Statistical sampling is the process of finding a representative portion of a large population. In quality control, a percentage of products will be selected at random for testing, and the results extrapolated to the population. For example, out of 100,000 cars produced, 30 might be crash-tested. The results will be assumed to hold for the entire population of cars actually sold.
In public opinion polls, samples must be carefully done in order to avoid several kinds of bias.