and trends.

Another example comes from debt collections. When assigning overdue accounts to external collection agencies, a lender might use stratified sampling to build truly comparable groups. The lender could split accounts into strata based on factors such as days past due, balance size, product type, or prior payment behavior. Then they would randomly assign accounts from each stratum to different agencies. This helps make sure the overdue customers in each group are comparable, so performance differences are more likely to reflect agency results rather than sample mix.