Sampling

The process of creating a sample that correctly reflects the makeup of the whole population. In market research, polling a whole target group or market is almost always cost-inefficient or impossible. Therefore a smaller sample is identified. Theoretically, the surveyed group would be representative of the total target market if it was selected perfectly randomly. However this is almost always not possible in market research, especially in many Asian settings. Some selected respondants may not be available, or simple as it may seem, giving each member of a target group an equal chance of being selected is not only time consuming but practically impossible.

In market research, most samples are chosen by stratified sampling, meaning that a sample is selected according to various criteria relating to variables that apriori would be expected to cause a difference in responses. These may include demographic factors such as gender, marital status, income, residence, ethnicity, language, or known consumer behaviour factors (e.g. heavy users vs. light users). This is a less perfect solution than random sampling, but if done correctly can be used to ensure that the sample can be generalised to the whole target group with some measure of confidence. However, most agencies use the term "probability sample" to indicate samples selected on the basis of probablilty of occurance or "incidence" of the target groups in the population a a whole.

Various formulaes can be applied by market research companies to estimate how large a sample needs to be to be generalised reliably to the whole population.