It is used when the characteristics of the universe or population are already known, and this is heterogeneous and diverse. Homogeneous units of analysis are grouped together by strata, categories or classes (age, sex, income, academic level; always in line with the research topic). Within each of the strata, a random selection (simple random) is subsequently performed. It is used to improve the representativeness of the sample, and not to lose its “singularity” with respect to the total sample (Gaitán and Piñuel, 1998: 153). The strata or segments can include a wide range of variables: age, sex, religion, income level, media consumption… The homogeneity of the subgroup contributes to reducing the sampling error. It is more likely to be representative. Stratified sampling can be proportional, disproportionate, and controlled (or uniform). The composition of the final sample can be selected respecting the proportionality of the strata in the original universe (proportional) or each stratum has the same quota of representation (fixed), regardless of its distribution in the original universe.